ci -NON-MEDICAL APPLICATIONS OF IMAGING TECHNIQUES: MULT! -DIMENSIONAL NMR IMAGING by V A S A N T H A K U M A R R A J A N A Y A G A M B.Sc. (Hons.), University of Peradeniya, Peradeniya, Sri Lanka, 1980 THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Department of Chemistry) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA November, 1986 © Vasanthakumar Rajanayagam, 1986 5 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of £ H£M t£TR j The University of British Columbia 1956 Main Mall Vancouver, Canada V6T 1Y3 Date J V W 2 3, SUPERVISOR : PROFESSOR LAURANCE D. HALL ii ABSTRACT The work described in this thesis concentrates on two aspects of Proton NMR imaging: development and evaluation of new/old experimental sequences and application of those techniques to study some non-medical systems that are of industrial importance. Two-dimensional Fourier transform spin warp imaging technique has been evaluated. Importantly, the adaptation of a conventional high resolution spectrometer to perform imaging has been demonstrated with means of "phantoms". This includes calibration of magnetic field gradients, mapping the static magnetic field and radiofrequency field distributions and intensity measurements related to proton spin densities. In addition, a preliminary study describes microscopic imaging of glass capillary tube phantoms containing water. Several different sequences related to Chemical Shift Jmaging including the one developed during the study have been described. A brief insight into chemical shift artifacts as well as some experimental methods of minimizing some of them have also been presented. The potential of NMR imaging to study non-medical systems has been explored in three different areas of interest: Chromatography columns. Porous rock samples and Wood samples. A variety of NMR imaging sequences have been used to study some interesting and challenging features of these systems which clearly extends the scope of NMR imaging science. ACKNOWLEDGEMENTS I wish to thank my research director, Professor L.D. Hall , for his guidance and constant encouragement to work in this new area of science which made the current work interesting. It is also a great pleasure to thank Dr. S. Sukumar for his help and for many useful hints on various aspects of NMR imaging. I would also like to acknowledge my gratitude to Mr. S.L. Talagala for many helpful discussions during the course of this work. While it is impossible to thank all those who contributed to the complet ion of this thesis, I am truly indebted to my collaborators; Mr. S.D. Luck for the work on NMR Microscopy, Ms . W.A. Stewart for obtaining the wood images on the whole -body NMR scanner, Drs. C. Hall and P.R. Steiner for providing the samples and proof reading parts of this thesis and to Dr. P.A. Salisbury for performing the mycological tests. I would also like to acknowledge the great help of Drs. L.G. Harrison and N.E. Burlinson for proof reading this thesis and to T . Marcus (electronics shop) and C. Neale (mechanical shop) for their expert technical contributions. Special thanks are also extended to Ms . Dawning Fung for her kind support and encouragement. Finally, I wish to thank UBC for granting me a University Graduate Fel lowship (1984-1986) and to all my friends who made my stay at UBC enjoyable. iv TABLE OF CONTENTS Page ABSTRACT ii ACKNOWLEDGEMENTS iii TABLE OF CONTENTS iv LIST OF FIGURES viii LIST OF TABLES xvii i LIST OF ABBREVIATIONS xix CHAPTER 1 - GENERAL INTRODUCTION 1 1.1 Background 2 1.2 Basic Pulse FT NMR Experiment 6 1.3 Effect of Magnetic Field Gradients 10 1.4 C lass i f icat ion of Imaging Techniques 12 1.5 Organization of this Thesis 14 CHAPTER 2 - TWO-DIMENSIONAL FOURIER TRANSFORM IMAGING 17 2.1 Introduction 18 2.2 Fourier Zeugmatography 20 2.3 Spin Warp Imaging 23 2.3.1 Ana lys is of the Technique 25 2.3.2 Phase Encoding Spatial Information 27 2.3.3 Spin Precession .-. 29 2.3.4 Choice of Parameters 31 2.3.5 Data Processing 34 2.4 Studies of Phantoms 42 2.4.1 Magnetic Field Gradients 43 2.4.2 Static Field B 0 Homogeneity 48 2.4.3 Radiofrequency Field B , Homogeneity 57 2.4.4 Miscellaneous Topics 63 2.4.4.1 Proton Density Measurements 63 2.4.4.2 NMR Microscopy 67 CHAPTER 3 - MULTI -DIMENSIONAL CHEMICAL SHIFT IMAGING 72 3.1 Introduction 73 3.2 Imaging Sequences 76 3.2.1 Two-Dimensional Chemical Shift Imaging - x,6 77 3.2.2 Two-Dimens ional Chemical Shift Imaging - y / 5 80 3.2.3 Three-Dimensional Chemical Shift Imaging - x,y,8 85 3.3 Four-Dimensional Chemical Shift Imaging - x,y,z,5 95 3.4 Chemical Shift Art i facts 102 CHAPTER 4 - NON-MEDICAL APPLICATIONS OF NMR IMAGING 111 4.1 Background 112 4.2 Appl icat ions to Chemistry 113 4.2.1 Introduction , 113 4.2.2 Metal Chelate Af f in i t y Chromatography 114 4.2.3 The Imaging Method 115 4.2.4 Results and Discussion 117 4.3 Appl icat ions to Material Science 120 4.3.1 Introduction 121 4.3.2 Oil Recovery Methods 121 4.3.3 The Imaging Methods 124 4.3.4 Results and Discussion 125 4.3.4.1 Water Relaxation 125 4.3.4.2 Imaging of Water 127 4.3.4.3 Imaging of Oil/Water 134 4.4 Appl icat ions to Forest Products 142 4.4.1 Introduction 142 4.4.2 Wood Growth 143 4.4.3 Water in Wood 145 4.4.4 Defects in Wood 146 4.4.5 The Imaging Methods 148 4.4.6 Results and Discussion 149 4.4.6.1 Internal Structural Features 149 4.4.6.2 Decay in Wood 156 4.4.6.3 Drying of Wood 166 4.4.6.4 Impregnation of Wood 174 4.4.6.5 Specialty Products f rom Wood 178 CHAPTER 5 - S U M M A R Y AND CONCLUSIONS 182 5.1 Summary and Conclusions 183 vii CHAPTER 6 - EXPERIMENTAL 189 6.1 The Imaging Spectrometers 190 6.1.1 Narrow Bore System 190 6.1.2 Wide Bore System 191 6.2 Setting up of an Experiment 193 6.3 Samples 195 BIBILIOGRAPHY 198 APPENDIX 1 212 APPENDIX 2 213 LIST OF FIGURES VIII Figure Page Chapter 1 1.1 Projection Reconstruction imaging. From a series of one-d imensional projections, obtained by acquiring the NMR signal in the presence of magnetic f ield gradients directed along different orientations relative to the sample, a two-d imensional NMR image of the sample can be reconstructed. The method is analogous to the one used in X - r a y CT. [From reference (16)] 1.2 The macroscopic magnetization vector M of a set of equivalent nuclei precesses about the magnetic f ield B 0 at its characteristic Larmor frequency. At any instant, M can be represented by its components M x , My and M z along the three directions x, y and z respectively. 1.3 The absorption mode (a) and the dispersion mode (b) lineshapes. 1.4 Scheme of c lassif icat ion of the various NMR imaging techniques. [From reference (30)] 13 Chapter 2 2.1 Partitioning of the timing axis in 2D FT NMR experiments. 19 2.2 Two-d imensional Fourier Zeugmatography. During the evolution period tu gradient G x spatially encodes the x dimension. S imi lar ly , gradient Gy during the detection period encodes y dimension. FID's are recorded in t 2 for different t1 values, and Fourier transformed in two dimensions to yield the image. 21 2.3 Two-d imens ional Spin Warp imaging. A variable amplitude Gy pulse is applied during the evolution period and the spin echo signal is detected. The spin echo is formed by gradient reversal in (A) and by 180° refocussing rf pulse in (B). In either case, the sequence is repeated for different values of Gy, and the image obtained by double Fourier transformation. 24 ix 2.4 The spins are prepared by a 90° pulse init ial ly. 28 Spatial encoding is accomplished, first by the application of the phase encoding gradient Gy and then by the frequency encoding gradient G x . Note that the y spatial information is encoded as phase differences. 2.5 Rotating reference frame representation of the spin 30 echo sequences described in figure 2.3. 2.6 Diagrammatic representation of the effect of 33 spatial resolution on 2D images obtained at 80 MHz. 2.7 (a) Pattern of the spin echo signal as n is 36 stepped through, in the different blocks of the data set S ( G v , t 2 ) . (b) Fourier transformed spectra S ( G y , F 2 ) . 2.8 (a) The pre-acquisi t ion information are reflected as 37 phase variations in the echo-interferogramsS(F2,Gy). (b) Fourier transformed spectra with magnitude calculation S(F2,y). 2.9 (A) Stacked plot display of the 4 cm diameter 40 spherical glass bulb containing C u ^ + doped water. (B) Contour plot display (8 levels) of the same. (C) Colorgraphic display of the same with a 8 level pseudo blue scale. 2.10 Evaluation of the imaging gradients on the 1.89 T 45 wide bore system using a phantom comprising three, 5 mm NMR tubes containing water and mounted 15 mm apart. 2.11 Evaluation on the 6.35 T narrow bore system 46 using a phantom mounted inside a 10 mm NMR tube. The proton decoupler coi l (15 mm diameter, 15 mm height) in the 1 ^C- { 1 H} probe was used as the observe co i l . The phantom used was assembled from five glass capillary tubes (1.6 mm id) f i l led with water and held in posit ion by a Teflon plug. 2.12 The pulse sequence used to measure the magnetic 50 f ield distribution. The evolution of spins during interval t l f after initial excitation by a 90° pulse, is determined by the static field homogeneit ies, since no gradients are applied. The rest of the sequence is similar to the Fourier Zeugmatography sequence. [From reference (66)] X 2.13 The phantom (A) is machined from a Teflon plug 51 with 1.2 mm diameter holes spaced every 2.3 mm along its length, and f i l led with acetone. This set -up is mounted inside a 10 mm NMR tube and placed inside the narrow bore system. The one-dimensional image in (B) was measured using the maximum value of the z gradient (ie: z shim). 2.14 Field plots along the z axis showing the effect of 53 the axial shims on the wide bore system. (A) Z 1 (B) Z 2 (C) Z 3 (D) Z\ The phantom used was a 5 mm NMR tube of length 7 c m , f i l led with doped water and aligned along the z axis. 2.15 Several planar 2D images in the xy plane, 54 obtained from a three-dimensional data set showing the effect of the x 2 - y J shim on the wide bore system. The f ield difference between each image is 1 ppm of the main f ie ld , or 1.89 uT. The phantom used was a thin circular disk (6.5 cm diameter, 1 cm thickness) containing doped water. 2.16 Several planar 2D images in the xy plane showing 55 the effect of the xy shim. The f ield difference between each image is 1 ppm of the main f ie ld , or 1.89 juT. The phantom and the experimental parameters are the same as that in figure 2.15. 2.17 Field plot of an extended 14 cm long, 5 mm NMR 56 tube sample containing doped water. The image (B) was obtained after good shimming of the phantom on the wide bore sys tem. Transposition of (B) y ields the image (C). Under these conditions the high resolution spectrum (A) yielded a linewidth of 3.9 Hz. However, residual inhomogeneities are sti l l present, as seen in the image. 2.18 Evaluation of the Bi homogeneity using a phantom 59 made up of four capillary tubes (1.6 mm id) containing water. The standard 5 mm probe on the narrow bore system yielded the contour plot image shown in (A). The image in (B) was obtained using the C - 1 3 probe as mentioned before and (C) represents a stacked plot image of (B) on a different plot scale. The intensity variations in these images reflects the B1 inhomogeneity across the larger sample. A l l images were obtained using the 2D spin warp imaging sequence. xi 2.19 Evaluation of the Bj homogeneity on the wide 60 bore sys tem, using a phantom consisting of 5 vials (1.3 cm diameter) containing doped water. The image shown in (A) was obtained from a 7.5 cm axial resonator probe and that in (B) from a 12.5 cm axial resonator probe. The separation between the vials are 3 - 4 mm and 2 - 3 mm respectively on the two resonator probes. 2.20 (A) Two-d imens ional images depicting proton 65 density variations of a phantom consisting of f ive vials (1.3 cm id), containing M n 2 " 1 " doped H 2 0/D 2 0 mixture with varying amounts of proton concentration. The vials are arranged as indicated in (B). 2.21 Two-d imens ional spin warp variant images of 70 glass capillary tubes containing water, located inside a 5 mm NMR tube. (A) 1.2 mm id tubes; (B) 250-300 Mm id ; 1.2 mm od ; (C) 140-220 nm id. Chapter 3 3.1 Schematic diagram of the pulse and gradient 78 sequence used for two-d imensional x, 6 imaging. 3.2 Two-d imensional chemical shift resolved images 79 of 2 vials containing water and acetone respectively. One dimension represents the spatial dimension, while the other represents the chemical shift. Transposition of the image in (A) yields that in (B), which represents the view along the chemical shift dimension. 3.3 Two-d imens ional chemical shift resolved images 81 of 4 capillary tubes mounted inside a 10 mm NMR tube, two containing water and two containing benzene respectively, are displayed as stacked plots (a) and contour plots (b). The image was obtained on the narrow bore system. 3.4 Two-d imens ional chemical shift resolved images 82 of a phantom consisting of 8 capillary tubes, four containing water and four containing benzene. This image is again from the narrow bore system. The intensity variations are due to the inhomogeneity of the Bj f ield of this particular probe. xii 3.5 The effect of increasing gradient strength is 83 illustrated on the same phantom as that used in figure 3.3 but contained inside a 5 mm NMR tube. (A) gradient strengths of 0.1 G/cm. (B) gradient strengths of 2 G/cm (C) gradient strengths of 4 G/cm. These are all two-d imensional chemical shift resolved images, obtained on the narrow bore sys tem. Note that the information is mixed up in (C). This is because, the gradient induced dispersion is greater than the chemical shift separation between the two resonances. 3.6 The pulse and gradient sequence to perform 86 three-dimensional chemical shift imaging. The spin echo signal is acquired after a composite 180° pulse in a homogeneous f ie ld , ie; all gradients are switched off during data acquisit ion. 3.7 Two-d imensional xz planar images obtained from 87 three-dimensional data set of a phantom, which comprises of a vial of water and a vial of acetone are shown in (A) and (B) respectively. 32 increments of each gradient G x and G z were performed. 3.8 An object with two chemically shifted resonances 88 and its three-dimensional image matrix are represented. Two axes for spatial information (x and y) and a third for frequency information (5). At the left, a conventional image, obtained by summing information along the 6 axis is seen. Note, that the spectral information can be extracted from any spatial x,y coordinate, or vice versa. The dark lines represent the curvature of the image planes due to distortion. [From reference (111)] 3.9 Two-d imensional chemical shift resolved images 91 obtained using the sequence described in section 3.2.1 for two sets of vials containing 10 mM C u S 0 4 (A) and (B), and 1 mM M n S 0 4 (C) and (D) respectively. 3.10 Effect of summing up data sets to minimize 92 distortion is illustrated by means of a three-dimensional imaging experiment. A and B represent stacked and contour plot display of a 2D xz planar image obtained from a single plane along the chemical shift dimension respectively. Similarly C and D represent 2D xz planar image obtained by summing images from two different planes along the chemical shift dimension. xiii 3.11 (A) shows the conventional high resolution NMR 94 spectrum at 270 MHz, of a phantom consist ing of two capillary tubes with water and another two with ethanol. The - O H and - C H 3 signals were deliberately folded over to decrease the spectral width in the chemical shift dimension. The images in (B), (C), and (D) represent two-d imensional xy planar display of water and the two high f ield components of the methyl triplet respectively; the latter demonstrates the chemical shift resolution attainable on our narrow bore system. [From reference (117)] 3.12 The pulse sequence for 4D chemical shift imaging. 96 (A) FID acquisition (B) spin echo acquisition after a composite 180° pulse. 3.13 (A) shows the high resolution NMR spectrum at 98 80 MHz of a phantom comprising of a vial containing CHCI 3 and H 2 0 / D 2 0 . (B) and (C) represent individual xy planar images of CHCI 3 and H 2 0/D 2 0 respectively, extracted from a 4D data set. 3.14 2D images of a vial containing motor oil and 105 water. The gradient is applied across the vial crosssect ion, and therefore, for both posit ive and negative values of the gradient the image appears the same as seen in (A) and (B) respectively. However, the image of one component is shifted with respect to the other due to the chemical shift differences between the two components. 3.15 2D images of the same phantom as in figure 3.14, 106 except that the. gradient is applied along the long axis of the v ia l . The images for both positive and negative gradients are not equivalent as seen in (A) and (B) respectively. Regions of low (in A ) and enhanced (in B) intensities are seen depending on the orientation of the phantom (see text for details). 3.16 Schematic diagram showing the effect of gradient 107 on the orientation of the phantom which contains equal amounts of water and o i l ; (A) and (B) represent gradient G v acting along +y and - y directions respectively. The length of the phantom is 2y and the chemical shift difference between the two resonances is 6. xiv 3.17 2D images of a phantom which consists of a 108 beaker of oil placed at the center of a petri dish containing doped water. The composite images of this system are shown in (A) and (B). The image in (B), however, has been obtained with reduced gradients (see text for details). Chapter 4 4.1 The pulse sequence used for three-dimensional 116 imaging together with the inversion-recovery preparation pulse. 4.2 Single slice images of the chromatography column 118 mounted vertically inside the wide bore system obtained using the sequence given in figure 4.1. 4.3 Two-dimensional images in (A) and (B) depict the 128 distribution of water protons within a berea sandstone sample and a limestone (lepine) sample which have been partially saturated with water, and contain 5.5% and 11.8% water by weight respectively. Both samples are cylindrical in shape and the images represent the projected intensity across the diameter of the samples. 4.4 Schematic diagram illustrating the effects of 130 gradient strength and rf frequency bandwidth on the thickness of the excited slice. 4.5 Three-dimensional x, y, z imaging sequence with a 132 180° refocussing pulse. 4.6 Two-dimensional slice images (A) to (E), obtained 133 from a three-dimensional data set of the same sample as that used in figure 4.3 A. The slice thickness is ca. 900 jum and each slice is separated by 2.8 mm. (F) shows the sum of images (A)-(E), and this image closely approaches the 2D projection image seen in figure 4.3 A . 4.7 Two-dimensional xy planar chemical shift resolved 135 images obtained from a 3D (x,y,5) data set of a composite berea sandstone sample. The images are shown as stacked plots and the normal high resolution NMR spectrum of the composite sample is shown in the inset. 4.8 A modified three-dimensional x, y, z imaging 137 sequence with inversion-recovery preparation. XV 4.9 Two-d imensional sl ice images obtained f rom a T1 138 sensitive 3D data set are shown in (A)-(E). The same sandstone sample has been saturated with a mixture of oil and H 20/DjO and these sets of images are those of water while the oil resonance was passing through its null. The slice thickness and the sl ice separation are the same as that in figure 4.6. (F) is a sum of (A)-(E). 4.10 Two-d imensional sl ice images of the same 139 sandstone sample obtained using the sequence given in figure 4.8. However, these represent the oi l present in the sample. The null time for water is 800 ms. (F) is a sum of (A)-(E). 4.11 Chemical shift resolved two-d imensional planar 141 images obtained from a 4D data set of a limestone sample saturated with North Sea crude o i l . The images illustrate the water, (a), and o i l , (b), present in the sample. 4.12 The different regions in the crosssect ion of an 144 oak tree. [From reference (210)] 4.13 Two-d imensional image of a douglas - f i r sample 150 saturated partially with water (A) is illustrated in (B). The thickness of the sample is 1 cm and the separation between the growth rings is ca. 1 mm. 4.14 Two-d imensional sl ice images (slice thickness = 1 151 cm) obtained from the hospital NMR scanner, depicting crosssectional and transverse views of wood samples saturated with water. (A) shows clear separation of heartwood/sapwood regions in a douglas- f i r sample, whi le, (B) shows a buried knot in the spruce sample. The head coi l was used to acquire these images. 4.15 Two-d imensional sl ice image (slice thickness = 152 1cm) in (B), obtained from the hospital NMR scanner using the head c o i l , depicts the crosssectional v iew of a freshly cut aspen sample. (A) is a photograph of the same sample. The growth rings are clearly visible in the image. 4.16 Two-d imensional multiple slice (1 cm thick) image 154 set (A -D) of an aspen sample (ca. 25 cm diameter). Each sl ice is separated by 1 cm. A hidden knot and a region of decay (rot) are highlighted. The head coil of the who le -body NMR scanner was used to acquire this image set. xvi 4.17 Two-dimensional slice image (slice thickness = 157 1cm) in (B), obtained from the hospital scanner, depicts a crosssectional view of a freshly cut aspen sample with high degree of decay. (A) is a photograph of the same sample. The body coil was used to acquire this image. 4.18 Two-dimensional image obtained on the 80 MHz 160 wide bore system of a circular disk section (3.5 cm diameter, 1 cm thickness) of the sample used in figure 4.17. The section corresponds to region 9 in Table 4.4, (B) is a photograph of the same section. 4.19 J1 sensitive image of the same section as in 161 figure 4.18 obtained by partial nulling of the water in zone (ii). Note that the zone lines are clearly visible. 4.20 Two-dimensional image obtained on the 80 MHz 163 wide bore system of a circular disk section (3.5 cm diameter, 1 cm thickness) of the sample used in figure 4.17. The section corresponds to region 3 in Table 4.4. (B) is a photograph of the same section. 4.21 T i sensitive image of the same section as in 164 figure 4.20 obtained by partial nulling of the water in zone (ii). Note that the zone lines are clearly visible. 4.22 Effect of drying wood is illustrated on a 167 douglas-fir sample. All 2D images were acquired on the wide bore system, with the same attenuation and are scaled the same. 4.23 Effect of drying wood (contd) (A) represents the 168 same image as in figure 4.22 D, but with different attenuation level on the spectrometer. 4.24 Typical moisture gradient curve across the 170 thickness of a piece of wood during drying. Note that the moisture content of the central section is higher. [From reference (213)] 4.25 Curve (a) represents the plot of the logarithm of 171 the water proton echo amplitudes of a moist spruce sample against the time measured from the 90° pulse in a Carr-Purcell sequence. Curve (b) represents the shorter component on a different time scale. [From reference (205)] xvii 4.26 Series of spectra reflecting the changes in NMR 172 lineshape for a douglas - f i r sample during the process of drying. 4.27 Effect of impregnation of wood with paramagnetic 176 ions is illustrated on another douglas - f i r sample. A l l 2D images are scaled the same and 10 mM MnCI 2 solution was used to soak the sample. 4.28 Effect of impregnation of wood with paramagnetic 177 ions. The dif fusion of the C u ^ + i o n s embedded within the douglas - f i r sample is fo l lowed. Note, that the dif fusion of the paramagnetic ions along the wood fibers is faster than across the fibers as expected. 4.29 Two-d imensional images of two similar white spruce p lywood samples with different levels of water saturation. (A) saturated under pressure and (B) by plain soaking. (C) is a photograph of a sample specimen (see text for details). 4.30 (A) Two-d imensional image of an edge-glued white spruce plywood sample with water saturation. The glue line can be visualized in the image. (B) is the photograph of the sample. LIST OF TABLES Table Chapter 2 2.1 2.2 Chapter 3 3.1 Chapter 4 4.1 4.2 4.3 4.4 4.5 4.6 4.7 Chapter 6 6.1 Integrated Image Intensities of 5 Vials Containing 10 mM CuSO, in Two RF Probes. Image Intensities and NMR Data of H 2 0/D 2 0 Systems Containing Paramagnetic Mn^ + ions. The Effect of the Increase in Dimension Size on the Total Experimental Time for Mult i -d imensional NMR Imaging Methods. Relaxation Rates of Water at 80 MHz in Different Regions of the Column. Proton Relaxation Times and Linewidths of Water in Porous Rocks and Sintered Glass Compacts at 80 MHz. Tj Values of the Different Regions of a Fresh Sample of Aspen Measured at 6 MHz on the Whole -body Scanner. NMR and Microbiological Properties of Different Regions in the Aspen Sample. NMR and M.C Variations of Region 9 in Table 4.4. NMR and M.C Variations of Region 3 in Table 4.4. NMR Properties and M.C Variations of Douglas - f i r Wood During Drying. Intensity Changes Corresponding to Each Color Level in Colorgraphic Display. xix LIST OF ABBREVIATIONS NMR Nuclear Magnetic Resonance FID Free Induction Decay CW Continuous Wave FT Fourier Transformation SPD Single Phase Detection QPD Quadrature Phase Detection RF(rf) Radio Frequency B. Static Magnetic Field B, Appl ied Radio Frequency Field T a Spin -Latt ice or Longitudinal Relaxation Time T 2 S p i n - S p i n or Transverse Relaxation Time 7 Gyromagnetic Ratio 6 Chemical Shift P Spin Density G k Magnetic Field Gradient along k-di rect ion (k=x,y,z) " o Larmor Precessional Frequency AB0 Static Field Inhomogeneity 2D Two-Dimensional 3D Three-Dimensional 4D Four-Dimensional M.C Moisture Content CT Computerized Tomography id Internal Diameter LC Liquid Chromatography HPLC High Pressure Liquid Chromatography M C A C Metal Chelate A f f in i t y Chromatography A Q N Acquis i t ion MHz Mega Hertz Hz Hertz PPM Parts Per Mi l l ion Dedicated to My Parents 1 CHAPTER 1 GENERAL INTRODUCTION 2 1.1 BACKGROUND The phenomenon of Nuclear Magnetic Resonance (NMR) was first discovered independently by Bloch, Hansen, and Packard (1-3), of Stanford University and Purcel l , Torrey, and Pound (4), of Harvard University more than three decades ago, work for which they were jointly awarded the Nobel Prize in 1952. NMR is a process undergone by nuclei which have a magnetic moment, when they are placed in a strong magnetic f ie ld . If these magnetized nuclei are exposed to rad io -waves of a specif ic frequency they absorb a signal of the same frequency as the one applied. Proton, being the most sensitive naturally occurring nucleus, coupled with its common occurrence in organic substances, continues to be the most widely studied nucleus. However, in 1966 Ernst and Anderson demonstrated the applicability of Fourier transformation to NMR (5), which undoubtedly started a new era of NMR Spectroscopy. Since then, NMR has become one of the major analytical research tools . The most widely used applications have been in Chemistry, and include chemical and structural identif ication, reaction studies (6-8). Many lower sensit ivity nuclei can also be studied routinely. Equally important, NMR has been used in developing the theories of solids and liquids and has provided Physics with a fertile test bed for quantum mechanical, spectroscopic and statistical quantum mechanical theories (9-12). A s a consequence there exists not only a body of diverse and growing applications, but also an elegant and extensive theoretical foundation. An interested reader is referred to the literature (13) as to how these experiments are carried out. 3 The early practitioners of NMR back in the 1950's had strongly in their minds the possibi l i ty of using NMR to study biological systems. Purcell and Hahn are reputed to have attempted "to place their heads inside the magnet in an effort to try and observe differences in NMR signals caused by heavy thinking as opposed to being relaxed (14)". Although the idea was evident then, it was not until in 1971 that NMR applications to medical science began. Damadian suggested that NMR could be useful for tumour detection based on the different proton relaxation times he observed for normal and malignant tissues (15). Shortly thereafter, in 1973, Lauterbur published the first NMR image of two water - f i l l ed capillary tubes (16). An image of an object is a representation of the spatial distribution of one or more of its properties. The approach used by Lauterbur was similar to that used in X - r a y Computerized Tomography (CT), the image being reconstructed from projections of the object along different directions using back-project ion image-reconstruction algorithms (figure 1.1). As time progressed other workers devised and demonstrated alternate NMR imaging schemes. The most startling development came in 1975, when Kumar et al . (17) extended the concept of two-d imensional Fourier transform (2D FT) methods to imaging (18,19). Since then several variants of that technique have been developed and implemented successful ly (20). However, it is important to note that the basic idea of mult i -d imensional i ty has not changed and this forms the central theme of this thesis. Most of the early studies were performed with modi f ied , conventional NMR spectrometers, and magnets which afforded a restricted 4 Fig 1.1 Projection Reconstruction imaging. From a series of one-dimensional projections, obtained by acquiring the NMR signal in the presence of magnetic field gradients directed along different orientations relative to the sample, a two-dimensional NMR image of the sample can be reconstructed. The method is analogous to the one used in X - ray CT. [ From reference (16) ] 5 sample access of about 1-3 c m , and imaging times as long as 2 hours. Nevertheless, the proton imaging results from water - f i l l ed phantoms and small vegetable samples were suff ic iently encouraging to prompt the formidable task of designing and constructing larger-scale systems. One of the first such systems was completed by Hinshaw et a l . (21) in 1977, yielding promising proton imaging results from human forearm, and live animals of up to 8 cm in diameter. A review by Bottomley gives a good account of this early development (22). Widespread clinical trials began only in 1980 largely as a result of commercial activit ies of increasing vigor. The imaging of NMR parameters other than nuclear density and relaxation times has received attention lately. Chemical shift , being the most important, has enabled in spatial mapping of more than one chemical species. This technique, also termed Chemical Shift Imaging (CSI), has opened up new areas, such as multinuclear (eg: P -31) imaging, where metabolic states of intact biological t issues may eventually be monitored, in order to assess their health and response to therapy. At present, NMR imaging provides an anatomical map of soft t issues with a contrast which is frequently better than that of X - r a y CT for many areas of the human body in particular, the head. While the potential medical applications of NMR imaging are many, it is important to note that this new imaging modality also has a number of potentially useful applications in non-medical areas of science. The first such major application was in studying unsaturated water f low in porous media, by Gummerson et al . (23) in 1979, fo l lowed by studies on polymer composites by Rothwell et al . (24) in 1984. As wi l l be seen, this forms one of the themes for this thesis. 6 It is appropriate at this stage to give a brief description of the basic NMR experiment itself. A s we shall see in the next Chapter, this basic treatment acts as the foundation in understanding all Fourier-based imaging methods. 1.2 BASIC PULSE FT NMR EXPERIMENT The classical magnetization vector model provides a convenient picture for understanding the basic feature of pulse FT NMR experiments. In this model , the effects of radiofrequency (rf) pulses on an ensemble of nuclear spins experiencing a static magnetic f ield B„ are described in terms of the behavior of a macroscopic magnetization vector M. At equilibrium, this can be represented as a vector precessing at a characteristic frequency (o)0) called the Larmor precessional frequency, with its axis of precession aligned along the direction of the f ield B 0 as shown in figure 1.2. This precessional frequency is given by, "o = 7 B 0 (1.1) where 7 is the gyromagnetic ratio of the spins. The concept of "rotating" reference frame introduced by Torrey (25) is a convenient means of describing most experiments in NMR. In contrast to the "laboratory" frame it uses a coordinate system (x',y',z') that rotates in the same sense and with the same frequency, CJ0, as the rotating f ield vector of the rf f i e ld ; as a result the net equilibrium magnetization vector from an ensemble of spins, lies stationary along the z' (=z) -direction. The effect of a rf pulse is to rotate the total magnetization vector from its posit ion along z, through an angle about the direction of the applied rf 7 Fig 1.2 The macroscopic magnetization vector M of a set of equivalent nuclei precesses about the magnetic field B„ at its characteristic Larmor frequency. At any instant, M can be represented by its components M x , My and M z along the three directions x, y and z respectively. 8 f ie ld , ( B i ) , which is usually applied along the x ' -d i rect ion. This angle is called the "flip angle" and is given by, a = 7B, t 7 (1.2) where t^ is the duration of the rf pulse. Thus, a — ir/2 corresponds to a 90° pulse, and this brings the magnetization into the x'y' plane creating transverse magnetization. The receiver coi l is designed to measure the induced voltage arising from this transverse magnetization. In single phase detection ( S P D ) spectrometers the y' component of the magnetization (My,), is detected fo l lowing a pulse. It is also possible to detect M^and M^simultaneously by quadrature phase detection ( Q P D ) . The amplitudes of M x ,and M^,show characteristics of cosine and sine functions respectively. The observed signal after a 90° pulse in a conventional S P D NMR experiment can be expressed as, M y , ( t ) = M 0cos(AcJt)exp(-t/T 2) (1.3) where Ac; = co0 - CJ' and CJ' is the frequency of detection. The exponential term in the above equation indicates the damping of the NMR signal with time due to transverse relaxation (T2). Fol lowing detection, the time domain signal is first digitized and then subjected to Fourier transformation, to yield a complex, frequency domain spectrum (26), S(CJ) = [M 0T 2/2{1+(ACJT 2)'}] - [iM0Acj/2{1+(AcoT2)*}] (1.4) The real and imaginary parts in equation 1.4 correspond to the absorption and dispersion mode signals respectively. The lineshapes are Lorentzian and are illustrated in figure 1.3. Fig 1.3 The absorption mode (a) and the dispersion mode (b) lineshapes. The effect of B„ f ield inhomogeneity is taken into account by introducing an effect ive transverse relaxation t ime, T 2 , that contains contributions from true relaxation as wel l as from inhomogeneity of the static f ield (AB 0), 1/T,*= 1/T, + 7 A B 0 / 2 (1.5) In practice, f ie ld inhomogeneity makes by far the most important contribution to the disappearance of the transverse magnetization and therefore, to the linewidth of the spectrum in high resolution NMR spectroscopy. A l l imaging methods employ magnetic f ield gradients of one sort or another to achieve spatial differentiation. Hence, the effect of these gradients on spins is dealt with briefly in the fol lowing sect ion. 1.3 EFFECT OF MAGNETIC FIELD GRADIENTS The idea of using magnetic f ield gradients in NMR is almost as old as NMR itself (27-29). The early pioneers in the field realized that gradients or inhomogeneities in the main f ield often l imited the widths attainable for absorption lines in liquids. The conventional NMR absorption techniques are inherently one dimensional in that, the absorption is measured as a function of one variable, the Larmor precessional frequency In all imaging techniques, the general objectives are to measure a number of NMR parameters (eg: spin density, Tlt T 2) as a function of their spatial coordinates. The spatial information is obtained by degrading the uniformity of the static magnetic f ield such that the magnetizations from different parts of the specimen lie in sl ightly different static f ie lds , and hence precess at different frequencies characteristic of their posit ions. Spec i f ica l l y , the precession frequency at x, when a gradient G x is applied along the x direction is , cox = 7(B 0 + G x x ) (1.6) The magnetic f ie ld gradient is, in general, described by a tensor with nine components, but for large B 0 , it is necessary to consider only the three components 9B z /9x , 9 B z / 9 y and 9 B z / 9 z (30). This precession frequency CJ^ is recorded as the NMR signal as a function of time which, on Fourier transformation yields the frequency space information. The resultant absorption profi le is the spin density projection along the x axis. For simple geometrical shapes and homogeneous distributions of spins, the projection profi les lead to a great deal of information concerning the c ross -sect iona l shape of the object. However, in order to obtain the actual shape of the object, it is necessary to observe a number of such projections corresponding to regularly disposed orientations of the gradient. The way in which such projections are obtained depends on the imaging methodology used, it is important to note, that, in order to employ these magnetic f ield gradients, one needs to know their magnitudes and their linearities across the sample under study. The determination of these parameters wi l l be discussed in section 2.4.1. The next section c lass i f ies different imaging techniques, together with their merits and demerits, at the time the work described in this thesis began. It should be stressed that, presently mult i -d imensional Fourier transform techniques have become predominantly popular due to their inherent versati l i ty. 1.4 CLASSIFICATION OF IMAGING TECHNIQUES NMR imaging techniques can be c lassi f ied into four types, depending on the type of volume that produces the perceived signal subsequently used to form the image: a single point, a l ine, a plane, and an entire three-dimensional object (figure 1.4). The single point method determines the points in the image most directly, by measuring points one at a t ime; on the other hand, the three-dimensional technique repeatedly measures all points simultaneously. The line and plane techniques fall between these two extremes. Brunner et al . (31) have reviewed a critical evaluation of the sensit ivity and performance time of all imaging techniques. They state that, since there is a large spread in these values for the various schemes, careful choices should be made for sucessful applications. The main problem with s ingle -point techniques is that data acquisition is s low. Planar and whole-object techniques which derive images based on information arising simultaneously from all points in a plane, have become popular. Because of their increased versatil ity and potential , the mult i -d imensional Fourier transformation techniques have outgrown the rest, and at present, it is generally believed that these techniques are superior. Some justif ication for this notion is that the original technique of Lauterbur (16) which triggerred off the era of NMR imaging is now virtually unused. y (a) (b) (c) (d) Fig 1.4 Scheme of classification of the various NMR imaging techniques; (a) sequential point measurement (b) sequential line measurement (c) sequential plane measurement (d) simultaneous volume measurement [ From reference (30) ] 1.5 ORGANIZATION OF THIS THESIS The rate of growth of the instrumental and methodological developments of NMR imaging continues to be rapid, with new experimental techniques as well as improved hardware developments being reported by numerous research groups (32-40). When Professor Hall's group, developed an interest in this f ield in 1981, it was possible using a home-bui l t , conventional high resolution imaging/spectrometer to develop and evaluate new experimental techniques using small sample models containing l iquids; it was impossible to study "real" objects for several technical reasons. Given the importance of the chemical shift parameter in structural NMR, incorporation of this parameter into imaging methods was actively pursued (41); part of the work described in this thesis forms a contribution to that development. However, the major thrust of this thesis is to explore possible applications of various imaging measurements to a number of areas which, it was felt , might be important to industry. These applications, all of which were without prior literature precedent at the start of this study, are summarized in Chapter 4. Because those results could not have been obtained without knowledge of the basic fundamentals of the NMR imaging methods used, and access to properly evaluated hardware, those studies were of necessity preceded by a substantial investment of t ime, and effort in development of methods, some of which produced results of interest in their own right. These are discussed in Chapters 2 and 3. The detailed description of the two-d imensional Fourier transform imaging technique given in Chapter 2, emphasizes an experimentalist's point of v iew and excludes complex mathematical formal isms. The account starts with the concept of two-d imensional Fourier transform, and is fo l lowed by an overview of the two principal imaging techniques, up to the stage of generating an image of an object. Mul t i -d imensional Fourier transform Chemical Shift imaging is dealt with in Chapter 3. Initial description of the various imaging sequences is fo l lowed by a discussion of, what is at present, a very promising, but not widely used because of time l imitations, Four-dimensional Chemical Shift imaging. In both of those Chapters, the emphasis is on the use of "phantoms" (model objects, designed to evaluate performance characteristics of the instrumentation). Appl icat ions of the methodologies developed during those studies are dealt in Chapter 4; these involve several non-medical systems such as, visualization of Chromatographic Separations, imaging of Oil/Water mixtures in porous rocks, and imaging of W o o d . Finally, the main conclusions derived from this study are summarized in Chapter 5. The experimental work of this thesis was performed using two NMR devices, both of which were in construction during the course of the work itself . The first of these, on which most of the initial experiments were done is based on a 54 mm narrow bore superconducting magnet supported by a high resolution NMR console. The second, also based on a high resolution console, uses a 31 cm wide bore superconducting magnet. The resonance frequencies of these systems are 270 MHz and 80 MHz respectively, for protons, which were the nuclei of interest in this work. During the course of this study, extensions by other laboratories of some of the author's work have appeared in the literature therefore, in order to make the description complete material from these references have been incorporated fo l lowing dicussion of the work of this thesis. One final point must be stressed. During the time scale of this work, the instrumentation used was being modif ied by two other graduate students, S.L. Talagala and S.D. Luck, along with two technicians, T. Marcus and C. Neale in the Chemistry Department. When knowledge of the performance of that hardware was necessary in order to support the work of this particular thesis, appropriate evaluations were made by this writer and are included in this thesis ; all such evaluations have been clearly identified in this thesis. 17 CHAPTER 2 TWO-DIMENSIONAL FOURIER TRANSFORM IMAGING 2.1 INTRODUCTION One of the major developments in NMR spectroscopy has been the introduction of the concept of two-d imensional Fourier transformation. Since 1966 it has been possible to obtain high resolution NMR spectra by Fourier transformation of the transient, t ime-domain signal excited by a short rf pulse. S ( t ) > S(co) (2.1) Although the novel idea of performing a double Fourier transformation with respect to two independent time parameters was formulated by Jeener in 1971 (18), it was not until 1976, that the generality of the idea was demonstrated experimentally by Aue et al . (19). Any signal that is a function of two independent time variables can be converted in this way into a function of two frequency variables. S ( t „ t 2 ) = = = = = = > S(uuu2) (2.2) The introduction of two time variables t^tj, necessitates the partition of the experiment into three time periods as illustrated in figure 2.1. The most important domain is the "evolution period", during which the spin system is al lowed to evolve and to sample its nuclear or molecular environment. Prior to that, the system is suitably manipulated during a "preparation period". Data col lection fo l lows the evolution period. If the evolution period is incremented n t imes with equal time increments of At^ the detected signal for the whole experiment wi l l be a function of two time variables, S(tj,t2). 2D spectroscopy is always possible if a systematic variation of the evolution period results in a periodic change of a property of the spin system (eg: scalar coupling). To Preparation Evolution Detection 4 Fig 2.1 Partitioning of the timing axis in 2D FT NMR experiments. appreciate the signif icance of the 2D FT technique, it is important to realize that the experiment relates the behavior of the spins during tj to their behavior during t 2. Fourier Zeugmatography, as proposed by Kumar et a l . (17), is a special case of 2D spectroscopy; in this case the spin system is made to evolve under the influence of the magnetic f ield gradients. In this Chapter the analysis of this basic technique wi l l serve to illustrate the underlying principles governing NMR imaging from a standpoint of a high resolution 2D NMR spectroscopist . This discussion wi l l be fo l lowed by a more general description of all Fourier -based NMR imaging methods, including the more versatile spin warp imaging sequence (42). The mathematical formulations introduced briefly in the description of these methodologies yields a better understanding of the techniques, and an interested reader is referred to the original literature for detailed reviews of these formulations (43-49). Data col lect ion and data handling strategies for two-d imensional imaging are focussed upon next. The latter part of the discussion concentrates on studies of phantoms made at the beginning of this work and emphasizes the evaluation and modif icat ion of a conventional high resolution NMR spectrometer to enable it to perform imaging studies. 2.2 FOURIER ZEUGMATOGRAPHY This method is based on the successive, pulsed application of orthogonal magnetic f ield gradients during the free induction decay (FID) of the nuclear spin system. The spins, therefore precess at different frequencies during the evolution and detection periods depending on the magnitude of the applied gradients. The radiofrequency (rf.) pulse and gradient sequence is illustrated in figure 2.2. At time t=0 , a short, nonselective 90° rf pulse nutates the equilibrium magnetization into the x'y' plane of the rotating reference frame. Linear f ie ld gradients G x and Gy are applied in succession during the course of the signal decay and the FID is sampled in the presence of Gy. The variable time period t1 is incremented in constant intervals and the corresponding signal is acquired as a function of t 2 , so as to build up a data matrix of the fo rm, S(t) = S ^ t , ) (2.3) 90° : AQN 2.2 Two-dimensional Fourier Zeugmatography. During the evolution period t l t gradient G x spatially encodes the x dimension. Similarly, gradient G v during the detection period encodes y dimension. FID's are recorded in t2 for different tj values, and Fourier transformed in two dimensions to yield the image. The 2D FT of S(t) is S(co) = S(coua)2) such that, S(w) = //SCt JexpC-iut Jc l t^t , (2.4) The observed signal S(t) is a composite of the contributions from the 2D volume elements of the sample and is given by, S(t) = J7p(r)s(r,t)dv (2.5) where s(r,t) dv is the contribution from the volume element dv = dxdy at posit ion r, and p(r) is the spatial spin density. Therefore, expressing S(CJ) as an integral over all 2D volume elements, one obtains a relationship of the fo rm, S(w) = //p(r)s(r,cj)dv (2.6) where s(r,co) is the Fourier transform of s(r,t). Considering a single resonance, the phase sensit ively detected signal in SPD, at the transmitter frequency of is given by, s(r,t) = M.cosKAu + TrG^Jt, + (AOJ+ 7 G y y) t J ]exp[ - ( t 1 +t 2 )/T J ] (2.7) where Ao> = OO0-CJ' is the frequency offset and ACJ^ = TG^k is the effect caused by the gradients. Restricting to the contribution of the resonance near AGO (which is usually the case), the FT of s(r,t) can be written as (17), s(r,cj) = 1/2[F(Au+ 7 6 ^ - 0 ) ! ) F(Au + 7 G y y-a> 2 ) ] (2.8) with complex lineshape function, F(w) = M0/[T2{(1/T2)2+w2}] + iMdU/Kl/TJ '+u' ] (2.9) Substituting for s ( rp) in equation 2.6, the 2D FT of the observed signal can be represented as, S(w) = 1/2[;;p(r)F(Acj + 7G x x -w l )F (Aa>+ 7 G y y -w 2 )dxdy] (2.10) In this data matrix, it is seen that the spatial information corresponding to x is coded along one dimension, while that corresponding to y is coded along the other. In a system which contains more than one kind of nucleus (ie: with different chemical shifts), equation 2.10 wi l l be further modif ied to include this parameter in both dimensions, since the gradients are applied during the magnetization decay. This particular sequence is very susceptible to inhomogeneities in the static magnetic f ield B 0 which is a serious l imitation of this technique. Furthermore, both dimensions simultaneously contain both chemical shifts and spatial information. Thus, in general, this technique is unsuitable for Chemical Shift imaging (see chapter 4). In v iew of these limitations it has to be realized that, although this is the technique which effect ively introduced the concept of mult i -d imensional FT imaging, as wi l l be seen in the next section modif icat ions were necessary to produce a more versatile and a practical imaging sequence. 2.3 SPIN WARP IMAGING Fundamentally, spin warp imaging is derived from the method of Fourier zeugmatography, but differs from it in several important aspects. Probably the most important of these is that the timing of events in each pulse sequence is identical, thereby yielding a practical imaging procedure which is remarkably tolerant of the inhomogeneities in the static magnetic f ie ld . The sequence itself is illustrated in figure 2.3 A . Instead of applying gradients of constant magnitude for varying periods of time as in the Fourier zeugmatography method, variable amplitude gradients are applied for fixed periods of time and a spin echo is formed by the inversion of a gradient. Alternatively , the spin echo can be formed by application of a Fig 2.3 Two-dimensional Spin Warp imaging. A variable amplitude G y pulse is applied during the evolution period and the spin echo signal is detected. The spin echo is formed by gradient reversal in (A) and by 180" refocussing rf pulse in (B). In either case, the sequence is repeated for different values of G v , and the image obtained by double Fourier transformation. ro 25 180° rf refocussing pulse at the end of the evolution period as shown in figure 2.3 B. These two sequences differ mainly in the way they generate the spin echo, and this wi l l be discussed in section 2.3.3. The amplitude and phase of the acquired signal are a function of the combined effect of the duration of evolution period plus the magnitude of the phase encoding gradient. S ince, the evolution period is kept constant, the signal in this pseudo time domain wi l l be modif ied only as a function of the magnitude of Gy, which is controlled solely by variations of its amplitude. In this section a more general approach is chosen to describe 2D spin warp imaging, which has the merit of being suitable for all Fourier imaging methodologies; although the discussion is restricted to two-d imensional imaging only, the logical extensions to mult i -d imensional imaging schemes are tr iv ial . 2.3.1 A N A L Y S I S OF JJHE TECHNIQUE In general, for a single resonance evolving under the influence of magnetic f ie ld gradients, the observed signal S(t) in quadrature phase detection (QPD) can be represented as (43), S(t) = ; ;p(x,y)exp[iAcj + i 7 G k . k - 1 / T 2 ] t d v (2.11) where the frequency offset Aco = o)0-co', the effect caused by the gradients ACJ^ = yG^'k, and k = x,y the two spatial coordinates, t is the time after the initial 90° excitation pulse and p(x,y), the 2D spin density distribution in a volume element dv = dxdy. If T 2 is long compared to the time over which the data are col lected, the T 2 term in equation 2.11 can be neglected (46). S impl i fy ing further, by observing on resonance, equation 2.11 can be rewritten as, S(t) = ; ;p (x ,y )exp[ i7tG k .k ]dxdy (2.12) Let q k = - 7 G k t then, S(t) = / ;p ( x , y )exp[ - iq k .k ]dxdy (2.13) This is the fundamental relation of NMR imaging; the signal detected phase-coherently is proportional to the spatial Fourier transform of p(x,y) (48). Let p(q) be this transform, then, P(t?) = J7p (x ,y)exp[ - iq k .k]dxdy (2.14) This implies that, S(t) = p(q) (2.15) Ensuring identical initial excitation of p(x,y) and by obtaining a suff iciently dense set of all q values with repeated measurements of S(t) for varying values of G k , an image of p(x,y) can be reconstructed, as its transform S(t) would be known everywhere. In other words, the inverse Fourier transform of S(t) y ields the distribution of p(x,y) which is the image. The general principle then, is that the spatial encoding process consists of sampling the NMR responses by means of time varying gradients; the decoding process consists of processing those data samples (usually by means of Fourier transform) to yield a discrete image which is an estimate of the original spatial distribution. In the case of the spin warp imaging sequences described in section 2.3, equation 2.12 can be further modif ied to accommodate effects arising from different chemical shifts in the sys tem, particularly during data acquisit ion, to yield S(t) = ; ;p (x , y )exp i[7G y y t 1 + (7G xx + 5)t2]dxdy (2.16) The 2D FT can be denoted as (11), S(u) = ; ;p (x ,y )F(7G y y -o ; 1 )F(7G x x + 5-cu2)dxdy (2.17) This data matrix has spatial information (7G y y) along one dimension and spatial/chemical shift information (7G x + 6) along the other. Although attention wi l l be focussed in the next section only on applications of a single resonance, the usefulness of this imaging scheme to perform chemical shift imaging wi l l be dealt in greater detail in the next chapter. 2.3.2 PHASE ENCODING SPATIAL INFORMATION The principle of phase encoding is illustrated in figure 2.4. During the phase encoding period, nuclei precess at different frequencies depending on their posit ion relative to gradient G . A l l spins in column "a" wi l l have precessed through a given phase angle; Spins in column "b" which are in a stronger f ie ld wi l l have precessed through a larger phase angle, etc. The end result is that distance information along the y -d i rect ion is phase encoded. The read (or observe) gradient, G creates a further distribution of frequencies along the x -d i rect ion . Spins in row "7" move at one frequency, spins in row "2" move at a greater frequency etc., which provides the final spatial encoding along the x -d i rect ion . The frequencies caused by G x combine to form the signal which is sampled and stored in the computer. This process is repeated for the desired number of incrementally increased values of the phase encoding gradient, to generate the 2D data matrix, S ( G y , t x ) . Data handling wi l l be discussed in section 2.3.4, but it is instructive at this stage to review first the behavior of the 28 £ > err p r e p a r e y Gy c? £> f r e q u e n c y Gx 2 A The spins are prepared by a 90° pulse initially. Spatial encoding is accomplished, first by the application of the phase encoding gradient 6 y and then by the frequency encoding gradient G x . Note that the y spatial information is encoded as phase differences. spin isochromats in the rotating frame of reference. 2.3.3 SPIN PRECESSION A s mentioned previously in section 1.2, the macroscopic magnetization vector remains stationary along z -d i rect ion in the rotating reference frame. The effect of a rf pulse is to f l ip this total magnetization vector about the direction of the applied rf f ield (ie: x ' -axis) creating transverse magnetization. In the imaging sequences illustrated in figure 2.3 A and B the initial 90° excitation pulse creates a net magnetization in x'y' plane (figure 2.5 A) . When this excited system is allowed to evolve , the spin isochromats begin to dephase during the evolution period t,, due to the presence of the applied gradients and also due to the static f ield inhomogeneities; for s impl ic i ty we shall ignore the effect of Gy. Relative to the transmitter frequency at which the reference frame is rotating (figure 2.5 B), the spin isochromats may be assumed to contain leading, "fast" components and trail ing, "s low" components. The effect of the 180° rf pulse is similar to that in the pulse sequence used to measure sp in -sp in relaxation rates in conventional high resolution NMR. The spin vectors are rotated about the direction of the pulse (x' -direction as indicated in figure 2.5 C). After continued precession of the leading and the trailing components during a time period tu all spin isochromats are brought back into phase along the negative y ' -d i rect ion . It is important to note that the inhomogeneities in the main magnetic f ield are also refocussed at this point (figure 2.5 E). z Fig 2.5 Rotating reference frame representation of the spin echo sequences described in figure 2 .3 . (A) A 9 0 ' pulse applied along x' at time t=0 , causes the magnetization to tip to the positive y' axis. (B) The macroscopic magnetization of nuclei in different parts of the sample dephase as a result of the applied gradients and also due to the static field inhomogeneities during the evolution period t,. Looking down the positive z axis, in the reference frame rotating at the chemical shift frequency, the leading components appear to move clockwise and the trailing components appear to move counter-clockwise. (C) A 180 ' pulse along x' after time t,, rotates the spin isochromats about that axis. (D) The "fast" components still moving clockwise and the "s low" components moving counter-clockwise. (E) The spin isochromats refocus after an equal time t, along the negative y' direction. (F) On gradient reversal, the precessional direction is reversed for the leading and trailing components. (G) After an equal time, the spin isochromats refocus along positive y' direction. to o Gradient switching after time tu however, changes the precessional frequency of the individual components by reversing their direction of precession (figure 2.5 F). For example, a component precessing at a frequency CJ0+ACO, where Aw is the frequency spread caused by the fixed gradient, wi l l precess at tJ 0 -Aco , when the gradient is reversed. Hence, after an equal time period, all the magnetization components come back into phase. Note that any inhomogeneities in B 0 are not refocused by this particular pulse sequence as illustrated in figure 2.5 G. Chemical shift information is suppressed in the pseudo time domain in both these sequences and this is of practical importance for many applications relevant to Chemical Shift imaging. 2.3.4 CHOICE OF PARAMETERS The choice of sampling parameters in the observe dimension is straightforward. In order to achieve acceptable spatial resolution the magnetic f ield gradients must broaden the resonances by much more than the broadening AB caused by the natural iinewidths plus the static f ield inhomogeneities. If we consider- a sample of length L represented by N pixels across the image, the applied gradient G x (G/cm), should sat isfy the condition described in equation 2.18 (50). 7G XL/2TTN > AB (2.18) This implies a minimum gradient strength necessary to overcome any undesired broadening effects AB. The f ield of view in this dimension is related to the magnitude of G x and is given by (51), W 2 (cm) = 27r/[7G xf] (2.19) where t' is the dwell t ime. Hence, the sweep width must be selected to cover the desired f ield of v iew. The relationship between the sweep width and the dwell time is , Sweep width (Hz) = 1/2[dwell time] (2.20) and this corresponds to a digital resolution given by, Digital Resolution (Hz/point) = 2[sweep width]/[block size] (2.21) Therefore, the spatial resolution in the observe dimension is given by, Spatial Resolution (cm/point) = 2W 2/[block size] (2.22) The phase encoding gradient is varied in constant increments for each sequence, such that, G y = pAGy (2.23) where p varies from — (/?— 1 )/2 to (/?-1)/2 in integer steps and A G y is the gradient increment. The number of sequences performed is n. Note that n different spin echo signals are collected with both' negative and posit ive gradient. The total f ield of v iew in this domain is given by, W, (cm) = 27T/ [7AG y t J (2.24) and this corresponds to a spatial resolution of , Spatial Resolution (cm/point) = \Njn (2.25) This represents the size of the smallest spatial element from which an NMR signal is sampled. In order to get a wel l defined image, the digital resolution should be adequate and equation 2.25 clearly implies that more experiments means better spatial resolution in the phase encoding dimension. This is illustrated in figure 2.6 which shows 2D images of two vials (1.3 cm diameter) containing water; note the increased definit ion of Fig 2.6 Diagrammatic representation of the effect of spatial resolution on 2D images obtained at 80 MHz. The phantom consists of two vials (1.3 cm diameter) containing doped water mounted 5 mm apart. The images corresponds to (A) 128 increments (B) 64 increments (C) 32 increments (D) 16 increments of Gx. The final image matrix size however, is 256x256. The f ield of v iew in both dimensions remain the same. Experimental parameters \ sweep width = ± 16129 Hz ; acquisit ion time = 15.87 ms ; block size = 1024 ; f ie ld of view = 7.6 cm the samples with more experiments. It has to be emphasized at this point that all the expressions discussed above in this section strictly apply only to rectangularly shaped gradient pulses, which have very short rise and fall t imes, that is , the time the gradients take to reach the peak values or to reach zero after being switched of f . Although low inductance gradient coi ls shorten rise t imes, gradient shaping becomes necessary, particularly when high magnitudes and/or rapid switching are desired. Under these conditions the magnitude of the "effective gradient" during the respective time periods must be considered. For example, in the case of the phase encoding gradient G , Effective Gradient = (1/f) / G y(t)dt (2.26) In order to get symmetric representation of the final image, the net effect caused by both the steady and the phase encoding gradients must be made equal, ie; W , = W 2 ; this leads to the fo l lowing relationship, G x t 2 = A G y t l (2.27) This condition is true only when the full 2D data set S ^ . t j ) is processed completely. A s the reader wi l l note in the next sect ion, in practice data processing can be restricted to the region of interest, thereby reducing the need for large computer storage space. Under these circumstances, care has to be exerted to produce a symmetrical image. 2.3.5 D A T A PROCESSING The general processing routine is summarized in scheme 1. The acquired signal is a function of two variables, S(G y , t 2 ) . Figure 2.7 a illustrates the pattern of this acquired spin echo signal as n (equation 2.23) Scheme 1 K64H s(t,.t2) 64—A SCFa.tt) DATA HANDLING 1sh FT (i) Baseline Correction (ii) Trapezoidal Multiplication or Sine-bell Multiplication 2 n d FT (i) Sine-bell Multiplication (ii) Zero-fill Fig 2.7 (a) Pattern of the spin echo signal as n is stepped through, in the different blocks of the data set S(G v , t2) . (b) Fourier transformed spectra S(G V ,F2) . 37 Fig 2.8 (a) The pre-acquis i t ion information are reflected as phase variations in the echo- interferograms S(F2,G y ) . (b) Fourier transformed spectra with magnitude calculation S(F 2 .Y). is incremented. These data were obtained from a 4 cm diameter spherical glass bulb, f i l led with water. It is the phase of the spin echo signal that is varying, and this variation depends on the magnitude of the phase encoding gradient. Therefore in the second time dimension (ie: phase encoding dimension) n, which determines the magnitude of the gradient, behaves as a pseudo time variable. This signal is first apodized with either a s ine -be l l or a trapezoidal function so as to enhance the symmetry of the spin echo, in addition, any unwanted signals present at the beginning of the acquisition period induced by incomplete dephasing during the evolution period are also el iminated. Fourier transformation with respect to t 2 y ields spectra, S(G y ,F 2 ) , shown in figure 2.7 b. The frequency of the component lines in this spectrum reflect both the chemical shift information (5) and the spatial information content induced by the static gradient G x > It is convenient at this juncture to select the block of data which encompasses the frequency range of interest, since this minimizes the computer storage problems and saves memory space. It is important for the block of data so selected to have some integral relationship with the orthogonal digital dimensional i ty ; furthermore, because equation 2.27 is no longer val id , care has to be exerted so as to end up with a symmetric image. This selected region is then transposed to yield S(F 2,Gy), (figure 2.8 a), these resultant echo- interferograms now encode only the spatial information imparted by the phase encoding gradient, G . Signal conditioning is conveniently performed at this stage. S ine -be l l multiplication of the data is fo l lowed by interpolation by ze ro - f i l l ing to improve the effect ive digital resolution; recall that the resolution in this domain is determined by the number of experiments performed (equation 2.25). Magnitude calculation fo l lowing the second Fourier transformation produces the final absorption mode spectra, S(F2,y) (figure 2.9 b), which can be displayed either in this form or, after transposit ion, as S(y,F 2). The one used most in this thesis is the former in which the phase encoded dimension (eg: y) is displayed along the horizontal axis and the frequency encoded dimension (eg: F2) along the vertical axis. There are many ways of displaying images, the obvious choice being either a grey -sca le or a colorgraphic display (figure 2.9 C). In some cases, stacked-plot or the contour-plot display used conventionally in 2D Spectroscopy seem appropriate and this is illustrated in figure 2.9 A and B. The intensity levels in a contour plot display vary in an exponential manner with the inner levels representing higher intensities. The exact variation on the 80 MHz wide bore system is the same for both the contour plot and colorgraphic display and this variation scheme wi l l be discussed later in Chapter 6. The perceived intensity of an NMR image is not merely a 2D representation of the proton density p. Rather, it is a spatial representation of the NMR signal and a full appreciation requires an in-depth understanding both of the fundamental relaxation processes of the nuclear spins and the behavior of their magnetization when subjected to a particular sequence of the rf pulses. In general, the NMR signal from a volume (contd) Fig 2.9 (A) Stacked plot display of the 4 cm diameter spherical glass bulb containing Cu^ + doped water. (B) Contour plot display (8 levels) of the same. (C) Colorgraphic display of the same with a 8 level pseudo blue scale. element at x,y,z can be given by (52), [pF(T 1 ,T 2 ( v) ] x y z (2.28) The precise form of the function F, of the relaxation times T l t T 2 and the f l u i d - f l o w velocity v, wi l l depend on the method of measurement. It should, therefore, be borne in mind that there is no universal intensity scale in NMR imaging, and the image intensity is simply a display of the acquired NMR signal. In practice it is the contrast between the signals from different regions of the sample which is useful. Attempts to explain and evaluate contrast wi l l not be dealt in this thesis, and the reader is referred to the literature (53,54) for detailed theoretical evaluation of contrast, s i g n a l - t o - n o i s e ratio and related details. The next section summarizes the studies performed on model objects to evaluate and characterize imaging systems. A s wil l be seen, this led to the recognition of several individual components pertinent to NMR imaging and to NMR in general. 2.4 STUDIES OF PHANTOMS Although the production of NMR images using commercial ly available devices has already from the experimental point of view become clinical practice, there are no widely accepted protocols for evaluating the NMR imaging systems as such. Furthermore, since the cost of purpose-bui l t imaging apparatus is substantial, there is a continuing interest in information concerning the possibi l i ty of adapting existing high resolution NMR spectrometers for imaging studies. This requires experimental evaluation of several aspects of the spectrometer performance, such as the magnitudes and linearity of the magnetic f ield gradients, and the homogeneities of the static magnetic f ield and the rf f ie ld . Such evaluations are most easily carried out using "phantoms", which are objects of known dimensions containing liquids of known composi t ions . In this work these have been constructed from arrays of liquid f i l led glass tubes (capillaries, NMR sample tubes, spherical bulbs, vials etc.) mounted on suitable supports (teflon, styrofoam etc.), and then located inside the probe. A description of several such studies which form part of this work, constitutes the remainder of this section. 2.4.1 MAGNETIC FIELD GRADIENTS It is important that the magnitudes and the linearities of the magnetic f ie ld gradients be known as accurately as possible. Several schemes have been used to measure gradient magnitudes; theoretical calculations (55), measurement of the magnetic f ield as a function of posit ion (56), measurement of a known dif fusion coeff icient (55), and measurement of the gradient directly by means of its effect on the NMR free induction decay (28,57). More recently, gradient magnitudes have also been evaluated by lineshape analysis of NMR spectra of simple objects (58). The diff icult ies with a theoretical calculation are the necessity of a careful measurement of coi l posit ion and induced f ie lds . Measurement of the change in magnetic f ield as a function of posit ion of a small sample as it is moved in a gradient demands that the displacements be known accurately. Measurements based on a substance having a known dif fusion coeff ic ient suffer from the problem of requiring good reference data; water is used most often, but there is considerable inconsistency in the published values of its diffusion coeff ic ient . Lineshape analysis suffers from the disadvantage of being diff icult to implement when larger gradients and objects are involved. Direct measurement of very large gradients brings the problem of sampling a shor t - l i ved time domain signal, but this is the method chosen here for relatively small gradients involved. The use of phantoms to evaluate the magnitude and linearity of gradients is a convenient and simple method (59). Figure 2.10 shows NMR projection spectra of a phantom which comprised three 5 mm NMR tubes f i l led with water, mounted 15 mm apart in an accurately machined teflon plug, and located at the center of a 31 cm horizontal bore magnet operating at 1.89 T for protons. The observed projections in figure 2.10 A and B were obtained after good shimming of the static B 0 f ie ld , using the phantom with its long axis aligned along x and z axes respectively. From the known geometry of the phantom and the observed separations in the projection spectrum, it is possible to estimate the gradient magnitudes. In this particular case, the magnitudes of the x and z gradients were found to be 0.77 G/cm and 0.78 G/cm respectively. One simple method to evaluate the gradient linearity is to measure the widths at hal f - intensi ty of the three s ignals ; that the linewidths together with the spacings between the tubes were identical implies that over the total size of the phantom (35 mm), these gradients are linear within an experimental uncertainty ca. ± 5%. A 9780 Hz B 10,004 Hz c D 825 Hz 778 Hz Fig 2.10 Evaluation of the imaging gradients on the 1.89 T wide bore system using a phantom comprising three, 5 mm NMR tubes containing water and mounted 15 mm apart. The phantom orientation depends on the gradient used. The axis of the phantom is aligned along the direction of the appropriate gradient. (A) and (B) are one-dimensional projections obtained with G x and G z set at 0.77 and 0.78 G/cm respectively. (C) ai (D) are the equivalent x and z traces, but with the gradient magnitudes reduced by about 92 %. Note the non- l inearit ies in the lineshape reflecting the influence of the static field inhomogeneity. Fig 2.11 Evaluation on the 6.34 T narrow bore system using a phantom mounted inside a 10 mm NMR tube. The proton decoupler coi l (15 mm diameter, 15 mm height) in the 1 3 C- { 'H } probe was used as the observe co i l . The phantom shown in (A) and (D) was assembled from five glass capillary tubes (1.6 mm id) f i l led with water and held in posit ion by a Teflon plug. Images (B),(E) were measured using the x gradient and (C),(F) using the y gradient. The shim coi ls were used as the gradient coi ls and in each case, the gradient was set under software control . The above approach is subject to systematic errors if the axis of the sample is not coincident with the direction of the applied gradient. Although this is not a problem with the horizontal, wide bore magnet which has sufficient room for sample posit ioning, alignment inside the vert ical , narrow bore magnet is nontrivial . In this case the relative orientation of the phantom has to be obtained by finding experimentally the maximum frequency separation for a particular orientation of the phantom. This approach was applied to a 54 mm bore magnet which operates at 6.35 T, for which the gradients are obtained from the normal shim co i l s . The phantom used consisted of f ive capillary tubes of internal diameter (id) 1.6 mm f i l led with water, mounted in a teflon plug and placed inside a 10 mm NMR tube. A s seen in figure 2.11, two sets of such projection spectra for two different orientations were obtained, and in each case the gradient magnitudes were found to be 0.1 G/cm for both x and y shim coi ls . This value is in good agreement with that measured from lineshape analysis (58). Careful inspection shows that the observed splittings are not always equal; in typical measurements, differences of up to 6% have been observed. These deviations are attributable to the effects of static f ield inhomogeneity and demonstrate that the technique does not work well when those broadening effects are large compared with the gradients used (also see figure 2.10 C and D). However, for large gradients the method works well as illustrated in figure 2.10 A and B; the projections then obtained are equivalent to those which can be obtained with good static f ie ld homogeneity. This emphasizes the fact that in imaging the applied gradients must overcome the residual line broadening if d istort ion- f ree images are to be obtained (equation 2.18). Based on these findings, it is clear that the shim coi ls of the narrow bore system cannot be used to image "real" objects, since such objects exhibit broader signals and the above mentioned criteria cannot be sat is f ied. However, we shall see later on in section 2.4.4.2 how this system can be modif ied to give gradients with higher strengths. 2.4.2 STATIC FIELD EL, HOMOGENEITY The measurement of the B 0 f ield distributions is important in NMR imaging, since the image quality can be dependent upon the strength and homogeneity of the magnetic f ields used; furthermore, for imaging chemically shifted species, high B 0 homogeneity is mandatory. Methods of mapping the f ie ld distribution wi l l be discussed in this sect ion. The description starts with a review of the existing methods, fo l lowed by that of the author's, and concludes with results obtained on our instrument using the method described by Maudseley et a l . (67). Image restoration from the undesired artifacts arising from poor B 0 homogeneity wi l l not be discussed, and the reader is referred to the literature (60-65), for some computer simulation studies. However, in the next Chapter experimental ways of minimizing these artifacts in Chemical Shift Imaging are described. Direct observation of the f ield distributions during magnet shimming is frequently achieved by a point by point plotting of the f ield using a NMR probe; however this is extremely time consuming. Alternatively, imaging of phantoms can be used to map f ield distributions of different regions simultaneously. Maudseley et a l . (66) in 1979, proposed an effect ive method to do this. The method is a simple modif ication of the Fourier zeugmatography technique (section 2.2). The accurate measurement of magnetic f ields requires that no externally applied magnetic f ield gradients be present during the acquisition period; however for spatial select iv i ty , gradients must be applied to encode various spin volume elements. A pulse sequence which satisf ies both these requirements is shown in figure 2.12. The evolution of spins during t^ when no gradients are present, is determined by the f ield inhomogeneities present over the spin volume; Fourier transformation of the acquired signal with respect to t^ yields the spatial distribution of the f ie ld . Clearly, either 2D or 3D field measurements are possible . More recently, in 1984, the spin warp imaging technique (section 2.3), and variations thereof, have been used to map f ie ld distributions of different regions (67-69). It should be recalled that spatial encoding is achieved by means of amplitude modulated gradients, and in this particular case, the spin echo signal formed by 180° rf refocussing pulse at the end of the evolution period is acquired in the absence of any gradient. This method possesses additional advantages over the former, in that the static f ield inhomogeneities are present only during the period of detection, whereas in the former they influence other time intervals as we l l . A full 2D or 3D f ield measurement is not always required, and it may be sufficient to measure the f ield over a specif ic volume. Alternatively , a one-dimensional projection spectrum of suitable phantom can yield information concerning the homogeneity, of the static f ie ld , provided the gradients are linear and capable of overcoming the static f ield inhomogeneity. Under these condit ions, the lineshape of the projection view gives a direct reflection of the B 0 homogeneity (59,70). 2.12 The pulse sequence used to measure the magnetic f ie ld distribution. The evolution of spins during interval t l t after initial excitation by a 90° pulse, is determined by the static f ie ld homogeneities, since no gradients are applied. The rest of the sequence is similar to the Fourier Zeugmatography sequence. [ From reference (66) ] An example of this approach used in this work is given in figure 2.11, which shows that over the volume of the phantom,-which consists of 5 capillary tubes (1.6 mm id) filled with water, the x and y axes have a Q reasonably homogeneous field; this is estimated to be 1 part in 10 . Interestingly, along the z-direction, B 0 homogeneity falls off rapidly as illustrated in figure 2.13 which represents the projection spectrum of a teflon plug with carefully drilled holes (1.2 mm diameter) filled with acetone. Thus, it is clear for this particular 6.35 T narrow bore magnet that the z shim is less effective for imaging than either the x or y shims. This approach is quick and simple. However, if simultaneous mapping of field distributions of different regions are required then the procedure of 2.3 mm o o o o o o o o o oo o 2.3 mm B Fig 2.13 The phantom (A) is machined from a Teflon plug with 1.2 mm diameter holes spaced every 2.3 mm along its length, and filled with acetone. This set-up is mounted inside a 10 mm NMR tube and placed inside the narrow bore system. The one-dimensional image in (B) was measured using the maximum value of the z gradient (ie: z shim). Maudseley et a l . (67) can be applied. Depending on the type of f ield measurement required, an appropriately shaped phantom can be used to provide the NMR signal. Field variation along the axial direction of the magnet can be measured by choosing as the object, a single narrow tube (5mm id) containing 10 mM copper sulfate solution and aligned along the z axis, the direction of B 0 . Two-d imensional imaging sequence with phase encoding along the z -d i rect ion , yields a two-d imensional image with one spatial dimension z, and the f ield variation (or frequency) along the other. Figure 2.14 A - D shows field variations along the z axis of the 1.89 T wide bore sys tem, for different settings of the axial shims Z 1 , Z 2 , Z 3 and Z 4 respectively. It is clear from these images that higher order axial shims contribute radial terms. Field distributions in the xy plane can be measured by using a thin circular disk phantom (6 cm diameter, 1 cm thickness) containing the same solut ion, positioned in the central transverse plane of the magnet. Three-dimensional imaging sequences (section 3.2.3) with phase encoding along x and y axes permit mapping the f ield in the xy plane. Several sl ice images from a full 3D data set are shown in figure 2.15; the difference between the successive planes is 1 ppm of the main f ie ld . The f ield distortion in this particular case was deliberately introduced by offsett ing the x 2 - y 2 shim. In a similar fashion, the effect of incorrectly setting the xy shim is shown in figure 2.16 using the same phantom. 53 F\e\d p\ots a\ong the z axis showing the effect of the axial shims on the wide bore system. (A) Z 1 (B) Z 5 x is given by, A w x = 7 G x x (2.31) where x is the length of the object and G x is the strength of the x gradient. In order to spatially resolve an object, the applied gradient must overcome any intrinsic broadening effects (equation 2.18). In addition, it is seen from equation 2.31 that to observe microscopic objects the gradient strengths have to be increased. Hence, when imaging microscopic objects with large natural l inewidths, higher gradients are mandatory. The shim coi ls used for imaging with the 6.35 T narrow bore magnet available in this work are capable of producing linear gradients of approximately 0.1 G/cm. This is sufficient to enable imaging of axially symmetric phantoms with a spatial resolution ca. 0.5 mm (see figure 2.18 A) . In the work of another student in this laboratory (S.D. Luck), a standard high resolution NMR probe was modif ied and fitted with new gradient coi ls to produce orthogonal gradients of higher strength. The x, y, and z gradient coi ls were wound on the outer surface of a perspex cylinder, which was then mounted inside the conventional high resolution probe. These gradient coi ls were capable of generating gradients of 3.7, 3.9, and 2.4 G/cm respect ively . Two-d imensional imaging was performed on glass capillary tube phantoms f i l led with water and mounted inside this probe; the images show a spatial resolution better than 50 jtxm (89). The results shown in figure 2.21 were obtained in collaboration with Mr. S.D. Luck, and the images are represented as contour plots for three sets of phantoms. Figure 2.21 A shows the image of four melting point capillaries (1.2 mm id) and figure 2.21 B represents that f rom four finer capillaries (250-300 (im id). The four capil lary tubes in each case were mounted inside a 5 mm NMR tube. The image illustrated in figure 2.21 C is from a phantom assembled inside a 1.2 mm id melting point capillary tube, which in turn is placed inside a 5 mm NMR tube. The larger of the capillaries has an inner diameter of ca. 220 Aim, the smaller 140 nm. It can be seen clearly that in addition to being able to resolve ca . 15 Mm in figure 2.21 C, the resolution of the four capillary phantom, particularly that in figure 2.21 A , is far better than that obtained using the shim coi ls on a similar phantom (see figure 2.18 A). The fact that these gradients are 10 - fo ld greater than those from the shim coi ls makes many imaging experiments of "real" objects feasible. This interest in microscopic imaging is evidenced by the reports of three groups of workers appearing almost simultaneously. These include Fig 2.21 Two-dimensional spin warp variant images of glass capillary tubes containing water, located inside a 5 mm NMR tube. (A) 1.2 mm id tubes; (B) 2 5 0 - 3 0 0 nm id; 1.2 mm od; (C) 1 4 0 - 2 2 0 ^m id. E)qxrimental parameters ; (A) sweep width = 4 0 0 0 Hz; acquisition time = 6 4 . 1 2 ms; number of scans = 4 ; relaxation delay = 4 s ; gradient increment = 0.05 G/cm; static gradient = 0.9 G/cm. (B) gradient increment = 0.06 G/cm; static gradient = 1.17 G/cm; (C) sweep width = 6 0 0 0 Hz; acquisition time = 4 2 . 5 7 ms; gradient increment = 0.06 G/cm; static gradient = 3.89 G/cm. o imaging of a rat brain (90), a plant stem (91), and a single cell (92), the spatial resolution reported in the latter two being 20 nm and 13 Mm respectively. The major l imitation to microscopic imaging is the s i g n a l - t o - n o i s e consideration; as the image-element (pixel) gets smaller a corresponding loss in signal occurs. Hence, signal averaging in microscopic imaging is mandatory and this leads to longer experimental t imes. However, imaging at 71A higher f ields should yield benefits as the s ignal -noise increases as CJ^ (93). Another criterion that has to be considered in the design and use of higher gradient strengths is the rise t imes, which have to be quite short. In this particular work, the gradient rise times were determined experimentally and are ca. 10 ms. CHAPTER 3 MULTI-DIMENSIONAL CHEMICAL SHIFT IMAGING 3.1 INTRODUCTION One NMR parameter in which cl inical interest has grown rapidly over recent years is the chemical shift. The phenomenon of chemical shift arises because of shielding of the nuclei from the external magnetic f ie ld , B 0 , by the surrounding electrons (12). This shielding can be taken into account by considering an effective f ie ld , B g ^ , at the nucleus given by, B e f f= B 0 (1 -a ) (3.1) where a is called the shielding constant of the nucleus. This modif ies the precise resonant frequency to , w = 7B.(1-a) (3.2) Thus, high resolution NMR spectra which display chemically shifted resonances provide information on the chemical species present in the system. The importance of the chemical shift in life sciences was realized by many early investigators studying biochemistry (94,95). NMR spectroscopy has been used to study enzymatic reactions (96,97), measure intracellular pH (98-100), study cell membranes (101-103), fo l low metabolic pathways of different states (104-107), and to monitor pharmaceutical interventions (108,109). Although early observations using NMR spectroscopy were limited to in -v i t ro measurements, more recently, techniques have been developed for acquiring metabolic data in - v i vo using NMR (109,110). One of the favorite method involves the use of surface coi l technology, in which a small receiver coil is used to acquire signals from small localized volumes within intact biological samples. Concurrent with these developments in in - v i vo NMR spectroscopy, NMR imaging has grown rapidly in recent years as discussed in Chapter 2. The reader wi l l note that, on the one hand, NMR spectroscopic studies require a highly homogeneous static magnetic f ield in order to monitor and categorize functions at the cellular levels. On the other hand, in NMR imaging, which offers superb anatomic detail and spatial information, the magnetic f ield gradients used to spatially encode the NMR signal to form the image automatically disrupt the f ield homogeneity fundamental for spectroscopic studies. Chemical Shift Imaging (CSI) combines these two approaches, thereby adding the spectral dimension to the spatial dimension (111). In principle, such approaches al low either NMR spectra to be extracted from all points within the image matrix, or alternatively separate images each corresponding to one of the chemically shifted resonances may be displayed. Although the initial work on CSI has been pioneered by several workers on suitable systems (41,44,64,112,113), the first application to human was demonstrated only in 1983 by Pykett et a l . (114). These workers obtained separate water and fat images from a human forearm using a three-dimensional Fourier transform imaging technique. The early work on proton imaging of human samples was done at suff ic ient ly low magnetic f ields (= 0.15 T) that the frequency separation between the water and fat components are very smal l . By design the gradients used were large enough to overcome these minor frequency differences, so in effect , the frequency separation between the water and fat resonances are not differentiated; hence, the final image represents a combined proton distribution of both these components. At higher magnetic f ields 1.5 T), however, these resonances are wel l resolved and now CSI becomes feasible. Chemical shift images have been obtained by a variety of techniques, the most widely used methods being image formation by mult i -d imensional FT techniques. These include phase encoding each spatial dimension with gradients and acquiring the signal in the absence of any f ield gradients (44,114-119). Other techniques described in the literature are; a modif ied projection reconstruction method, in which the magnetic f ie ld gradient strength is varied to cause a corresponding change in the relative effect ive spatial offset (64,112,120,121); projection reconstruction at higher magnetic f ields (41); use of rf gradient f ields which retain high static f ield homogeneity (113); varying the temporal offset of the spin echoes formed by 180° rf pulses relative to those formed by magnetic f ield gradient reversal so that the phase of different spectral lines vary accordingly (122,123); selective excitation of the desired lines (124,125); selective saturation of the undesired lines (35,124,126,127); encoding chemical shift information in a similar method to that used in high resolution 2D NMR homonuclear correlation (128) and scalar J coupling (49) experiments. More recent techniques include the use of stimulated echoes to excite resonances selectively (129) and the use of relatively long 180° rf pulses to form selective spin echoes, whereby only the desired frequency wi l l experience the 180° rotation to form the spin echo (34). Most of the discussion in the fo l lowing sections pertains to mult i -d imensional CSI. Initially, sequences relating the chemical shift with one and two spatial dimensions are discussed, together with their advantages and l imitations. The latter part discusses the ultimate form of CSI which was developed during the present study (119). A brief insight into chemical shift artifacts is also included to enable a clear understanding of the possible distortions observable by neglect of the chemical shift parameter. The principal applications derived from these techniques, however, wi l l be described in Chapter 4. 3.2 IMAGING SEQUENCES The original Fourier zeugmatography technique of Kumar et a l . (17) encodes chemical shift information along both dimensions in addition to the spatial information (section 2.2). With the advent of phase encoding gradients, chemical shift information can be suppressed in one or more of the dimensions. This, then only yields spatial information along the phase encoding dimensions (section 2.3). Thus, spectroscopic imaging can be performed, provided that the chemical shift differences are greater than the frequency dispersions caused by the magnetic f ield gradient applied during data acquisit ion. That is , AS > A w x ( = 7 G x x ) (3.3) For water and fat resonances this condition is satisf ied easily at very high magnetic f ields of the order of 6.35 T. For lower f ields (— 1.89 T), however, the gradient during the acquisition period destroys the chemical shift information. From this stems the necessity to switch off the read gradient so as to preserve this useful information. The simplest form of CSI is to phase encode only along one spatial dimension and this sequence is described in the next sect ion. 3.2.1 TWO-DIMENSIONAL CHEMICAL SHIFT IMAGING: x,5 The basic pulse sequence used here is represented in figure 3.1. The NMR signal is acquired in the absence of any magnetic f ield gradients. The result of turning all gradients off during the acquisition period is that spatial information can no longer be frequency encoded. Hence, a single phase encoding gradient, G is used to differentiate spins spatially along the x -d i rect ion . The observed signal S(t) can be represented using the formulation developed in Chapter 2 by, S(t) = S(tut2) = 5 / p 5 (x) exp i [ 7 G xtj + S . t Jdx (3.4) where Pg(x) is the one-dimensional spin density distribution for nuclei with chemical shift 5. The 2D FT of S(t) is denoted by, S(u) = S C U i P , ) = S(x,6) = Z / p 5 (x) F f rG x - u , ) F(6.-o>2)dx (3.5) 1 i This data matrix represents one dimension exhibiting intrinsic chemical shift differences 5 , and the other exhibiting spatial differentiation along the x axis. To illustrate this approach two vials (1.3 cm id) containing water and acetone were imaged using this sequence; the resultant images are shown in figure 3.2 as stacked plots. 128 increments of G were used and the final image matrix size is 128x128. Delay interval 9 0 e Image -development Data acquisition I80 c Gradient X Fig 3.1 Schematic diagram of the pulse and gradient sequence used for two-dimensional x, 5 imaging. oo A 79 Fig 3.2 Two-d imensional chemical shift resolved images of 2 vials containing water and acetone respectively. One dimension represents the spatial dimension, while the other represents the chemical shift . Transposition of the image in (A) y ie lds that in (B), which represents the view along the chemical shift dimension. Note the curvature present in this image. Experimental parameters \ sweep width = ± 2000 Hz ; acquisition time = 128 ms ; block size = 1024 ; gradient increment = 0.016 G/cm ; no of experiments = 128. It is noticeable that the image data corresponding to a particular chemical shift are distributed about a curved surface, the precise shape of which is determined by the distribution of the nonuniformities in the static magnetic f ie ld . A discussion of distortions and the ways of minimizing them wi l l be dealt with in the latter part of this section on imaging sequences. These distort ions, however, are not prohibitive and studies of other nuclei (eg: P -31) have sucessful ly yielded information, especial ly on phosphorus metabolites from different spatial regions of t issues in - v i vo (115). 3.2.2 TWO-DIMENSIONAL CHEMICAL SHIFT IMAGING: y,(x6) The pulse sequence used here has been described previously for a single resonance in section 2.3. This imaging scheme is capable of producing chemical shift images, only if the chemical shift separations between resonances are greater than the gradient induced dispersion, since the signal is acquired in the presence of a static gradient (equation 3.3). The final image has one spatial dimension (y) and another combined chemical shift (5)/spatial dimension (x). To illustrate this approach, figure 3.3 represents images from four capillary tubes (1.2 mm id), two containing acetone and two containing benzene respectively, placed inside a 5 mm NMR tube and studied on the 54 mm narrow bore magnet operating at 6.35 T for protons. It is seen that the images corresponding to the two resonances are well resolved. Further illustration of this technique is seen in figure 3.4, which represents images from eight such capillary tubes arranged in an alternating fashion, and Fig 3.3 Two-dimensional chemical shift resolved images of 4 capillary tubes mounted inside a 10 mm NMR tube, two containing water and two containing benzene respectively, are displayed as stacked plots (a) and contour plots (b). The image was obtained on the narrow bore system. Experimental parameters ; sweep width - 4000 Hz ; acquisition time = 128.12 ms ; block size = 1024 • gradient increment = 0.004 G/cm ; no of experiments = 32 ; relaxation delay = 10 s. co Y A Fig 3.4 Two-d imens ional chemical shift resolved images of a phantom consist ing of 8 capillary tubes, four containing water and four containing benzene. This image is again from the narrow bore sys tem. The intensity variations are due to the inhomogeneity of the B, f ield of this particular probe. Experimental parameters ; sweep width = 4000 Hz ; acquisition time = 64.12 ms ; block size = 512 ; gradient increment = 0.007 G/cm ; no of experiments = 32 ; relaxation delay = 10 s. 83 Fig 3.5 The effect of increasing gradient strength is illustrated on the same phantom as that used in figure 3.3 but contained inside a 5 mm NMR tube. (A) gradient strengths of 0.1 G/cm. (B) gradient strengths of 2 G/cm (C) gradient strengths of 4 G/cm. These are all two-dimensional chemical shift resolved images, obtained on the narrow bore system. Note that the information is mixed up in (C). This is because, the gradient induced dispersion is greater than the chemical shift separation between the two resonances. placed in a 10 mm NMR tube. The intensity variations reflect the B L inhomogeneity of the 10 mm probe used in this study (section 2.4.3). Figure 3.5 illustrates the effect of increasing the gradient strengths, on the phantom containing four capillary tubes, so as to violate the condition given in equation 3.3. However, the capillary tubes are mounted inside a 5 mm NMR tube. The image in figure 3.5 A is similar to that in figure 3.3 but is represented with chemical shift along the horizontal axis. Images 1 and 2 are that of benzene and 3 and 4 are that of water. With gradient magnitudes of 2 G/cm for the 5 mm sample, the expected image is seen (figure 3.5 B). But when the gradient magnitude is increased to 4 G/cm the two images overlap creating confusion (figure 3.5 C). This emphasizes the fact that the gradient magnitudes should be known when imaging with this particular sequence. A useful extension of this technique under these circumstances is to use some other inherent property of the individual resonances and to combine this property together with the imaging sequence. For example, if two resonances have different sp in - lat t ice relaxation t imes, then inversion-recovery sequence (130) can be combined with the imaging sequence. This enables one resonance to be observed selectively while the other is at its null. Selective excitation or selective saturation can also be combined with this imaging sequence to yield chemical shift resolved images (124). 3.2.3 THREE-DIMENSIONAL CHEMICAL SHIFT IMAGING: x,y,6 The generalized pulse sequence is given in figure 3.6. The NMR signal is acquired in the absence of any applied gradients. Phase encoding gradients are applied along both the x and y directions for spatial differentiat ion. Three-dimensional FT of the experimental data S(G x ,Gy,t 2 ) , produces a matrix with three dimensions, two representing x and y spatial axes, and a third containing the chemical shift information. The images are usually displayed as two-d imens ional xy spatial plots corresponding to a particular chemical shift 6. Figure 3.7 A and B illustrate xz projection images (since G and G X z were used for phase encoding) corresponding to the individual chemical shifts of water and acetone contained in 1.3 cm id vials respectively. These images have been computed from a total of 1024 data sets, corresponding to 32 increments of each gradient. The loss in definit ion of the phantom is due to the low digitization along the phase encoding dimensions. It is seen that the images are distorted due to static magnetic f ield inhomogeneity AB 0 . The rest of the discussion in this section focusses upon the origin of this distort ion, the methods of minimizing it experimentally and an illustration of what can be achieved with highly homogeneous static magnetic f ie lds . Let us consider a three-dimensional image matrix obtained after triple FT; as mentioned earlier, two dimensions reflect the spatial axes while the third reflects the chemical shift dimension (figure 3.8). It can be seen that the image data for the two resonances, S . a n d 5 R , are distributed about Delay interval Image-development ' | 8 0 o Data acquisition Gradient X Gradient Y Fig 3.6 The pulse and gradient sequence to perform three-dimensional chemical shift imaging. The spin echo signal is acquired after a composite 180' pulse in a homogeneous field, ie; all gradients are switched off during data acquisition. 00 CT> B >• x 3.8 cm 1.8 cm Fig 3.7 Two-d imens ional xz planar images obtained from three-dimensional data set of a phantom, which comprises of a vial of water and a vial of acetone are shown in (A) and (B) respectively. 32 increments of each gradient G x and G z were performed. Experimental parameters \ same as that for figure 3.2. 88 Fig 3.8 An object with two chemically shifted resonances and its three-dimensional image matrix are represented. Two axes for spatial information (x and y) and a third for frequency information (6). At the left, a conventional image, obtained by summing information along the 6 axis is seen. Note, that the spectral information can be extracted from any spatial x,y coordinate, or vice versa. The dark lines represent the curvature of the image planes due to distortion. [From reference (111)] curved surfaces. The width of the distribution on either side of this surface depends on the linewidth of the particular peak. Furthermore, the precise shape of the surface is defined by the static field inhomogeneities within the sample volume observed (114). This situation can be visualized easily by looking at the image in figure 3.2 in section 3.2.1, which is a two -d imens iona l analogue of this x, y, and 8 sequence. The projection of the image in figure 3.2 B across the chemical shift dimension yields the high resolution spectrum of the two resonances, whose linewidths are determined by the normal l ine-broadening mechanisms as in conventional NMR spectroscopy. These include contributions from the intrinsic relaxation properties as wel l as the static f ield inhomogeneity. Thus, it is clear that the shape of the curvature in this chemical shift image is determined only by A B 0 , although both contributions determine the width of this curved surface. This distortion wi l l be present in all similar imaging sequences used to acquire chemical shift resolved images. However, if the B 0 f ield is homogeneous over the sample then the distortion is minimized. It should be recalled that this provided the basis for mapping the magnetic f ield distribution discussed in section 2.4.2. There are various methods available to minimize this distort ion. One is to integrate the NMR spectrum by summing up all the reconstructed image data planes passing through that particular resonance peak. Further improvement can be obtained by decreasing the digitization along the observing domain. In certain instances, combination of these methods have to be applied. In this work, for s impl ic i ty , these correction features are illustrated by considering a single resonance. Figure 3.9 represents two-d imensional x, 6 images, displayed as stacked (A & C) and contour (B & D) p lots , of two sets of vials (1.3 cm id) containing water doped with copper sulfate (10 mM) and manganous sulfate (1 mM) respectively. Note the absence of the distortion in figures 3.9 C and D. This can be explained as f o l l o w s ; the 2+ 2A-linewidth of the Cu solution is 11.7 Hz while that of the Mn is 45.1 Hz. This implies that the percentage contribution of A B 0 to the observed linewidth is smaller in the case of the latter solution since the broadening contribution from its natural T 2 relaxation is higher. Furthermore, using a decresed digital resolution along the observing domain suppresses this distort ion. This is because the frequency spread between two digitally sampled data points (Hz/point) is greater than the frequency spread caused by AB 0 . Combination of both these factors eliminates the distortion from figures 3.9 C and D. However it should be noted that a resonance peak has to be characterized by a minimum amount of data points and hence, the digital resolution just cannot be made arbitrarily low. The effect of summing several images is illustrated in figure 3.10. These images have been obtained from a three-dimensional x, z, 5 imaging sequence and represent xz planar views displayed as stacked plots and contour plots respectively. The phantom consists of two similar vials containing water doped with M n 2 + ions. Figure 3.10 A has been obtained from a single plane taken through the resonance peak (ie: along the 6 dimension), and the distortion is apparent in the image. However, it is seen that in the integrated image in figure 3.10 C obtained by summing images 91 F«9 3.9 obtained using T w o - P e n s i o n . , c ^ c s , X L * two systems are 7 8 m / P | § t ,„ » these two the curvature is . same as that dimension int absent in the latter. m figure 3.2. 92 Effect of summing up data sets to minimize distortion is illustrated by means of a three-dimensional imaging experiment. A and B represent stacked and contour p\ol display of a 2D xz planar image obtained from a single plane along the chemical shift dimension respectively. Similarly C and D represent 2D xz planar image obtained by summing images from two different planes along the chemical shift dimension. f rom two adjacent planes taken through the same peak, the distortion has been reduced. A s stated before, this distortion wi l l be absent if the static magnetic f ie ld is very homogeneous and to achieve this calls for long and extensive magnet shimming procedures which are laborious. However, in the case of narrow bore high resolution magnets with small regions of very high homogeneity, CSI can be performed with a high degree of resolution as illustrated by Hall and Sukumar (117). Figure 3.11 illustrates 3D chemical shift images of a phantom consist ing of two capillary tubes (1.6 mm id) containing water and two of ethanol mounted inside a 5 mm NMR tube and placed inside the 54 mm narrow bore magnet. Figure 3.11 A shows the composite spectrum of the phantom. The contour plot display in figure 3.11 B shows the location of the two tubes of water and that in figures 3.11 C and D show the high f ield and the central transition of the methyl triplet; it is clear that these two ethanol images are identical, as they should be. Obviously such high resolution can only be achieved at higher f ields (270 MHz for protons) and with a high degree of magnetic f ie ld homogeneity 9 (1 part in 10 ). Though we cannot achieve this degree of homogeneity on our wide bore magnet over a larger sample, we have seen that CSI can be performed with moderate toleration of AB 0 . Most imaging magnets, in fact, can be shimmed to perform studies of this nature. The three-dimensional CSI sequence has been widely used in both Proton and Phosphorus imaging, but applications of P -31 imaging in human studies have yet to be described. The versati l ity of this technique stems from the fact that extensions and modif icat ions are easi ly achieved; for Fig 3.11 (A) shows the conventional high resolution NMR spectrum at 270 MHz, of a phantom consist ing of two capillary tubes with water and another two with ethanol. The - O H and - C H } signals were deliberately folded over to decrease the spectral width in the chemical shift dimension. The images in (B), (C), and (D) represent two -d imens iona l xy planar display of water and the two high f ield components of the methyl triplet respectively; the latter demonstrates the chemical shift resolution attainable on our narrow bore system. 32 increments of each gradient G x and Gy were acquired. The final matrix display size is 64x64. [ From reference (117) ] example, an inversion-recovery sequence can be incorporated to yield T t sensitive images and specif ic applications with such modif ications wi l l be described in Chapter 4. 3.3 FOUR-DIMENSIONAL CHEMICAL SHIFT IMAGING: x,y,z,8 The ultimate form of CSI is an experiment that generates a single four -d imensional data matrix which encodes the three spatial dimensions x, y, z along with the complete chemical shift information 6, for every volume element within the data set. The experimental sequences required to achieve this are shown in figure 3.12. The nonselective rf pulse during the preparation period is used to excite spins which then evolve under the influence of three phase encoding gradients applied along the x, y and z axes. Chemical shift information is encoded by acquiring the signal in the absence of any gradients (figure 3.12A). Alternately, a 180° rf refocussing pulse can be applied at the end of the evolution period so as to generate a spin echo, which is then acquired as before (figure 3.12 B). The signal coming from a particular volume element and a particular chemical shift from a 4D sequence wi l l be, S'(G k ,t 2) = S e x p i f r t ^ - k + o y j (3.6) Relaxation ef fects have been neglected and k=x,y,z. The observed signal S(t) which is a contribution from ail volume elements is given by, S(t) = S ^ t , ) = Z ; ; / p 5 (x,y,z) S'(G k ,t 2)dxdydz (3.7) 1 i where Pg(x,y,z) is the spin density distribution for nuclei with chemical shift 8. Delay interval image-9 C j „ development Gradient X -Gradient Y Gradient Z Data acquisition B Delay interval Image -9 0 o development ! | 8 0 „ Data acquisition Grodient X -• — Gradient Y Gradient Z Fig 3.12 The pulse sequence for 4D chemical shift imaging. (A) FID acquisition (B) spin echo acquisition after a composite 180° pulse. The 4D FT of this signal can be denoted by, S(x,y,z,5) = 5///P 5 F ( 7 G k k-co k )F (5 r co 2 )dxdydz (3.8) / In this data matrix the resonance frequency CJ^ wi l l represent the three spatial dimensions and the resonance frequency a>2 wi l l represent the chemical shifts . Thus, the signals corresponding to a particular chemical shift represent the 3D spatial distribution of that chemical species, which can then be processed in a variety of ways , f inally leading to, for example, an xy planar image corresponding to a particular z s l i ce . Figure 3.13 A shows the normal spectrum of a phantom consist ing of a vial (2.5 cm id) containing chloroform and water/deuterium oxide (1:1), and the final 2D images (64x64) within the xy plane extracted from the 4D data set are displayed as contour plots, in figures 3.13 B and C respectively. The choice of parameters for data acquisition and processing are essential ly the same as described before in sections 2.3.3 and 2.3.4, with logical extensions. The complete data processing routine is summarized in scheme 2. The acquired signal is apodized and Fourier transformed with respect to t 2 , yielding a data matrix of the form S(G x ,Gy ,G z ,F 2 ) . Then a slice through a particular chemical shift is chosen in F 2 and the whole data matrix is transposed to give information concerning the three spatial dimensions, S(G x ,Gy ,G 2 ) , for that particular chemical shift . Fol lowing apodization and data interpolation to improve the digital resolution, a second FT with respect to G z , y ields a data matrix of the form S(G x ,Gy,z) . If another sl ice is chosen, this time along the z spatial dimension, and the data matrix transposed, information regarding the two spatial dimensions, S(G x ,Gy) , for 98 Fig 3.13 (A) shows the high resolution NMR spectrum at 80 MHz of a phantom comprising of a vial containing CHCI 3 and H 2 0 / D 2 0 . (B) and (C) represent individual xy planar images of CHCI 3 and H 2 0/D,0 respectively, extracted f rom a 4D data set. A total of 4096 experiments were performed corresponding to 16 increments of each gradient G x , G y , and G z . The final image display size is 64x64. 99 Scheme 2 SCGx.Gy.Gz , t 2 ) prp * S ( G x , G y , G 2 ,F 2 ) 2 n d F T I Transpose T at *, S(G x ,G y ,Z ) Transpose . r at z1 3 r d G x t F T — Gy S ( G X f G y ) I, 4th F T S(y.x) JTranspose fx S(x.y) t - • G z S(G X ,G y ,G Z ) G x t STGX.Y) Transpose that particular z sl ice is obtained. This amounts to selecting a volume slice of the whole object along the z -d i rec t ion , digitally without use of sl ice selection routines based on selective excitation. A third Fourier transformation with respect to G y fo l lowing apodization, data interpolation, yields a data matrix of the form S(G ,y). This matrix when transposed, signal conditioned and Fourier transformed for the fourth time with respect to G yields the final data matrix S(y,x). This can be displayed either as it is , or transposed to give S(x,y), which now represents the 2D spatial distribution in the xy plane corresponding to a particular z sl ice of one of the chemically shifted resonances in the original data S(G x ,G y . ,G z , t 2 ) . In this measurement a total of 4096 data sets corresponding to 16 increments of each gradient were acquired with a total experimental time of » 150 min. Thus the main l imitation of the experiment is the overall duration. It is instructive to contemplate the outcome of acquiring a data set with adequate digital resolution. Table 3.1 summarizes the effect of dimension size on the experimental time for mult i -d imensional imaging methods. The values shown have been calculated for a s ing le - scan , 1 s pulse repetition. Based on this, if the final display is to include an adequate definit ion of the actual shape of the object, then a final display of 128x128 derived from a 32x32 data set is the absolute minimum; the overall time of 9.1 hrs is acceptable to study certain non-medica l systems. Clearly though, any substantial increase in spatial resolution rapidly results in totally unacceptable t imes. However, it is important to note that not all studies require equal digital resolution in all three spatial dimensions. 101 TABLE 3.1 The Effect of the Increase in Dimension Size on the Total Experimental Time for Multidimensional NMR Imaging Methods. Dimension Size 16 32 64 128 1D 1 s 1 s 1 s 1 s 2D 16 s 32 s 64 s 128 s 3D 4.3 min 17.1 min 1.2 hr 4.6 hr 4D 1.2 hr 9.1 hr 3 days 24.3 days Hence, resolution in one dimension can be sacrificed in order to improve resolution along the other dimensions. For example, a 64x64x8 data set can replace a 32x32x32 data set, whereby the resolution along x and y are doubled at the expense of a four - fo ld reduction along z. The foremost advantage of this technique is that the whole object is observed and a true chemical shift image is obtained. Since slice selection is achieved digitally, very narrow slices can be imaged. For inanimate objects like rock core samples bearing oil , this technique would produce a wealth of information in just one experiment. The volume of the object that can be studied using this approach is limited by the dimensions of the sensitive volume which encompasses 102 suff ic ient ly high B„ homogeneity. Large data storage faci l i t ies together with fast array processors are mandatory for performing 4D imaging. The most important future prospect stems from the use of rapid imaging methods based on multiple spin echo formation (131,132) which, by giving n v iews of the object for every pulse delay can, in principle, provide a /7 -fold reduction in the overall acquisition t ime. Moreover, incorporating shorter preparation pulses can cause a substantial reduction in the experimental t imes since longer relaxation delays are avoided (133,134). These modif icat ions, though promising, are yet to be demonstrated. 3.4 CHEMICAL SHIFT ARTIFACTS In low f ield clinical NMR imaging, it is generally assumed that the chemical shift separation between water and fat resonances are negligible since the gradients employed are large enough to overcome these separations. Hence, the observed image is a composite of both water and fat components. However, one may predict that the presence of non-negl igible chemical shifts can cause distortions in the final NMR image especially at higher f ield strengths. Hricak et a l . (135) in 1983, saw an artifact in the kidney images of humans and they termed this a "linear edge artifact". In 1985, two groups of workers independently demonstrated the presence of this artifact in model phantoms and in clinical scans at low fields of 0.26 T and 0.35 T respectively (136,137). They attributed this to the chemical shift differences of the different species present in the system under study. They also pointed out that asymmetries in the images are found particularly near the boundaries of some organs having fatty deposits bordering predominantly wate r - f i l l ed compartments. The compartmentation al lows the chemical shift differences between the water protons and the mobile lipid protons of the adjacent fat, to distort the spatial information. In these planar 2D FT imaging methods, the spatial information is encoded in terms of frequency along one axis and in terms of phase along the other (section 2.3.2). Thus anomalous frequency changes caused by chemical shift differences between structures is displayed in the image as artifactual displacements along the direction of frequency encoding; ie: the observe dimension. In the work of this thesis, this was illustrated as fo l lows . The composite image of a phantom consisting of a vial (1.3 cm id) containing equal amounts of water and motor oil (10W30) is shown in figure 3.14. It can be noted that the image of the oil is slightly shifted along the observe dimension relative to the water image. This shift of 268.8 Hz is almost equal to the chemical shift separation between the water and oil resonances which is 302.1 Hz (3.8 ppm); this slight discrepancy is due to the uncertainty of each digitally sampled point in the image, since the digital resolution is calculated to be 63.0 Hz/point. Depending on the relative orientation of the phantom with respect to the frequency encoding gradient, different effects caused by these displacements can be observed. Thus regions of high and low intensities are clearly visible when G y is employed as the observe gradient, as illustrated in figure 3.15. The image in figure 3.15 A was obtained with posit ive value of G y and that in figure 3.15 B with negative value of G y . This amounts to inverting the orientation of the sample with respect to the gradient; the effect of this is illustrated in figures 3.16 A and B respectively. The precessional frequency change of a species under the influence of a magnetic f ield gradient (in this particular case G y ) is given by, Aco y = ? G y A y (3.9) Let 6 be the chemical shift difference between the water and oil resonances. If it is assumed that the phantom is of length 2y, and it contains equal amounts of water and o i l , then the fo l lowing relationships hold for figure 3.16 A. ^ w a t e r = ?V ( 3 J 0 ) ^ o i l = 7 G y Y + 6 ( 3 J 1 ) This results in the projection of the oil resonance being displaced from that of the water resonance by an amount 6, and therefore the image is displaced as seen in figure 3.15 A. However, the opposite effect occurs in the other orientation. Since the oil resonance is observed first with respect to the gradient, as seen in figure 3.15 B, the change in the precessional frequency modify as, ^ o i l = 7 G y y " 6 < 3- 1 2) Ad) „ = 7G y (3.13) water y \ • / This leads to the overlap of both projections, and therefore in the image regions of enhanced intensities are seen as in figure 3.15 B. In effect then, an extended image of the phantom is seen in the former case and a compressed image is seen in the latter. The amount of this 105 A B Fig 3.14 2D images of a vial containing motor oil and water. The gradient is applied across the vial crosssect ion, and therefore, for both posit ive and negative values of the gradient the image appears the same as seen in (A) and (B) respectively. However, the image of one component is shifted with respect to the other due to the chemical shift differences between the two components. Experimental parameters : sweep width = ± 8064.51 Hz ; acqusition time = 15.87 ms ; block size = 512 ; gradient increment = 0.02 G/cm ; static gradient = 0.59 G/cm ; no of experiments = 128. 106 Fig 3.15 2D images of the same phantom as in figure 3.14, except that the gradient is applied along the long axis of the v ia l . The images for both posit ive and negative gradients are not equivalent as seen in (A) and (B) respectively. Regions of low (in A ) and enhanced (in B) intensities are seen depending on the orientation of the phantom (see text for details). 107 Fig 3.16 Schematic diagram showing the effect of gradient on the orientation of the phantom which contains equal amounts of water and o i l ; (A) and (B) represent gradient G y acting along +y and - y directions respectively. The length of the phantom is 2y and the chemical shift difference between the two resonances is 6. Note the change in the projection spectra that wi l l lead eventually to extended and compressed images respectively. 108 Fig 3.17 2D images of a phantom which consists of a beaker of oil placed at the center of a petri dish containing doped water. The composite images of this system are shown in (A) and (B). The image in (B), however, has been obtained with reduced gradients (see text for details). Experimental parameters \ (A) sweep width = ± 11111.1 Hz ; acquisition time = 11.52 ms ; static gradient = 0.95 G/cm. (B) sweep width = ± 3267.97 Hz ; acquisition time = 39.17 ms ; static gradient = 0.26 G/cm. overlap/displacement in either case is equal to 268.8 Hz as before. It is important to note that these distortions become much more prominent when higher magnetic f ields are used and hence very large gradients are required to minimize these distort ions. The application of increasingly stronger gradients, apart from being technically di f f icult , wi l l necessitate a proportional increase in the detection frequency bandwidth. Furthermore, the S/N ratio deteriorates rapidly, since more noise is introduced without concomitant increase in the signal strength. At constant f ields however, application of moderately higher gradients does give better images as illustrated in figures 3.17 A and B with data from a phantom which consisted of a beaker (2.5 cm id) containing oil placed in petri dish (4.8 cm id) containing water. This is to be expected, since the increased gradient strength causes the physical boundary to be represented by an increased frequency spread, while the chemical shift is not affected. Thus, the chemical shift displacement becomes negligible. Note further, that the image in figure 3.17 B is affected by static f ie ld homogeneities as we l l . As seen in the images in figure 3.17, the dark regions are always detectable, but the bright regions are not always seen. This is due to the fact that there exists a finite wall thickness between the two systems and in order to observe the enhanced region, the chemical shift displacement must be greater than the frequency resolution of the wall thickness. This enhanced region is just visible in figure 3.17 B, which was obtained with a decreased gradient magnitude. Attempts have been made in this section to give some understanding of the origin of the "chemical shift artifact". It is important that it be not mistakenly identified as something spatially real, specially in cl inical studies. The extent of this distortion in two-d imensional FT imaging depends on the gradient magnitude, the orientation of the system with respect to the read gradient, the interfacial thickness and, of course, on the chemical shift differences between the species. CHAPTER 4 NON-MEDICAL APPLICATIONS OF NMR IMAGING 4.1 BACKGROUND NMR Spectroscopy has been used extensively for over thirty years in such diverse f ields as organic, inorganic, physical and analytical chemistry (6,8), sol id state physics (9), food chemistry (138), geological studies (139), biological studies and biochemistry (140,141). In contrast, although early developmental studies on NMR imaging often involved fruits and plants (87,88,134,142-144), its continued success in cl inical studies has prompted very few workers to apply the technology to study non-medical systems such as those constantly profited from NMR spectroscopy. The present study is intended to explore non-medica l applications of NMR imaging. The first such application was reported in the literature in 1979 by Gummerson et al . (23), where they observed unsaturated f low of water within porous inorganic materials which have been saturated with water by capil larity. The sensitive point imaging technique (21) was used to monitor the dynamics of the internal water content distribution during capillary f low. From that study, it was clear that monitoring these permeation processes by NMR imaging would be of special interest to soil scientists and chemical engineers. Subsequently, Rothwell et al . (24) described the absorption of water by g lass - re inforced epoxy resin composi tes . The short and long term processes of water exposure are known to give reversible and irreversible effects which are important for the design of composite structures having enhanced physical properties and moisture resistance. It is clear that NMR imaging offers excellent possibi l i t ies of studying absorption and di f fusion processes. Basically, any non-metallic sample which contains a liquid-like component, and for which it is necessary or desirable to obtain qualitative or quantitative information non-destructively as a function of spatial location, is a candidate for analysis by NMR imaging. The novel applications pertaining to chemistry, material sciences and forest products pursued in the course of this study, are described in this chapter. The imaging techniques used are those developed during the studies which have been discussed in the earlier Chapters of this thesis. In some cases, these techniques have been further modified to enhance the detection of certain features of the system under study, and those modifications are discussed in relation to that particular application. 4.2 APPLICATIONS TO CHEMISTRY This section deals with NMR imaging studies of a Chromatography column which consists of three chemically different regions. 4.2.1. INTRODUCTION Since as early as 1951, NMR has been used in various studies of flowing fluids. For example, flow rates, dynamics of flow and spin relaxation times of flowing fluids have been studied quite extensively (145-150). With the advent of Fourier transform techniques, NMR has also been used as an on-l ine detector for high performance liquid chromatography (HPLC). Watanabe et al. (151) using a stopped-f low technique, directly coupled MH-NMR to LC. Many workers then exploited this idea and have published numerous papers on LC-NMR (152). Thus, a new analytical tool for rapid structure elucidation of components present in complex mixtures had been developed. The main interest is the analysis of petroleum and synthetic fuels (153-155). A recent review gives a good discussion on the technical considerations, advantages, and disadvantages of the L C - N M R technique (152). Given that NMR imaging yields spatial information of the system under consideration, instead of the output of the column being analyzed as in L C - N M R , it is conceivable that the column itself could be observed directly. Thus, there exists the possibi l i ty of mapping the. f low of fluids as a function of operating conditions in separation columns, reactor beds etc. A simple illustrative application is described in this thesis, which demonstrates the feasibi l i ty of such studies (156). The methods used are those developed earlier in this thesis with slight modif icat ions. 4.2.2. METAL CHELATE AFFINITY CHROMATOGRAPHY Metal chelate aff inity chromatography (MCAC), al lows selective extraction of materials on the basis of their aff init ies for chelated metal ions attached to insoluble matrices. The variability in extent of this interaction determines the degree of separation. Porath et a l . (157) described the first such application, where metal chelate linked to Sepharose was used to separate proteins. Since then, applications of M C A C to separate and purify proteins and nucleotides have grown tremendously (158). Sepharose is a trade name given to agarose gels. Agarose is a polymer of D-galactose and 3 ,6 -anhydro -L -ga lactose . Depending on the percentage of agarose used to prepare the gel , different pore sizes can be achieved. For example, the Sepharose 6B gel contains 6% agarose and has a pore size of 15 nm (159). Chelating Sepharose 6B consists of a chelating agent, iminoacetic acid, coupled to Sepharose 6B after epoxy activation and the structure is represented as, (Sepharose 6B) - O - C H J - C H ^ H ^ C H J - O - ^ H J V O - C H J - C H ^ H J - C H J - N ^ H J C O O H ^ The biscarboxymethylamino moieties provide the loci for metal -binding and are coupled to Sepharose 6B via a 12-atom long hydrophilic spacer arm. 2 + In this study, gel -bound chelate of Cu ions was used and a crosssection of the loaded column is illustrated in figure 4.2 A . The column consists of three separate regions; f ree-water , gel-copper/water and gel/water, and the discrimination between them depends on the changes in Tj relaxation rate of the water induced by the copper ions. 4.2.3. JJHE IMAGING METHOD The method used is an extension of the 2D spin warp imaging technique described in section 2.3. In addition to the inclusion of the third spatial dimension, the scheme was further modif ied to incorporate the inversion-recovery sequence; the complete pulse sequence is illustrated in figure 4.1. The preparation period is used to establish T1 contrast. Since only one mobile chemical species is present (ie; water), the raw data, S(G ,G ,G ), wi l l correspond to the three spatial dimensions x, y and z, x y z respectively. Note that the sl ice selection is achieved digitally along the z -d i rect ion by means of the refocussing gradient, G Preparation 180° 90° Image-development Pi Data-acquisition Delay-interval Gradient X Gradient Y Gradient Z Fig 4.1 The pulse sequence used for three-dimensional imaging together with the inversion-recovery preparation pulse. The preparation period, P, is varied for different experiments. The chromatography column was mounted vertically along the y - a x i s , inside the horizontal bore magnet with the magnetic f ie ld along the z -ax i s . The first FT of the raw data yields a series of "s l ices" along the z -d i rec t ion , ie: S (G x ,G y , z ) . Double FT of an appropriately chosen s l ice, wi l l then yield a xy-planar image, showing the projection of the distribution of spins in that plane along the column axis. A total of 256 experiments were done corresponding to 16 increments of G x and G y and the final image matrix size is 64*64 after data interpolation by ze ro - f i l l ing during the second and third FT. 4.2.4. RESULTS AND DISCUSSION The set of images shown in figure 4.2 represent single, Tj sensitive sl ice images (ca. 0.13 mm thickness) obtained for different values of the variable delay period, (Pj), between the initial 180° and 90° rf pulses. It is wel l known that paramagnetic metal ions change spin relaxation rates by providing an alternate, faster mechanism for proton relaxation. In clinical NMR imaging, the compounds which affect the relaxation parameters so as to enhance the contrast in the final image are termed as "contrast agents". It should be recalled that the image contrast is the perceived signal intensity differences between the different regions in the system under study (section 2.3.5). In this particular sys tem, the copper ions bound to the Sepharose 6B gel relax the water protons in that region, thereby providing contrast. Hence, the T, sensitive images wil l reflect the spatial distribution of the C u ^ + b o u n d region as seen in figure 4.2 C and D which were obtained by setting P, to equal the nulling times of the C u ^ + b o u n d region and the remaining regions respectively. The images in figure 4.2 A (a) (b) (c) (d ) x Fig 4.2 Single slice images of the chromatography column mounted vertically inside the wide bore system, (a) The image was obtained using sufficiently long values for P, (3 s), that all the protons contribute equally to the intensity of the final image, (b) The image shown has P, = 10 s, so that the magnetization of water in contact with the C u ^ + i o n s was its null. (c) Used P, = 3 0 0 ms to null all but the water in contact with the C u ^ + i o n s . (d) Shows an image with P, = 2 s, indicating partially relaxed water protons. A total of 256 experiments corresponding to 16 increments of each gradient G x and G y were performed. Experimental parameters ; sweep width = ± 2 5 0 0 Hz ; acquisition time = 51.2 ms ; block size = 5 1 2 ; gradient increments = 0 .015 G/cm ; TABLE 4.1 Relaxation Rates of Water in Different Regions of the Column Medium Relaxation rates1 (s ) Free water Water/Chelating Sepharose 6B Water/Chelating Sepharose 6B/Cu 1. Spin-lattice relaxation times were obtained using the null point method. 5*10 6 * 1 0 - 1 16 and D represent a completely relaxed image which is equivalent to a normal image, and a partially relaxed image respectively. Besides locating the spatial distribution of the species it is also possible to characterize some of their properties. This is illustrated by individual measurements of the spin-lattice relaxtion rates of the water in the different regions in the column. These data, listed in Table 4.1 were obtained using the null point method. It is seen that the copper ions cause nearly a 30- fo ld increase in the relaxation rate. Microanalysis for copper present in the gel gave a value of 12.6 ^mols and this agrees reasonably well with that estimated from references (157) and (160). In effect, 2 + approximately 10 Aimols of Cu ions produce contrast indirectly by inducing a 3 0 - f o l d enhancement in relaxation of the surrounding water protons. This brings in a concept of "molecular -ampl i f icat ion" in which the information of interest is encoded in the responses of a probe species present in high concentration (ie; water in this case), whose properties are directly influenced by interaction with target molecules which are themselves present in low abundance (ie; copper ions). In other words, even though direct detection of 10 M I T I O I S of Cu^ + ions by NMR is completely impossible, indirect detection is feasible by means of an imaging measurement, as illustrated here. The application described in the present study was the first performed on the wide bore system. Although not quantitative, it clearly proves the feasibi l i ty of such studies. The main l imitation in this work was the performance of the receiver coi ls which had very long 90° and 180° pulse lengths (ca. 880 M S for a 180° pulse). This makes selective nulling of different components with smaller Tj differences dif f icult . Since then the receiver coil designs have been modif ied especially to observe f lowing fluids by many workers, both in cl inical NMR imaging (161) and in L C - N M R (152). Hence a challenging array of new opportunities exist, particularly to study many chemical phenomena which occur in, or on columns. 4.3 APPLICATIONS TO MATERIAL SCIENCE Spatial mapping of fluid distributions within porous rock samples is described in this sect ion. 4.3.1 INTRODUCTION A s mentioned previously, the sensitive point NMR imaging technique has been used to monitor the dynamics of the internal water content distribution in porous inorganic materials (23). Sedimentary rocks, especial ly those encountered in o i l - f i e l d exploration and recovery processes, are a scient i f ical ly and technically important group of materials. It is conceivable that visualization of f luid phases within porous opaque rocks might lead to better understanding of the processes occurring within rocks. Clearly NMR imaging has the potential to provide means of mapping those processes. Furthermore, it may be also possible to study the porosity distribution, l iquid-phase d i f fus ion, interfacial processes and porous media fluid dynamics. Interest in the present work is focussed upon applying NMR imaging techniques to map either a single fluid phase or, simultaneously two fluids of different chemical composit ion. It is appropriate at this juncture to give a brief insight into o i l - recovery methods, so as to highlight the importance of characterizing f luids, particularly oil and water in porous rocks. 4.3.2 OJL RECOVERY METHODS Ore reservoirs are usually comprised of permeable rocks, such as, sandstones, l imestones and dolomites. The production of oi l and gas using the natural reservoir energy constitutes the "primary" o i l - recovery process. Generally, more than 70% of the original oil content remains underground fo l lowing this initial recovery process (162). Hence, there exists a need for further recovery. The "enhanced" recovery processes are those which apply external or artif icial forces to dislodge the oil from the reservoir rock pores and move it to the production wel ls . Such recovery methods have attracted widespread attention and numerous different strategies have been proposed and implemented (163). The most commonly employed recovery process is water - f lood ing (164), where treated water is injected into the o i l -bear ing part of the reservoir to force the oil to f low toward the production wel ls . In spite of increased oil production, the eff ic iency of water - f lood ing is hindered by two major l imitations. Due to the immiscibi l i ty of water and o i l , 2 5 - 5 0 % of the oil is left behind in the form of small droplets as the water moves through the reservoir rock. Surfactant additives are used to minimize or eliminate this problem and thereby increase oil production (165,166). The second limitation is that, due to diff icult wel l placements or unexpected geological configurations, the advancing waterfront generally bypasses significant portions of the reservoir space. Information regarding the success rate for this type of oil stimulation is scarce. A l l that is known is that sedimentary rocks bearing oil are inhomogeneous in nature and their configurations are not necessarily conducive to uniform fluid f low . Hence, the predictions of fluid f low patterns between injection and production wel ls cannot be made with reasonable accuracy. A further problem is that detailed understanding of fluid f low through porous media is stil l scarce. 123 However, efforts have been made by many workers in this f ield to try to understand these processes in model systems (167-169). A s the reader might have noticed, NMR imaging offers excellent capabilities of characterizing the physical distribution of fluid residues in o i l -bear ing strata at each stage of recovery. Furthermore, Chemical Shift imaging can be used advantageously to distinguish several fluid phases which are of different chemical composi t ion, present within the same pore space (170). At the time the work described in this section began, there were no reports related to mapping fluids within porous rocks. Very recently, however, interest has been seen in petrophysical applications and a number of workers have reported NMR images of oil and water .distributions within porous samples; the work of Baldwin et al . (171) demonstrated the capability of distinguishing oil from water within rocks by paramagnetic doping of water with manganous ions. Shortening of transverse relaxation times (T2) of water in the paramagnetic environment eliminates the water signal completely thereby enabling only the oil resonance to be observed. Blackband et al . (172) published chemical shift resolved images on a test sample using the steady state free precession technique coupled with back-project ion image-reconstruction algorithms. Maerefat et a l . (173) reported the use of two-d imensional FT imaging technique coupled with inversion-recovery sequence to select ively null one component while imaging the other. A few other workers have also reported NMR images of a single species (eg: oil) contained in these porous rock samples (174-176). Most work has been done on whole body NMR scanners at very low magnetic f ields (— 0.5 T) with selective excitation techniques employed to localize smaller sample regions and standard 2D FT imaging used for spatial discrimination. The present work in this thesis demonstrates, in addition to being able to visualize fluids within porous rocks, the feasibi l i ty of simultaneous mapping of oil and water with the aid of Chemical Shift imaging methodology discussed in chapter 3. Both 2D projection images across the samples and thin sl ice high resolution 3D images wi l l be discussed in the fo l lowing sect ions. An important requisite for this technique to be effect ive is that the fluid bearing rock samples exhibit reasonable linewidths. It is wel l known that paramagnetic impurities broaden NMR linewidths because of short T 2 relaxation, and samples containing such impurities render the imaging technique less attractive. Initially, therefore, a survey of different samples was done in order to select those with optimum relaxation properties. Procedures for saturating the rock samples are decribed in Chapter 6 (section 6.3). 4.3.3 T_HE IMAGING METHODS The sequences used to image porous samples are the 2D, 3D and 4D FT methods described in sections 2.3, 3.2.3 and 3.3 respectively ; the latter two encode chemical shift information. Tj sensit ive images were obtained by incorporating the inversion-recovery sequence, as described in section 4.2.3 except that 180° rf refocussing pulses were used instead of gradient inversion. Speci f ic features of this modif ied pulse sequence wi l l be highlighted in the appropriate sections. 4.3.4 RESULTS AND DISCUSSION 4.3.4.1 Water Relaxation Several sedimentary rocks and a small number of sintered glass compacts impregnated with water were analyzed to determine the relaxation characteristics. Table 4.2 summarizes the results obtained here. Compared with pure water, water in heterogeneous mineral systems is known to show a somewhat more, rapid T1 relaxation and a strongly enhanced T 2 relaxation. This general trend is seen in the data presented in Table 4.2. The sedimentary rock samples exhibit larger variations in their linewidths than in their T1 values; however, the two sandstone samples have similar relaxation properties. On the other hand, the sintered glass compacts show broader linewidths and the Tj values decrease as the particle size decreases. This agrees well with the observations of Glassel (177), that the larger- the particle size, the larger the corresponding Tj relaxation t imes. Although many studies related to NMR relaxation in heterogeneous systems have been made (177-189), there exists no unique interpretation of all experimental results. It is generally assumed that in the "two phases in rapid exchange" model (186), water molecules exchange rapidly between a s lowly relaxing environment (free water phase) and a rapidly relaxing environment (bound water phase) and a single weighted relaxation rate is observed. Furthermore, cross- re laxat ion between water protons and magnetic nuclei or paramagnetic sites located at the wall is also known to contribute to the shortening of Tj (184,187,188). However, contribution to the enhanced T 2 relaxation may arise from dif fusion in local f ield gradients TABLE 4.2 126 Proton Relaxation Times and Linewidths of Water in Porous Rocks and Sintered Glass Compacts Spin-Lattice Relaxation1 Linewidth2 Time (ms) (Hz) LIMESTONE Lepine/Lavoux Lepine/Lavoux Courteraie/Savonnieres Hauteroche SANDSTONE Berea (2 mD permeability) Berea (20 mD) Berea (50 mD) SINTERED GLASS Sample 1; 177-210 nm original particle size, 0.33 volume fraction porosity Sample 2: 44-53 Mm original paricle size, 0.38 volume fraction porosity Sample 3: 88-105 Mm original particle size, 0.13 volume fraction porosity 395±11 230±7 240±12 370±10 335±4 280±4 1260161 415150 745+35 122.1 234.4 317.4 390.6 151.4 1445.3 166.0 483.4 512.7 673.8 1. Spin-lattice relaxation times (at 80 MHz) were calculated using the three parameter fit routine to analyze the inversion-recovery data. 2. Linewidths were calculated as described in section 6.3 (15%). caused by magnetic susceptibil ity differences between the matrix and the pore f lu id . It is also possible for effects to arise from dissolved paramagnetic impurities derived from soluble mineral constituents but that was excluded in this study. Based on these f indings, sandstone samples and limestone (lepine) samples were chosen for imaging studies as they exhibited moderate linewidths and Tj values. However, most of the work to be described in the fo l lowing sections concerns sandstone samples, because of their small sample size, which is convenient to analyze in the wide bore sys tem. 4.3.4.2 Imaging of Water 2D FT imaging technique was used to illustrate the distribution of water within porous rock samples. The images shown in figure 4.3 A and correspond to cylindrical samples of sandstone (2.5 cm diameter, 2.5 cm height) and limestone (3.5 cm diameter, 7.7 cm height) containing 5.6% and 11.8% water by weight, respectively. These images represent projection views across the diameter of these samples and the water distribution is encoded on an 8 - leve l color scale with white representing regions of high intensity. It can be seen that the images are almost similar , except that the limestone sample exhibits a continuous distribution of water, probably due to the presence of higher water content. In addition, this non-uni formity may also arise due to porosity differences between the samples. Dif fusion measurements could give a better understanding of the porosity. However, such measurements were not attempted in this work. Another feature present in figure 4.3 B is the image intensity variations 1 2 8 B 2.5 cm 2.5 cm SANDSTONE 7.7 cm 3.5 cm LIMESTONE Fig 4.3 Two-d imens ional images in (A) and (B) depict the distribution of water protons within a berea sandstone sample and a l imestone (lepine) sample which have been partially saturated with water, and contain 5.5% and 11.8% water by weight respectively. Both samples are cylindrical in shape and the images represent the projected intensity across the diameter of the samples. Experimental parameters : sweep width = ± 8064.51 and 10000 Hz ; acquisition time = 15.87 and 12.8 ms ; block size = 512 ; gradient increment = 0.024 and 0.013 G/cm ; static gradient = 0.38 G/cm ; no of experiments = 128. due to B1 inhomogeneity of the 7.5 cm diameter axial resonator probe over the larger limestone sample. In order to minimize the Bj f ield variations, smaller samples (eg: sandstone) were used in subsequent studies. A major drawback in these 2D images is that the image information has been obtained from the whole sample (ie: projection image) and thus, lacks detail . Although this serves as a means of mapping fluid distribution within the samples, it would be better if smaller sections of the samples could be viewed. One of the most common methods of achieving smaller sections is by employing slice selection routines (190,191). Such techniques usually consist of selective rf pulses of long duration applied in the presence of a magnetic f ield gradient, so as to select narrow regions of excitation along the applied f ield gradient direction (figure 4.4 A and B). The slice thickness is determined by the duration of the rf pulse and the magnitude of the applied gradient. Note that the duration of the rf pulse is inversely proportional to the frequency bandwidth of excitation. A s illustrated in figure 4.4 C and D, thin sl ices could be achieved by the application of very long rf pulses and/or very high gradient magnitudes respectively. The rf pulses cannot be infinitely long since there exist inherent l imitations determined by the spin relaxation times of the probe species. Moreover, higher strengths of gradients wi l l result in lower S/N. Hence, achieving thin sl ices are demanding and the main criterion that has to be considered is the observable S/N. A convenient method to achieve very thin sl ices is to observe the complete object by means of the 3D. imaging technique (figure 4.5); sl ice selection is achieved here, digitally along the observe dimension by data 130 Fig 4.4 (A) Selective excitation of a plane in the presence of a magnetic field gradient, produces a linear variation of the resonance frequency with position. (B) Rf frequency bandwidth (Aco) excites a slice of thickness AZ along the z-direction. (C) Effect of gradient strength on slice thickness : a higher strength gradient (M) causes the slice thickness to reduce to A Z M at constant rf frequency bandwidth excitation. (D) Effect of rf frequency bandwidth on slice thickness : a decrease in the bandwidth (Acdiy|) causes the slice thickness to reduce to A Z M a t constant gradient . strength. processing. Depending on the thickness desired, the strength of the gradient and the digitization along the observe dimension can be varied. The cost of observing very thin sl ices is the increased experimental time when 3D FT imaging is pursued. However, many sl ices can be obtained in just one experiment. Figure 4.6 shows single sl ice images of the same sandstone sample as before, but now obtained from the 3D FT technique. A total of 4096 experiments were done corresponding to 64 increments of gradients, G and Gy and the spin echo signal was acquired in the presence of the static gradient, G z (see figure 4.5). Sl ices were selected digitally along the z -d i rect ion , fo l lowing the first FT. The sl ice thickness is ca. 900 (zm and in this particular case images of f ive sl ices spaced apart by 2.8 mm are illustrated in figure 4.6 A - E . In effect a total of 27 s l ices spaced side by side can be obtained from this single imaging experiment. Clearly these images contain more information regarding the fluid distribution, pore space etc. For example, the water is predominantly distributed along the outer surfaces, as seen by the higher signal intensity. A l s o note that lower intensity at the center of the image, which reflects the dif f iculty of water penetration into the central portion of the sample, was not available previously in the 2D projection image (figure 4.3 A) . Figure 4.6 F shows an image obtained by summing the above f ive images, and it is clear that this particular image c losely approaches the 2D projection image. In fact this 2D projection image can be obtained by summing all 27 sl ice images. However, note that in this particular 3D sequence only 64 increments have been done compared to 128 in the 2D sequence, for the phase encoding 1 Delay interval Gradient X i 90 Image -0 development Data acquisition 80« Gradient Y Gradient Z Fig 4.5 Three-dimensional x, y, z imaging sequence with a 180' rf refocussing pulse to generate the spin echo signal. Digital slices can be , selected along the z-direction after the first FT. LO 132a Fig 4.6 Two -d imens iona l sl ice images (A) to (E), obtained from a three-dimensional data set of the same sample as that used in figure 4.3 A . The slice thickness is ca . 900 um and each slice is separated by 2.8 mm. 64 increments corresponding to each of the two gradients G x and Gy were performed. (F) shows the sum of images (A) - (E), and this image c losely approaches the 2D projection image seen in figure 4.3 A . Experimental parameters : sweep width = ± 10000 Hz ; acquisit ion time = 12.8 ms ; block size = 512 ; no of experiments = 4096 ; gradient increments = 0.023 G/cm ; static gradient = 0.2 G/cm. gradient. Although this approach provides a convenient means of mapping fluid distribution within opaque porous rocks, it is more challenging to map simultaneously two f luids, based on their chemical shift differences. This is the subject of discussion in the in the next section. 4.3.4.3 Imaging of Oil and Water The practicality of chemical shift imaging depends on being able to produce chemical shift separation of resonances from different fluid substances comparable to or greater than the effect ive linewidth of the individual resonances (chapter 3). In order to illustrate the feasibi l i ty of CSI to map oil and water, a composite sample was made up of two cylindrical specimens of berea sandstone, one impregnated with water (5.6% by weight) and the other with n-dodecane (4.3% by weight), and then placed coaxially end to end. This composite sample was then imaged using the 3D FT CSI sequence described in section 3.2.3. Figure 4.7 shows chemical shift resolved stacked plot images of water and n-dodecane absorbed separately by capillarity into two halves of the composite sample. A s seen in the normal high resolution spectrum of the sample, the proton resonances of water and n-dodecane are separated by 3.5 ppm. A pair of two-d imensional xy planar images were reconstructed, each corresponding to one of these two resonances; they show a clear separation of the two f luids. The presence of a small amount of water distributed uniformly in the top half of the composite sample is apparent; this is due to the inherent moisture present in the sandstone sample. This Fig 4.7 Two-dimensional xy planar chemical shift resolved images obtained from a 3D (x,y,{)data set of a composite berea sandstone sample. The images are shown as stacked plots and the normal high resolution NMR spectrum of the composite sample is shown in the inset. A total of 1024 data sets were acquired, corresponding to 32 increments of each gradient G x , Gy. Experimental parameters ; sweep width = + 5 0 0 Hz ; acquisition time = 25.6 ms ; block size = 5 1 2 ; gradient increments = 0 .022 G/cm ; final image display = 2 5 6 * 2 5 6 . demonstrates the ability to resolve two chemically distinct fluids coexisting as immiscible phases intimately mixed in the same pore space. In order to simulate a more natural case, one sandstone sample was saturated partially with an emulsion made from a mixture containing 1;1 standard motor oil (10W30) and H 2 0/D 2 0 mixture (1:1). In this particular example, the proton resonances of water and oi l are separated by 3.7 ppm; hence obtaining chemical shift resolved 2D projection images using the 3D (x,y,5) CSI sequence is easy. However, it becomes a challenge if thin s l ices are to be observed, since the presence of a static gradient during detection period (see figure 4.5) wi l l destroy the chemical shift dispersion. Recall that the sl ices are to be obtained digitally after the first FT. In order to achieve this objective, the system has to be modif ied so that only one component is present during detection. Fortunately, in this sample the oil and water resonances differ in their T t values; the spin - lat t ice relaxation times determined by the null point method are 115 ms and 1.15 s respectively. This clearly represents an opportunity to manipulate the chemical properties of the system advantageously. High resolution chemical shift resolved imaging was performed by incorporating the inversion-recovery sequence into the three-dimensional x,y,z imaging sequence as illustrated in figure 4.8. The initial preparation period is used selectively to null one component so that spatial encoding could be perfomed on the other. The final spin echo signal is detected in the presence of the static gradient G . Note that this enables the selection Gradient X Gradient Y Gradient Z 180 Preparation o 9C Image-0 development 180 Data acquisition o ...llllllllllllll.. Delay-interval *> Fig 4.8 A modified three-dimensional x, y, z imaging sequence with inversion-recovery preparation to selectively null the undesired chemical component. co 137a Fig 4.9 Two-dimensional slice images obtained from a T , sensitive 3D data set are shown in (A) - (E). The same sandstone sample has been saturated with a mixture of oil and H20/DjO and these sets of images are those of water while the oil resonance was passing through its null. The slice thickness and the slice separation are the same as that in figure 4.6. (F) is a sum of (A) - (E). Experimental parameters : same as that in figure 4.6. The null time = 80 ms. 138a Fig 4.10 Two-d imens ional sl ice images of the same sandstone sample obtained using the sequence given in figure 4.8. However, these represent the oi l present in the sample. The null time for water is 800 ms. (F) is a sum of (A) - (E). Experimental parameters : same as in figure 4.6. 139 of sl ices after first FT. Separate T t sensit ive images of oil and water were generated from two experiments; the oil image is obtained while the water is at its null and vice versa. Figures 4.9 and 4.10 represent sets of single sl ice images depicting the distributions of oil and water respectively within the sandstone sample. The sl ice thickness is ca. 900 Mm and these s l ices are separated by 2.8 m m . This method works well if the Tj differences between the two resonances are large. Unfortunately, if these differences are smal l , then selective nulling becomes dif f icult . In such instances 4D FT imaging is an alternative technique to differentiate these two resonances. Such was the situation in a limestone sample (3.7 cm diameter, 6 cm height) saturated with North Sea crude o i l . In this sample, the spin - lat t ice relaxation times of the two resonances, oil and water were ca. 289 ms and 354 ms respectively. 4D FT imaging was performed on this sample and the chemical shift resolved images are illustrated in figure 4.11, together with the normal high resolution spectrum. A total of 8192 experiments were performed corresponding to 32 increments of gradients, G x , G y and 8 increments for G 2 . These xy planar images have been obtained from a z - s l i c e of thickness ca. 0.74 mm. Obviously , these images are of low resolution and hence, many details are not seen. However, the water present in the sample is clearly differentiated. Quantitation of NMR image parameters, particularly p and T l f should yield information about fluid permeabil it ies, porosity and pore -s i ze distributions in these porous samples. Although NMR relaxation studies have been used to determine these matrix properties in sandstone samples based Fig 4.1 1 on statistical approaches (187-189), extension of such studies to perform imaging wi l l demand extensive sample analysis and computer data manipulations. In addition, measurements of di f fusion coeff ic ients (192) should yield information regarding porosity distributions. Such studies, though beyond the scope of this thesis, wi l l help to understand fluid processes within porous media and eventually lead to clearer insights towards primary and supplementary oil recovery processes. 4.4 A P P L I C A T I O N S T O F O R E S T P R O D U C T S Several different features pertinent to the forest industry are discussed in this section. 4.4.1 INTRODUCTION One of the most important natural materials used by man is wood . Its origin as a product of metabolism of the living tree makes it an inherently variable substance; therefore, wood characteristics vary both between and within species. The chemical composit ion of wood consists of a complex admixture of cellulose (40-50%), hemicellulose (20-35%), lignin (15-35%) and extractives (2-30%) (193-195). Most information regarding the chemical structure of wood has been obtained from studies of its components isolated destructively, rather than from intact wood itself. NMR has played a significant role in the study of wood -wate r systems as early as in 1962 (196). Since then many workers have used NMR to determine the moisture content and to study the interaction of cel lulosic materials with water (197-204). Transverse relaxation rates have been measured as a function of moisture content (205), and recently di f fusion of water in wood has also been studied (206). While such studies have been carried out to try and understand the distribution and behavior of water in wood , sol id state Carbon-13 NMR has been pursued recently in an attempt to analyze the chemical components present in wood and wood products (207-209). NMR imaging studies of wood , which wi l l be discussed in this sect ion, extend the scope of NMR further. The importance of wood as a massive source of renewable biomass, added to the increased interest in improving the eff ic iency of its use as a valuable industrial and construction material , has heightened recognition of the need for new insights concerning its internal structural features. NMR imaging promises to be a useful tool to study these features. It is appropriate now to discuss briefly some characteristics of w o o d ; its growth, wood -water relationship, and defects'. 4.4.2 WOOD GROWTH A straightforward approach by Gray and Parham (210) describes clearly the physiological processes responsible for tree growth and wood development. Figure 4.12 depicts a crosssect ion of an oak tree together with the various tissue regions present in the sample. The cells responsible for wood growth are in the cambium which is actually a single cell layer around the tree. The living cel ls (phloem), responsible for the movement of food and hormones from the crown of the tree (leaves, needles, buds) down to the tree trunk, branches and roots constitute the inner bark. The outer bark is composed of dead cells that have been pushed outward by 4.12 The different regions in the crosssection of an oak tree. [From reference (210)] the expanding tree trunk. The light colored zone next to the cambium is the part of the xylem called sapwood and this region contains many l iving cel ls . The non- l i v ing cells are responsible for moving sap (water and minerals) upwards from the root system to the crown. The very center of the tree is referred to as the pith, and the dark colored xylem region between the pith and the sapwood is called the heartwood. The heartwood region is formed by gradual conversion of sapwood, and it contains all physiological ly dead cel ls . This region provides the support and strength to the tree. In most tree species, there exist two distinct annual growth increments; a light colored portion (lower specif ic gravity) formed in the early part of the growing season (earlywood), and a darker colored portion (higher specif ic gravity) formed later in the season (latewood). The earlywood cells are th in -wal led and have large cell cavities as opposed to the latewood cel ls . The extent of these variations determines the demarcation of the growth rings. 4.4.3 WATER JN WOOD Water is a fundamental constituent of the living tree. The wood being a hygroscopic material has aff inity for water in both liquid and vapor fo rm. The temperature, humidity of the surrounding atmosphere and the moisture content of the wood determine, whether the wood absorbs or loses water. Generally, the most recently formed sapwood transports water in w o o d . The moisture content (M.C), is usually expressed by the fo l lowing relationship, M.C. = [Weight of water in wood/Oven dry weight of wood] * 100 % (4.1) The standard method for determining the amount of water present in wood is to dry it in an oven to constant weight at 105°C. The weight of water present is therefore the difference in weights before and after drying. For example, consider a wood sample which weighs 100 g and if after drying the weight is 40 g, then using equation 4.1 the moisture content of the wood sample is 150%. Percentages like this are very common in the forest industry since the reference, which is the oven dry weight is always very much smaller than the weight of the moist sample. The wood -water system is an extremely complex one. The water is known to be associated dynamically within w o o d , and is therefore distributed nonuniformly (211). It may exist as bound water in the cell wal ls or as free water in the cell wall cavit ies. The understanding of the moisture behavior and its distribution in wood are important in predicting and controll ing wood properties. 4.4.4 DEFECTS ]N WOOD Defects in wood are c lassi f ied mainly into two groups; growth-related defects, such as c ross -gra in , knots and pitch; and defects caused by wood -des t roy ing , wood-s ta in ing fungi and deterioration caused by insects. In addition to these fo l lowing fe l l ing, there are also defects caused by seasoning, such as warping, honeycombing etc. and by machine burns (212). A major problem in sawing logs into lumber has always been that often little is known about the size and relative posit ion of the defects which determine the quality of wood produced. With a v iew of increasing the yield of lumber, a number of defect detection methods (213-215) for scanning logs prior to sawing are being explored. An X - r a y Computerized Tomography (CT) scan of a log provides a detailed crosssectional view of what is inside (216-218); unfortunately this technique is not capable of differentiating defects. This is because the technique is based on the actual density of the wood , and the X - r a y s are sensitive to the presence of moisture (particularly the variation in moisture content) which is known to cause errors in the identification of knots (219). In order to identify and distinguish knots properly interactive computer programs have to be used; these have been developed on the basis of general shape and occurrence of the knots and decay present in the sample. This undoubtedly can sti l l lead to errors in identif ication. Furthermore, the use of X - r a y s requires safety measures. Thus there st i l l exists a need for an accurate evaluation of lumber quality. In the present work, the potential of applying NMR imaging techniques to detect defects, as wel l as to study the distribution of water in wood is explored. Based on the properties of water (proton densit ies, relaxation characteristics), we have shown that several internal structural features such as growth rings, .knots and decayed regions are detectable (220,221). Furthermore, drying of wood can be progressively fo l lowed as the total moisture content changes with drying t imes (222). Kiln dried wood unfortunately cannot be observed directly. However, if the dried wood is saturated with water, the characteristic features can again be. visualized. This also offers the possibi l i ty of studying impregnation of wood by aqueous solutions of preservatives and f lame-retardent chemicals. Finally, imaging of some specialty products f rom wood is discussed. Saturation of the dry wood samples with water was usually done either under pressure (ca. 15 mm Hg) or by soaking for a long period of t ime. The procedures for these are described in Chapter 6 (section 6.3). 4.4.5 JJHE IMAGING METHODS The two-d imensional imaging sequence described in section 2.3 was used in most of the studies described. In some cases, the sequence was modif ied with the inversion-recovery preparation to yield Tj contrasted images. Small samples of wood were studied on the 80 MHz imaging system in the Chemistry department, while large samples of freshly cut wood were imaged on the 6 MHz who le -body NMR scanner at the UBC hospital using both the head coil and the body co i l . Sl ice selection was used on all samples studied on the latter with slice thickness of 1 cm. The same basic technique was used in both instruments except in the case of one particular sample (figure 4.16) where multiple sl ice images were obtained on the hospital scanner. 4.4.6 RESULTS AND DISCUSSION 4.4.6.1 Internal Structural Features The proton image measured at 80 MHz shown in figure 4.13 B, is that of a thin crosssectional piece of douglas - f i r {Pseudotsuga menziesii) wood saturated partially with water by plain soaking to a moisture content ca. 172% (figure 4.13 A). The dimensions of the sample is ca. 3.5x3.5x1 c m 3 . The growth rings are clearly v is ib le , probably due to the higher and lower moisture content of the earlywood and latewood respect ively ; the earlywood is known to contain larger cell cavities which can encapsulate more water. This rationale is further justif ied by the fact that, the T t and T 2 sensitive images did not show any marked differences compared to the normal 2D image, implying that the image intensities in figure 4.13 B are governed to a greater extent by water proton densit ies. A large rectangular sample of white spruce (Picea glauca) wood (dimensions 4.5x4.5x10 cm 3 ) impregnated with water under pressure and imaged in the whole -body NMR scanner reveals, in addition to the growth rings, a clear separation of the heartwood and sapwood regions as seen in figure 4.14 A . Figure 4.14 B shows images obtained at 6 MHz, of two samples, aspen (Populus tremuloides) and white spruce, each of dimensions 9x4x30 c m 3 . A buried knot is clearly seen in the spruce sample. These samples also have been impregnated with water under pressure and the inherent differences in the moisture content between different regions A B 150 Fig 4.13 Two-d imens ional image of a douglas- f i r sample saturated partially with water (A) is illustrated in (B). The thickness of the sample is 1 cm and the separation between the growth rings is ca. 1 mm. Experimental parameters \ sweep width = ± 5000 Hz ; acquisition time = 25.6 ms ; block size = 512 ; gradient increment = 0.016 G/cm ; static gradient = 0.44 G/cm ; no of experiments = 128. Fig 4.14 Two-dimensional slice images (slice thickness = 1 cm) obtained from the hospital NMR scanner, depicting crosssectional and transverse views of wood samples saturated with water. (A) shows clear separation of heartwood/sapwood regions in a douglas-fir sample, while, (B) shows a buried knot in the spruce sample. The head coil was used to acquire these images. Fig 4 .15 Two-dimensional slice image (slice thickness = 1cm) in (B), obtained from the hospital NMR scanner using the head coi l , depicts the crosssectional view of a freshly cut aspen sample. (A) is a photograph of the same sample. The growth rings are clearly visible in the image. tn TABLE 4.3 Tj Values of the Different Regions of a Fresh Sample of A s p e n 1 Measured at 6 MHz on the Who le -body Scanner Regions TV (ms) Heartwood 188 - 230 Sapwood 343 - 395 Outer Bark 202 - 206 Inner Bark 244 1. The same sample as in figure 4.15 B. 2. Tj values were obtained using the standard PICKER software. technique, the knots are easi ly identifiable in a direct manner without any need for computational manipulations. A more useful test of the method would be with a freshly cut sample; this could be assumed to preserve its natural characterist ics, provided it is stored properly. Sample disks (ca. diameter 22 c m , thickness 11 cm) cut from freshly fel led aspen logs from Grand Prairie, Alberta were stored in sealed plastic bags and kept cold below 2°C prior to examination. The NMR image obtained on the who le -body scanner 4 days after cutting is seen in figure 4.15, together with a photograph of the sample. As it can be noted, all different regions, in addition to the growth rings, can be seen very clearly , including the bark, heartwood and sapwood Fig 4.16 Two-dimensional multiple slice (1 cm thick) image set (A-D) of an aspen sample (ca. 25 cm diameter) Each slice is separated by 1 cm. A hidden knot and a r e g i o n of decay (rot) are highlighted. The head coil of the whole-body NMR scanner was used to acquire this image set sample. A s it can be noted, all different regions, in addition to the growth rings, can be seen very clearly, including the bark, heartwood and sapwood regions. The spin- latt ice relaxation times of these different regions were computed on the hospital scanner and are listed in Table 4.3. It can be seen that most regions exhibit similar Tj values except the sapwood region which shows a slightly enhanced T l t Although T2 measurements were attempted, it was not possible to get precise values for these regions. However, a qualitative estimate was made for the heartwood and sapwood regions and the values varied from 85-108 ms. This implies that Tj and T 2 contributions to the image intensity in figure 4.15 are small for these two regions. Hence, the brightness differences reflect more of the proton density differences between the two regions, ie: moisture content di f ferences. This agrees wel l with the observations in aspen w o o d , in that the sapwood is known to contain more moisture than the heartwood (223). Another fresh aspen sample studied next on this scanner exhibits interesting internal architecture as seen in the images in figure 4.16 A - D , which represent four alternating parallel s l ice images obtained as a part of a contiguous mul t i - s l i ce data set. Although this sample did not show any visible signs of decay to the naked eye, the presence of the dark and bright zones in these images indicate decay and this leads to the fo l lowing section which discusses decay in wood . Before concluding this section on internal structural features it has to be mentioned that, very recently, Wang et a l . (224) have reported some similar work in which they have used NMR imaging to highlight annual growth rings, worm holes and an embedded knot in a large chunk of a cherry f i rewood sample that had been soaked in water for four weeks. 4.4.6.2 Decay in Wood The natural durability of wood is associated with i t s . ability to resist the attacks of foreign organisms like fungi, insects and marine borers. The reason why trees have a spectacular survival record is , because they have evolved into highly compartmented organisms which enable them to wall off injured and infected wood (225). Decay is the final stage of a long and complicated process that is initiated by wounds, and involves the interactions of microorganisms among themselves, and with the tree. It was intended therefore to explore the possibi l i ty of using NMR imaging to detect decay. The single slice image (1 cm sl ice thickness) obtained at 6 MHz, shown in figure 4.17 is that of a freshly fel led aspen sample (ca. 25 cm diameter, 12 cm thickness) with a high degree of decay. The sapwood appears as a bright region and the decayed heartwood region appears dark. Interestingly, a very bright region is also seen in between. Aspen is known to have very low resistance towards decay (226). Furthermore, microorganisms have been isolated and reported to be the cause of decay in aspen (227). These were confirmed for the above sample by mycological studies undertaken in the Microbiology laboratory in the Department of Chemistry. These data, together with moisture content distribution and the NMR parameters (at 80 MHz) are summarized in Table 4.4 for different regions for this particular sample. The microbiological tests were done solely for the purpose of confirming the existence of Fig 4.17 Two-dimensional slice image (slice thickness = 1cm) in (B), obtained from the hospital scanner, depicts a crosssectional view of a freshly cut aspen sample vyjth high degree of decay. (A) is a photograph of the same sample. The body coil was used to acquire this image. cn T A B L E 4.4 NMR and Microbiological Properties of Different Regions in the Aspen Sample 1 158 Region Microorganism 2 M.C J (%) T , 4 (ms) Linewidth5 (Hz) 1 Basidiomycete 226.2 572120 156.3 2 w 202.4 504123 218.8 3 n 123.7 451123 34.4 4 Trichoderma 177.8 517126 51.6 5 n 251.8 460120 31.3 6 n 147.6 387121 50.0 7 ti 255.7 773122 84.4 8 n 253.0 528114 42.2 9 Bacterial Growth 228.9 529124 96.9 10 275.0 512120 28.1 1. The same sample as in figure 4.17. 2. Mycological tests done by Dr. P A . Salisbury, Department of Chemistry, UBC. 3. Moisture content values determined as described in section 6.3 (11%). 4. Spin-lattice relaxation times (at 80 MHz) were calculated using the three parameter fit routine to analyze the inversion-recovery data. 5. Linewidths were calculated as described in section 6.3 (+5%). (contd) 159 microorganisms. In doing this it was also possible to identify certain classes of fungi like basidiomycetes and ascomycetes (imperfect form Trichoderma). Circular disk samples (3.5 cm diameter, 1 cm thickness) cut from this large sample were then imaged on the 80 MHz wide bore sys tem, in an attempt to further analyze the decayed regions. Two illustrative sets of results wi l l be discussed pertaining to regions 3 and 9. Interesting results are seen in figure 4.18 A , which is an image of the sample (figure 4.18 B) corresponding to region 9. There are three zones of different image intensity; sapwood, an intermediate zone and heartwood. Recall , that in figure 4.17 (A), a similar situation existed. The differences in Fig 4.18 Two-d imensional image obtained on the 80 MHz wide bore system of a circular disk section (3.5 cm diameter, 1 cm thickness) of the sample used in figure 4.17. The section corresponds to region 9 in Table 4.4. (B) is a photograph of the same section. Experimental parameters \ same as in figure 4.13. TABLE 4.5 NMR and M.C Variations of Region 9 in Table 4.4 Zone M.C 1 (%) T , J (ms) Linewidth 5 (Hz) (i) 175.8 350±40 273.4 (ii) 203.6 512±31 46.8 (iii) 228.9 529124 96.9 1. Moisture content values determined as described in section 6.3 (11%). 2. Sp in - lat t ice relaxation times (at 80 MHz) were calculated using the three parameter fit routine to analyze the invers ion- recovery data. 3. Linewidths were calculated as described in section 6.3 (15%). Fig 4.19 T, sensitive image of the same section as in figure 4.18 obtained by nulling the water in zone (ii). Note that the zone lines are clearly v is ible. intensities can be rationalized as fo l lows. . From Table 4.5 and figure 4.18 B, one finds that the three zones have different T x values with zone (/') having the shortest value. With a repetition time of 2 s , one cannot expect to see any changes in image intensity due to Tj contrast. On the other hand, the moisture content varies from 175-225%; although those variations can contribute to minor intensity differences, the major contribution comes from T2 variations as seen from the linewidth data in column 4 of Table 4.5. Zone (/') has the broadest linewidth of the three whereas zone (//) has the narrowest linewidth. In other words, zone (//') has the longest T 2 which contributes more to the image intensity. Hence, in figure 4.18 A the higher intensity of zone (//') is due mainly to T2 contrast. The most obvious feature in this disk sample is the so called "zone l ines". Although not well understood, these zone lines are generally thought to arise because of separation between different mycel ia . ie; they are demarcation zones which are usually less decayed than the portions of wood they separate (228). In this particular sample, the zone line separates zones (//') and (///), both of which are infected. Although these lines are clearly visible in the photograph (figure 4.18 B), they are not easily seen in the normal 2D image (figure 4.18 A) . However, if a Tj sensitive image is obtained, then these zone lines are very prominent as seen in figure 4.19, in which an inversion-recovery sequence was coupled to the imaging sequence, and the null time was set close to the Tj of zone (//). Let us now turn to another disk sample, corresponding to region 3. Again three different areas are seen both in the image and the photograph (figure 4.20 A and B) zone (//'/') appears dark in the image and this is 163 Fig 4.20 Two-d imens iona l image obtained on the 80 MHz wide bore system of a circular disk section (3.5 cm diameter, 1 cm thickness) of the sample used in figure 4.17. The section corresponds to region 3 in Table 4.4. (B) is a photograph of the same sect ion. Experimental parameters : same as in figure 4.13. TABLE 4.6 164 NMR and M.C Variations of Region 3 in Table 4.4 Zone M.C 1 (%) T , J (ms) Linewidth 3 (Hz) (i) 116.3 524115 39.1 (ii) 123.7 451+23 34.4 (iii) 206.3 279121 332.0 1. Moisture content values determined as described in sect ion 6.3 (11%). 2. Spin - lat t ice relaxation times (at 80 MHz) were calculated using the three parameter fit routine to analyze the invers ion-recovery data. 3 . Linewidths were calculated as described in section 6.3 (15%). Fig 4.21 Tj sensit ive image of the same section as in figure 4.20 obtained by nulling the water in zone (ii). Note that the zone lines are clearly v is ib le . again due to short T 2 relaxation (see Table 4.6). Zones (/) and (//') almost have similar linewidths which are considerably narrower and moisture contents that are nearly equal and hence they appear almost equally intense in the image. Because they are flanked by zones with similar characteristics on either s ide, these zone lines do not show up in the image, although visible to the naked eye. However, a Tj sensitive image provides a better contrast, thereby highlighting the presence of this demarcation (figure 4.21). A wide range of fungi are known to occur in wood including certain mucoraceous species , ascomycetes, basidiomycetes and fungi imperfecti (228-231). These can be broadly c lass i f ied into 3 groups. (1) Wood-dest roy ing fungi. (2) Fungi causing soft rot. (3) Wood-s ta in ing fungi and true molds. In addition to these, bacteria can also play an important role in wood decay. Although it was beyond the scope of this work to isolate and characterize these microorganisms, certain general features regarding decay deserve mention in the context of the particular sample studied here. A more general review is available in references 228-231. It is known that decayed wood absorbs moisture more rapidly than sound wood (232). This may account for the high moisture content reflected in regions 5, 7 and 8 in Table 4.4. Another common feature caused by bacterial infection is the phenomenon called "wetwood"; these are regions in heartwood, with abnormally higher moisture content than the adjacent sapwood (233). Moreover, this feature is very common in many hardwoods, such as aspen and this may account for the high moisture content of regions 9 and 10. Although certain fungi were identified in some regions of this sample, these results could not be used to characterize any particular kind of decay. This is because identification of a particular kind of microorganism does not necessari ly rule out the presence of others. In our preliminary studies in particular, the detection of fungi in the presence of bacteria was diff icult as the latter outgrow the former in the culture medium. Hence, a much more extensive microbiological characterization is necessary. Although it was not our intention to study this vast and varied f ie ld from a biological standpoint, these preliminary measurements show that NMR imaging may help to localize fungal populations; obviously much more work, particularly of interdisciplinary nature is needed. 4.4.6.3 Drying of Wood Kiln drying is an important commercial component of the lumber industry. Generally, the water has to be reduced from 2 0 0 - 2 5 0 % in a fresh sample to somewhere between 4 - 2 5 % depending on the subsequent use of the dried sample. With high energy costs , an accurate means for quantitating moisture distribution could lead to insight, and improvements of the drying process. A sample of douglas - f i r wood (dimensions ca. 4x4x1 cm 3 ) impregnated with water by soaking, was imaged on the wide bore system at 80 MHz; the image is shown in figure 4.22 A. This sample was then dried in an oven at 105°C for different time periods and the drying was progressively fo l lowed by NMR imaging. 167 Fig 4.22 Effect of drying wood is illustrated on a douglas - f i r sample. (A) Reference image, M.C. = 178 % (B) M.C. = 100.6 % (C) M.C. = 74.9 % p ) M.C. = 34.6 % . A l l 2D images were acquired on the wide bore system, with the same attenuation and are scaled the same. Experimental parameters ; same as in figure 4.13. 168 B Fig 4.23 Effect of drying wood (contd) (A) represents the same image as in figure 4.22 D, but with different level of attenuation on the spectrometer. (B) represents M.C. = 16.9 % The 2D images shown in figure 4.22 B - D , represent moisture distribution in the sample at different levels of drying. The image intensities in figure 4.22 A - D are scaled the same as each other, but those in figure 4.23 A and B have been acquired with a different attenuation on the spectrometer. This was necessary because the moisture content, being low, yields very little signal and the receiver gain had to be adjusted in order to observe the signal . The images in figures 4.22 D and 4.23 A represent the same moisture content level ; a slight enhancement of intensity is seen in figure 4.23 A . It is clear from this set of images that the water is lost progressively from the surface to the center. These results agree completely with the typical moisture distribution for wood during drying given in figure 4.24 which illustrates a moisture gradient curve across the thickness of a moist wood sample that is being dried. It could be noted that the outer layers dry before those in the interior (213). From the literature the early NMR work on wood -water systems provides evidence for the presence of two water components (189-204). The NMR absorption spectrum exhibits a relatively narrow line attributable to "free water", superimposed upon a much broader line resulting from "bound water". Riggin et a l . (205) have measured the transverse relaxation t ime, T 2 as a function of moisture content for white spruce sapwood and have found the dependence illustrated in figure 4.25. It is clear f rom curve (a) in figure 4.25, that the water has at least two different relaxation t imes, one short and one long. The shorter relaxation time is associated with water molecules that are adsorbed in the cell wal ls and the longer relaxation t ime, with more mobile molecules in the cell cavity. Lately, new 170 surface center surface Fig 4.24 Typical moisture gradient curve across the thickness of a piece of wood during drying. Note that the moisture content of the central section is higher. [From reference (213)] 171 1-Or TIME (mS) 20 40 60 CO | I- . TIME (mS) Fig 4.25 Curve (a) represents the plot of the logarithm of the water proton echo amplitudes of a moist spruce sample against the time measured from the 9 0 ° pulse in a Carr -Purcel l sequence. Normally a single relaxation time T2 leads to straight line with slope (-1/T2) in such a plot, but it is clear from the figure that atleast two different time constants, one long and one short are required to describe the experimental data. Curve (b) represents the shorter component on a different time scale. [From reference (205)] 4.26 Series of spectra reflecting the changes in NMR lineshape for douglas - f i r sample during the process of drying. TABLE 4.7 NMR Properties and M.C Variations of Douglas-fir Wood During Drying 1 Figure 4.26 M.C 2 T, 3 (ms) Linewidth4 (Hz) A 182.6 122±6 30.0 B 148.4 111±5 45.0 C 118.0 10517 58.2 D 90.6 90±9 67.6 E 63.3 81± 10 69.4 F 36.4 59±13 67.6 G 16.6 16±6 63.8 1. Column 1 corresponds to spectra in figure 4.26. 2. Moisture content values determined as described in section 6.3 (±1%). 3. Spin-lattice relaxation times (at 80 MHz) were calculated using the three parameter fit routine to analyze the inversion-recovery data. 4. Linewidths were calculated as described in section 6.3 (±5%). 174 interest has developed in characterizing water in wood by NMR lineshape analysis. Hailey et al . (234) have reported a method of analyzing the FID signal to calculate the amount of mobile or free protons, from which they can compute the moisture content. Getting back to our system under study, figure 4.26 shows series of NMR spectra obtained in this work at 80 MHz, corresponding to a similar douglas - f i r sample as that used to obtain images shown in figures 4.22 and 4.23. The linewidth data and Tl values were measured at different levels of moisture contents and are listed in Table 4.7. It can be seen that the spectra in figure 4.26 D and E deviate more from the Lorentzian behavior indicating the presence of at least two components. However, it is the narrow linewidth component (ie: free water) that contributes to the image intensity, and on drying the amount of free water decreases, and this causes the image intensity to decrease. Furthermore, the T1 values also decrease due to the removal of free water. Although these NMR spectroscopic data provide a qualitative understanding of the drying process, what is more important is to get quantitative information from the NMR image itself. Of course, this is challenging and therefore future work should be directed towards achieving this goal . 4.4.6.4 Impregnation of Wood Impregnation of w o o d , frequently under pressure, provides an important means for protecting it mainly from rot and f ire. In order to illustrate the use of NMR imaging to fo l low such impregnation processes, a sample of douglas - f i r wood (dimensions, ca . 4x4x1 cm 3 ) was treated with 10 mM MnCI 2 solution and imaged at different levels of impregnation. The series of 2D images in figure 4.27 show at different time intervals the consequences of impregnation. Figure 4.27 A represents an image at time t = 0 ; that is the normal image of the sample that had been soaked in water for over 7 days. The sample was then partially immersed in a solution of 10 mM MnCI 2 for different time periods as indicated in the figure caption. As the manganous chloride s lowly diffuses into the w o o d , it 2 + carries with it the paramagnetic Mn ions into the wood f ibers ; these ions cause the water in its environment to relax so rapidly that any signal from that region is eliminated. In an alternative approach a small portion of a paramagnetic salt was embedded inside a similar moist wood sample and the dif fusion of the paramagnetic metal ions was fo l lowed . The 2D image in figure 4.28 A is that of a normal sample (dimensions ca. 4x4x1 cm 3 ) with a small circular cavity (diameter 8 m m , depth 7mm) in the center. This cavity was ' then f i l led with solid CuS0 4 .5H 2 0 and the diffusion of the Cu^ + ions was fo l lowed by NMR imaging. The images at varying time intervals are shown 2 + 2 + in figures 4.28 B -F . As in the experiments using Mn ions, the Cu ions change the relaxation rates drastically in their environment so that no signal is seen from those regions. This clearly demonstrates an inherent advantage of NMR imaging, in that the changes induced by chemical means can be detected and fo l lowed. Using paramagnetic labels, it is also possible to evaluate both the depth of penetration as well as the speed of penetration from these images especial ly in the latter example. A s expected the dif fusion of the ions 175a Fig 4.27 Effect of impregnation of wood with paramagnetic ions is illustrated on another douglas- f i r sample. (A) Reference image, t = 0 (B) t = 15 mins (C) t = 6 hrs (D) t = 3 days (E) t = 4 days (F) t = 7 days . A l l 2D images are scaled the same and 10 mM MnCI 2 solution was used to soak the sample. Experi mental parameters : same as in figure 4 .13. 76a Fig 4.28 Effect of impregnation of wood with paramagnetic ions. The diffusion of the C u 2 + ions embedded within the douglas-fir sample is followed. (A) Reference image, t = 0 (B) t = 4 hrs (C) 15 hrs (D) 3 days (E) 10 days (F) 24 days . Note, that the diffusion of the paramagnetic ions along the wood fibers is faster than across the fibers as expected. Experimental parameters ; same as in figure 4.13. along the length of the fibers (ie: perpendicular to the plane of the image in figure 4.28) are much more rapid than across the fibers and it takes nearly 15 hrs for the signal intensity f rom the cavity to be zero; ie: it takes approximately 15 hrs to penetrate 3 mm. 4.4.6.5 Specialty Products from Wood A s wood f rom mature trees comes into increasingly short supply, the forest industry has had to resort to the use of smaller, younger trees; these result in s m a l l - s i z e d low quality timber, and a high proportion of juvenile wood . Fortunately, new manufacturing techniques have enabled the production of high quality f inished products f rom such low quality timber. P l ywood , for example, is prepared by cross - laminat ing and gluing thin veneers that results in strong, warp-resistant panels (235,236). Clearly, evaluation of the internal architecture of such composites is a potential use for NMR imaging. Figure 4.29 A and B illustrate crosssectional images obtained from two samples of white spruce p lywood (dimensions ca. 1.9x1.6x0.9 cm 3 ) saturated with water under two different circumstances but for the same length of t ime, one under pressure and the other by plain soaking for 12 hrs. A s seen in these images, penetration of water is more diff icult along certain layers than along others. This reflects the fact that the water conducting fibers are shorter along one dimension compared to the other as seen in the photograph of the sample (figure 4.29 C). Thus, on soaking water scarcely penetrates into those layers,, as seen in the image in figure 4.29 B. However, under pressure the situation is different and equal image c Fig 4.29 Two-d imens iona l images of two similar white spruce plywood samples with different levels of water saturation. (A) saturated under pressure and (B) by plain soaking. (C) is a photograph of a sample specimen (see text for details). Experimental parameters \ sweep width = ± 7042.25 Hz; acquisition time = 36.35 ms ; block size = 1024 ; gradient increment = 0.02 and 0.016 G/cm ; static gradient = 0.62 G/cm ; no of experiments = 128. B Fig 4.30 (A) Two-d imens ional image of a edge-glued white spruce p lywood sample with water saturation. The glue line can be visualized in the image. (B) is the photograph of the sample. Experimental parameters : sweep width = ± 7042.25 Hz; acquisition time = 36.35 ms ; block size = 1024 ; gradient increment = 0.016 G/cm ; static gradient = 0.62 G/cm ; no of experiments = 128. intensities are observed from all layers, reflecting equal extent of penetration. Edge-gluing is another technique used to produce specialty products. This study now shows that provided the sample has enough water, the glue line can be visual ized. This is shown in figure 4.30 A which illustrates a two-d imensional image of an edge glued white spruce p lywood sample (dimensions ca . 4.4x3.2x0.9 cm 3 ) , saturated with water by soaking for 4 days (figure 4.30 B). The demarcation in the middle indicates the glue line. In addition the earlywood/latewood separation (ie: growth rings) are also seen in the image. Given that one of the major problems in p lywood manufacture is that nothing is known about the continuity of the glue line when the veneers are glued together, it may be that as demonstrated here, NMR imaging can help to provide insight since if the line is not continuous, water can penetrate into the adjacent layers which should be easily detectable in an image. CHAPTER 5 SUMMARY AND CONCLUSIONS 5.1 SUMMARY AND CONCLUSIONS In this Chapter, an attempt wi l l be made to summarize briefly the principal findings of this study and to speculate on certain future possibi l i t ies in this rapidly advancing f ield of NMR imaging. Although there were a number of different imaging techniques described in the literature in 1982, little was known outside the imaging community ; a clear understanding, particularly in terms of practical details and di f f icul t ies , was hard to come by, even in 1986. One of the primary concerns of this thesis is to provide this insight, concentrating on the two-d imensional FT imaging methods on which we chose to focus our interest. In recent years, medical imaging studies have almost exclusively been done by this imaging . technique and hence, the discussion in section 2.3 should serve as a review, outlining the basic concepts and formal isms. This clearly helped our laboratory to pursue 2DFT imaging very widely in our studies which have been appropriately reported in the literature. "Phantoms" are an integral part of NMR imaging sc ience; their use in imaging studies is essential in assessing image quality and in evaluating imaging systems. Moreover, when one is in pursuit of new experimental sequences use of phantoms is mandatory. The application of phantom studies to evaluate conventional high-resolution NMR spectrometers modif ied to perform imaging has been presented in section 2.4. This also includes the first demonstration of using imaging techniques to map the Bj homogeneity (section 2.4.3). In addit ion, gradient magnitudes and homogeneity of the static magnetic f ield were also evaluated using this simple approach. Quantitation of image parameters (eg: p, Ju and T 2) and computing separate images is quite challenging. Although this has been done on phantoms, applications to real systems are yet to be explored. However, a convenient experimental procedure was described to calculate the relative spin density values directly from the images by means of phantoms (section 2.4.4.1). The search for an NMR Microscope led to the work described in section 2.4.4.2, in which a conventional 10 mm NMR probe fitted with new gradient coi ls was used to image phantoms, which consisted of thin glass capillary tubes (internal diameter 140-220 jum) containing water. This work, done in collaboration with Mr. S.D. Luck, was the first demonstration of achieving microscopic resolution in NMR imaging. As mentioned previously in section 2.4.4.2, this is one of the areas of imaging where much interest is being focussed now. Chapter 3 dealt with Chemical Shift Imaging (CSI), the potential of which, though recognized by our laboratory in the early stages, was relatively less appealing to many other workers. It was felt that the preservation of the chemical shift data in imaging wi l l indeed provide more information regarding the chemistry of the system and hence, the work in our laboratory was directed towards this objective. This motivation led to the pursuit of initially high f ield 2D CSI (section 3.2) and subsequently to the first demonstration of 4D CSI (section 3.4). The main limitation of this technique is the total experimental t ime, and a detailed description of the effect of dimension size on the experimental time has been presented. However the fact remains that all the information is contained in one single experiment and the general potential of the use of rapid imaging schemes to shorten the experimental t imes remains to be fully explored. This Chapter also includes a discussion on Chemical Shift Art i facts to enable understanding of the origin of these undesirable features. In addition data manipulation methods have been described to minimize and eliminate these artifacts from chemical shift imaging data. Chapter 4 summarizes all applications pertaining to non-medical systems undertaken in the work of this thesis. Most of this work has dealt with an exploratory study of the use of NMR imaging methods and three areas of interest have been covered; Chromatography, Porous rocks and W o o d . The imaging techniques used are those described in the earlier Chapters. However, in some cases these techniques have been modif ied to suit specif ic needs. Most of the work has been performed on the 80 MHz sys tem. Some studies have also been performed on the who le -body NMR scanner at the UBC Health Sciences Hospital . The initial system studied was in relation to Chromatography, and it was demonstrated that by changing the chemistry of the system useful spatial information can be obtained (section 4.2). Two points merit comment; one concerns T x and the other is the idea of "molecular ampl i f iers" . Although both these features go together, the latter offers the possibi l i ty of indirect determination of the paramagnetic metal ion which is present in such very small amounts that direct detection by NMR is impossible . This preliminary investigation warrants further study and the general potential to study Chromatography Columns depends on the ability to image f lowing fluids and to observe chemical reactions in columns. Such analytical techniques are yet to be fully investigated for chemical systems. NMR imaging application to porous rock samples was presented in section 4.3. In addition to being able to map the distribution of a single f luid species (eg: water or oi l ) within porous rocks, it was shown for the first time that the distributions of fluids of different chemical composit ion coexisting in the same porous sample can be mapped simultaneously by means of CSI (section 4.3.4.3). Furthermore, Tj sensitive chemical shift resolved thin slice images of a sandstone sample saturated with oil and water/D 2 0 mixture was also presented. This clearly demonstrated a method of manipulating the differences in J1 values of the oil and water components advantageously. However, if these differences are signif icantly smaller, as in the case of a limestone sample saturated with North Sea crude o i l , then 4D CSI is the alternate imaging technique to pursue (section 4.3.4.3). Thus, selection of the correct imaging technique is crucial and should be done, based on the chemical properties of the system under investigation. Finally, applications pertinent to Forestry were described in section 4.4. Several different features including internal structure, defects, decay, drying and impregnation have been studied from an NMR imaging standpoint. As mentioned previously, large samples were imaged on the who le -body NMR scanner, including all freshly felled samples and some dry samples saturated with water under pressure. One decayed sample of fresh aspen in particular was later studied on the 80 MHz system with a view of generating more insight in identifying decay. Microbiological tests have been performed on this sample to confirm decay init ial ly, and then to identify certain microorganisms associated with decay. Although the results are not conclusive in characterizing microorganisms responsible for decay by NMR, the potential exists and therefore more extensive NMR studies of many different sample species together with detailed microbiological tests are necessary. Another interesting feature that has been presented involves studying different processes such as drying and impregnation of wood by NMR imaging (sections 4.4.6.3 and 4.4.6.4). It has to be mentioned that determination of both the different components of water present and the moisture content of wood have been a long standing interest of many NMR Spectroscopists . Although detailed quantitation of all components was not attempted in this study, based on the studies of these previous workers it can be assummed that the major contribution towards the image intensity is from the free water component. The work described on impregnation of wood is based on the relaxation properties of the paramagnetic metal ions which help to locate their presence. Lastly, NMR images of some specialty products f rom wood impregnated with water have been presented. These offer the possibi l i ty of studying f inger- joint ing and edge-gluing processes and other special products such as part ic le-boards and chip -boards . In general, provided suff icient water is present, all these can be studied by NMR imaging. Thus, it can be seen that there exists a whole array of new opportunities. The results discussed in Chapter 4 are new and thus serve as the first demonstration of NMR imaging applications to these respective non-medica l areas. At the time the author started these studies, there was only one literature report relevent to non-medica l applications. Interest has grown very s lowly with one more report on polymer composites during the course of the author's work. However, very recently more reports are beginning to appear, particularly in relation to wood and some porous rock samples. Clearly, this demonstrates the increased interest of imaging non-medica l systems and it is felt that 1986 and subsequent years wi l l witness many more reports in the literature exclusively focussing on these systems. CHAPTER 6 EXPERIMENTAL 6.1 THE IMAGING SPECTROMETERS Principally two systems were used in this study; both located in the Department of Chemistry at U.B.C. 6.1.1 NARROW BORE SYSTEM The initial studies of this thesis were started with this system which provided some of the results described in sections 2.4 and 3.2.3. The magnet is an Oxford Instruments superconducting solenoid with a bore size of 54 m m , and a f ield of 6.35 T corresponding to a resonance frequency for protons of 270 MHz. The system is supported by a Nicolet 1180 computer and a 293 B pulse programmer. The standard Nicolet NTCFTB programme modif ied to incorporate the imaging software developed at U.B.C. by Dr. S . Sukumar was used for data acquisit ion. However, data processing was performed using the routine' 2D NMR software available from the manufacturer. Data storage is on 1 Mbyte disks which are driven on a Diablo (model -30) disk drive. Home-bui l t 5 mm and 10 mm probes were used; since the latter was a probe designed for observing C - 1 3 , the decoupler coi l was used to observe protons. The x and y shim coils were used to produce the magnetic f ield gradients for most studies on this system. However, microscopic imaging was performed on a 10 mm probe fitted with x and y gradient coi ls which are capable of providing gradient f ie ld strengths of 4 G/cm (section 2.4.4.2). The display of all images generated on this system is by means of "stacked p lots" or "contour p lots" made with a Zeta digital plotter using the same plotting software used for Chemical 2D NMR Spectroscopy. 6.1.2 WIDE BORE SYSTEM The majority of the work described in this thesis was performed with this sys tem, which consists of ; a 31 cm bore 1.89 T superconducting solenoid fitted with gradient co i ls , f rom Oxford Instruments, a Nicolet 1280 computer and a 293 C pulse programmer, a Phoenix disc drive capable of storing 80 Mbytes of data on the fixed platter and 16 Mbytes of data on the removable disk cartridge, a raster display scope, a Ramtek colorgraphic display sys tem, and a Ze ta -8 digital plotter. A s for the narrow bore system, the imaging software developed at U.B.C. by Dr. S . Sukumar was incorporated into the standard NMR programme, so as to enable data acquisit ion. The software for colorgraphic display was again written at U.B.C. by Dr. S . Sukumar. A number of rf probes have been used on this s y s t e m ; these have been developed by Mr. S.L. Talagala, another graduate student, together with two Departmental technicians, Mr. T. Marcus and Mr. C. Neale. Initially, a 2 turn Helmholtz coi l (10 cm diameter) was used; the work described in section 4.2 has been done using such co i l s . As mentioned previously, these coi ls tend to have very long pulse widths; for example, a 180° pulse on a 4 cm diameter spherical bulb is ca. 880 MS when the coi l is driven by a 140 W rf amplif ier . These rf probes were modified from time to time and at present, it is possible to obtain a 180° pulse length of 120 MS for the same 4 cm sphere using the same rf amplif ier and a resonator rf probe. Most studies were performed with a 7.5 cm diameter axial resonator c o i l ; others used a 7.5 cm transverse, and 12.5 cm axial, resonator co i ls . The system itself has also been continually modi f ied. Some major modif icat ions include; installation of a new set of gradient coi ls capable of producing 1.6 G/cm gradients and with reduced noise level , use of a rf power amplifier (ENI model LP I — 10) capable of generating 1 KW power, and a Nicolet f loppy disk (8") system mainly for data archiving. With the new power amplif ier , the 180° pulse length is reduced to 33 us for the same sample. The maximum matrix size that can be displayed on the colorgraphic system is 256x256 and almost all color images, particularly those in Chapter 4, have been optimized to this aspect. Usually, 128 experiments were acquired, and during data processing, zero f i l l ing was done to extrapolate the size of this second dimension to 256 data points. However, no zero f i l l ing was done along the observe dimension. Unless stated otherwise, all 2D images have been displayed with the observed dimension along the vertical axis and the horizontal axis displays information along the phase encoding dimension. In the case of 3D imaging, the final image displays information from both phase encoding dimensions. However, zero f i l l ing is done to yield a final image matrix size of 256x256. A 35 mm Yashica camera (model F X - D , SE) with Nikon zoom lens and close up lens kit (model 3T and 4T) are used to photograph images direct f rom the Ramtek screen. It should be noted that both slides as well as prints can be obtained from this s e t - u p . 193 The color scale and the. intensities corresponding to each level are shown in Table 6.1. This pseudo blue scale represents intensity variations as f o l l o w s ; high intensity regions are displayed white and progressively low intensity regions are progressively a darker blue. The intensity levels fo l low the same pattern as the contour levels in the standard NMR programme (237). The top level represents 70% of the highest intense peak and each subsequent level represents 70% of the former level . A s it can be seen with this 8 level color scale it is possible to go as low as 8.1% of the highest intensity. 6.2 SETTING UP OF AN EXPERIMENT A s there exists no standard rf probe for the wide bore sys tem, each different probe has to be tuned and matched to the characteristic resonant frequency of the sys tem, that is for protons at 80 MHz. At the beginning of this work manual shimming of the f ield was performed on the sample itself. The FID signal amplitude and shape were optimized by adjusting the currents passing through the different shim co i ls . With the later addition of an automatic computer shimming unit, the shimming routine could be pre-programmed leaving the computer to do the shimming. Depending on the system studied, the shimming times can vary from 10 to 45 mins. In all cases, the shimming routine described by Conover (238) was fo l lowed. Pulse length determination involved observation of the signal intensity after a simple FT NMR experiment. The 180° pulse length, for example, is determined by observing the first zero intensity c ross -over point. S imi lar ly , TABLE 6.1 194 Intensity Changes Corresponding to Each Color Level in Colorgraphic Display 1 Level Color Intensity (%) 1 White 70.0 - 100 2 Light Blue 49.0 - 69.9 3 34.3 - 48.9 4 24.0 - 34.2 5 16.8 - 23.9 6 11.8 - 16.7 7 Dark Blue 8.2 - 11.7 8 Black 0 - 8.1 1. These intensity estimates are valid only if the tallest peak in the data set is scaled as the highest intense peak. This was adhered to all results described in this work. the first intensity maximum is sought for a 90° pulse. Whenever 180° pulses were used, the pulse imperfections were minimized by applying composite pulses (239). Although detailed descriptions of individual timing sequences are not particularly useful because they depend on the system studied, certain features merit comment. It was found necessary in all imaging experiments to al low a delay before and after the gradients are switched on and off respectively to account for the rise and fal l t imes of these magnetic f ield gradients. The magnitudes of the applied f ie ld gradients and the relaxation properties of the system investigated determine these delay t imes, but generally delays between 5 - 4 5 ms were used. The magnitudes of the gradients were selected on the basis of the f ield of view criteria outlined in section 2.3.4 and on the inhomogeneity of the main f ie ld . Data processing was done as described previously in Chapters 2 and 3. The standard Nicolet 2D NMR spectroscopy software was used to perform this. The normalization constants for scaling the spectra were obtained first by locating the largest s ignal , and then setting the scaling factor with respect to this spectrum prior to Fourier transformation. In the updated version (version no. 50222) of this software, this procedure can be done by the computer itself . 6.3 SAMPLES Al l phantoms were constructed at U.B.C. and most credit goes to Mr. C. Neale of the Mechanical workshop in the Department of Chemistry. Chelating Sepharose 6B used in section 4.2 was obtained from Pharmacia Fine Chemicals. A l l solvents used were of spectral grade, and all aqueous solutions were prepared quantitatively fo l lowing the usual procedure. Porous rock samples were kindly supplied by Dr. C. Hall of Schlumberger Research, Cambridge, UK. The limestone sample saturated with north sea crude o i l , studied in section 4.3.4.3, however, was supplied by the British Petroleum Corporation. The saturation of these rock samples were done in the laboratory as f o l l o w s ; one end of the sample is al lowed to sit inside a beaker containing water until it gets partially saturated by capil larity. More water is then added to completely immerse the sample and with the beaker placed inside a dessicator, vacuum suction (ca. 15 mm of Hg) is applied usually for 30 mins. After releasing the pressure more suction is again applied. This process is repeated until the weight of the saturated sample becomes almost constant. In the case of the emulsion containing oil and H 2 0/D 2 0 (section 4.3.4.3) this process took nearly 7 days. The wood samples were kindly supplied by Dr. P.R. Steiner of Forintek Research, Vancouver, BC. Freshly fel led samples of aspen, in particular, were sent from Grand Prairie, Alberta. These disk samples were stored in sealed plastic bags and kept cold (below 2°C) immediately after cutting. They were examined 4 days after fe l l ing. A l l other wood samples were either saturated under pressure by a similar process described above or by immersing them in water over a long period of t ime, usually for about 10-14 days. The samples are usually sealed in Saran wrap during examination. Initially, the freshly fel led samples were analyzed on the whole body PICKER NMR imaging system (Cryogenic model 2055) operating at a proton resonance frequency of 6 MHz and located at the U.B.C. Health Sciences Hospital . Certain sets of images in sections 4.4.6.1 and 4.4.6.2 were obtained on this tomograph by Ms . W.A. Stewart in this laboratory. Sl ice selection was employed on all samples studied on this scanner and the slice thickness is 1 cm in all cases. The T t values which were directly computed from the image are also listed in section 4.4.6.1. One of the infected samples was later studied on the 80 MHz system in the Chemistry Department. In this system, the T, values and the linewidth data are obtained by operating the system in the spectroscopy mode. As mentioned previously the T t values are either computed using the three parameter single exponential fit routine to analyze the the inversion-recovery data or determined by the null point method. The linewidths were calculated from plots of spectra by determining the width at half -height. The moisture contents of the wood samples were calculated according to the standard procedure described in section 4.4.3, in which the sample is dried at 105°C to constant weight. A l l weights were recorded on a precision balance (± 5 mg). A l l mycological tests were made by Dr. P.A. Salisbury of the Chemistry Department Microbio logy laboratory. 198 BIBILIOGRAPHY. 1. Bloch, F., Hansen, W.W., Packard, M. Phys. Rev. (1946) 69, 127. 2. Bloch, F. Phys. Rev. (1946) 70, 460. 3. Bloch, F., Hansen, W.W., Packard, M. Phys. Rev. (1946) 70, 474. 4. Purcell, E.M., Torrey, H.C, Pound, R.V. Phys. Rev. (1946) 69, 37. 5. Ernst, R.R., Anderson, W.A. Rev. Sci. Instrum. (1966) 271. 93. 6. Shaw, D. Fourier Transform NMR Spectroscopy, Elsevier: Amsterdam, 1984. 7. Martin, M.L., Martin, G.J., Delpuech, J.J. Practical NMR Spectroscopy, Heyden: London, 1980. 8. Mullen, K., Pregosin, P.S. Fourier Transform NMR techniques: A Practical Approach, Academic Press: London, 1976. 9. Slichter, C.P. Principles of Magnetic Resonance, Springer-Verlag: New York, 1980. 10. Farrar, T.C., Becker, E.D. Pulse and Fourier Transform NMR, Academic Press: New York, 1971. 11. Abragam, A . The Principles of Nuclear Magnetism, Oxford University Press: Oxford, 1961. 12. Harris, R.K. Nuclear Magnetic Resonance Spectroscopy, Pitman: London, 1983. 13. Fukushima, E., Roeder, S.B.W. Experimental Pulse NMR: A Nuts and Bolts Approach, Addison-Wesley : Massachusetts, 1981. 14. Odeblad, E., Lindstrom, G. Acta Radiol. (1955) 43, 469. 15. Damadian, R. Science (1971) 171. 1151. 16. Lauterbur, P.C. Nature (1973) 242, 190. 199 17. Kumar, A., Welti, D., Ernst, R.R. J. Magn. Reson. (1975) 18, 69. 18. Jeener, J . Ampere International Summer School, Basko Polje, Yugoslavia, 1971. 19. Aue, W.P., Bartholdi, E., Ernst, R.R. J. Chem. Phys. (1976) 64, 2229. 20. Mansfield, P., Morris, P.G. NMR Imaging in Biomedicine, Academic Press: New York, 1982. 21. Hinshaw, W.S., Bottomley, P.A., Holland, G.N. Nature (1977) 270. 722. 22. Bottomley, P A . Rev. Sci. Instrum. (1982) 53, 1319. 23. Gummerson, R.J., Hall, C., Hoff, W.D., Hawkes, R., Holland, G.N., Moore, W.S. Nature (1979) 28_L 56. 24. Rothwell, W.P., Holecek, D.R., Kershaw, J.A. J. Polym. Sci: Po/ym. Lett. (1984) 22, 241. 25. Torrey, H.C. Phys. Rev. (1949) 76, 1059. 26. Gadian, D.G. Nuclear Magnetic Resonance and its Applications to Living Systems, Clarendon: Oxford, 1982. 27. Gabillard, R. Rev. Sci. Paris. (1952) 90, 307. 28. Carr, H.Y., Purcell, E.M. Phys. Rev. (1954) 94, 630. 29. Bradford, R., Clay, C., Strick, E. Phys. Rev. (1954) 84, 157. 30. Page 33 of reference 20. 31. Brunner, P., Ernst, R.R. J. Magn. Reson. (1979) 33, 83. 32. Frahm, J., Merboldt, K.D., Hanicke, W., Haase, A . J. Magn. Reson. (1985) 64, 81. 33. Hayes, C.E., Edelstein, W.A., Schenck, J.F., Mueller, O.M., Eash, M. J. Magn. Reson. (1985) 63, 622. 34. Joseph, P.M. J. Comput. Assist. Tomogr. (1985) 9, 651. 35. Axel, L., Dougherty, L. J. Magn. Reson. (1986) 66, 194. 200 36. De Luca, F., Luzzi, M., Crescenzi , A . , De Simone, B.C., Campanella, R., Casier i , C., Maravigl ia, X.Y. J. Magn. Reson. (1986) 67, 7. 37. Frahm, J . , Haase, A. , Matthaei, D. J. Comput. Asst. Tomogr. (1986) JO, 363. 38. Hornak, J.P., Ceckler, T.L., Bryant, R.G. J. Magn. Reson. (1986) 68, 319. 39. Granot, J . J. Magn. Reson. (1986) 68, 575. 40. Park, H.W., Cho, Z.H. Magn. Reson. Med. (1986) 3, 448. 41. Hal l , L.D., Sukumar, S . J. Magn. Reson. (1982) 50, 161. 42. Edelstein, W.A., Hutchison, J.M.S., Johnson, G., Redpath, T. Phys. Med. Biol. (1980) 25, 751. 43. Ljunggren, S . J. Magn. Reson. (1983) 54, 338. 44. Brown, T.R., Kincaid, B.M., Ugurbil, K. Proc. Natl. Acad. Sci. USA. (1982) 79, 3523. 45. Twieg, D.B. Med. Phys. (1983) JO, 610. 46. Hinshaw. W.S., Lent, A .H. Proc. IEEE. (1983) 71, 338. 47. Locher, P.R. Philips Tech. Rev. (1984) 41, 73. 48. King, K.F., Moran, P.R. Med. Phys. (1984) JJ., 1. 49. Manassen, Y., Navon, G. J. Magn. Reson. (1985) 61., 363. 50. page 174 of reference 20. 51. Edelstein, W.A., Bottomley, P.A., Hart, H.R., Leue, W.M. , Schenck, J.F., Redington, R.W. Proceedings of the Symposium on Nuclear Magnetic Resonance Imaging, W i t c o f s k i , R.L., Karstaedt, N., Partain, C.L. Eds., Bowman-Gray School of Medicine, Wins to n - S a le m, NC, 1982, p 139. 52. Andrew, E.R. Acc. Chem. Res. (1983) J6, 114. 53. Edelstein, W.A., Bottomley, P.A., Hart, H.R., Smith, L.S. J. Comput. Assist. Tomogr. (1983) 7, 391. 201 54. Edelstein, W.A., Bottomley, P.A., Pfei fer , L.M. Med. Phys. (1984) 1_1, 180. 55. Trappeniers, N.J., Gerritsma, C.J., Oost ing, P.H. Phys. Lett. (1965) 18, 256. 56. Murday, J .S. , Cotts, R.M. Naturforsch. (1971) 26a, 85. 57. Murday, J .S . J. Magn. Reson. (1973) 1 0 , 111. 58. Hall , L.D., Luck, S.D., Sukumar, S . J. Magn. Reson. (1985) 63, 183. 59. Hal l , L.D., Rajanayagam, V., Sukumar, S . J. Magn. Reson. (1984) 60, 199. 60. Lai , C - M . Phys. Med. Biol. (1983) 28, 925. 61. Lai , C - M . J. Phys. E: Sci. I nst rum. (1983) 16, 34. 62. O'Donnell, M., Edelstein, W.A. Med. Phys. (1985) 12, 20. 63. Hutchison, J .M.S. , Sutherland, R.J., Mal lard, J.R. J. Phys. E: Sci. Instrum. (1978) U, 217. 64. Lai , C - M . J. Phys. E: Sci. Instrum. (1982) 15, 1093. 65. Sekihara, K., Kuroda, M., Kohno, H. Phys. Med. Biol. (1984) 29, 15. 66. Maudsley, A .A. , Oppelt, A. , Ganssen, A . Siemens Forsch.-u. Entwickl.-Ber. (1979) 8 , 326. 67. Maudsley, A .A . , S imon, H.E., Hi Ial, S.K. J. Phys. E: Sci. Instrum. (1984) 17, 216. 68. Sekihara, K., Matsui , S. , Kohno, H. J. Phys. E: Sci. Instrum. (1985) Jj$, 224. 69. Sekihara, K., Matsui , S. , Kohno, H. IEEE Trans. Nucl. Sci. (1985) 32 , 799. 70. Bendel, P., Lai , C - M . , Lauterbur, P.C. J. Magn. Reson. (1980) 38 , 343. 71. Hoult, D.I., Richards, R.E. Proc. R. Soc. Lond. A. (1975) 344, 311. 72. Hoult, D.I. Prog. NMR Spec. (1978) 12, 41. 202 73. Grist , T.M., Hyde, J .S. J. Magn. Reson. (1985) 61, 571. 74. Cross, T.A., Muller, S., Aue, W.P. J. Magn. Reson. (1985) 62, 87. 75. Hal l , L.D., Marcus, T., Neale, C , Powel l , B., Sa l los , J . , Talagala, S.L. J. Magn. Reson. (1985) 62, 525. 76. Schnal l , M.D., Subramanian, V.H., Leigh JR., J.S., Gyulai , L., Mclaughlin, A. , Chance, B. J. Magn. Reson. (1985) 63, 401. 77. Bax, A . Two-Dimensional Nuclear Magnetic Resonance in Liquids, Delft University Press: Holland, 1982, p 26. 78. Schneiders, N.J., Post , H., Brunner, P., Ford, J.J., Bryan, R.N., Wi l lcott , M.R. Med. Phys. (1983) 10, 642. 79. Schneiders, N.J., Ford, J .J . , Bryan, R.N. Med. Phys. (1985) 12, 71. 80. Bakker, C.J.G., de Graaf, C.N., van Dijk, P. Phys. Med. Biol. (1984) 29, 1511. 81. Lerski , R A . , Straughan, K., Orr, J .S. Phys. Med. Biol. (1984) 29, 271. 82. Hall , L.D., Holme, K., Sukumar, S . J. Magn. Reson. (1985) 61, 52. 83. Bloembergen, N., Purcell , E.M., Pound, R.V. Phys. Rev. (1948) 73, 679. 84. Carrington, A. , McLachlan, A.D. Introduction to Magnetic Resonance, Harper & Row: New York, 1967, p 193. 85. Lauterbur, P.C. Pure Appl. Chem. (1974) 40, 149. 86. Hinshaw, W.S. J. Appl. Phys. (1976) 47, 3709. 87. Delayre, J.L., Ingwall, J.S., Mal loy , C., Fosse l , E.T. Science (1981) 212. 935. 88. Bottomley, P.A. J. Phys. E: Sci. I nst rum. (1981) 14, 1081. 89. Hall , L.D., Luck, S.D., Rajanayagam, V. J. Magn. Reson. (1986) 66, 349. 90. Johnson, G A . , Thompson, M.B., Gewalt , S.L., Hayes, C.E. J. Magn. Reson. (1986) 68, 129. 91. Eccles, C.D., Callaghan, P.T. J. Magn. Reson. (1986) 6 8 , 393. 92. Aguayo, J.B., Blackband, S.J., Schoeniger, J., Mattingly, M A . , Hintermann, M. Nature (1986) 3 2 2 190. 93. Hoult, D.I., Richards, R.E. J. Magn. Reson. (1976) 2 4 , 71. 94. Bersohn, R. Annu. Rev. Phys. Chem. (1960) H , 369. 95. Shulman, R.G. Annu. Rev. Phys. Chem. (1962) . 1 3 , 325. 96. Cohn, M., Rao, B.D.N. Frontiers in Physiochemical Biology, Pullman, B. Ed., Academic Press: New York, 1978, p 191. 97. Brown, T.R. J. Comput. Assist. Tomogr. (1981) 5 , 301. 98. Cohn, M., Hughes Jr., T.R. J. Biol. Chem. (1960) 2 3 5 , 3250. 99. Moon, R.B., Richards, J.H. J. Biol. Chem. (1973) 2 4 8 , 7276. 100. Hollis, D.P., Nunnally, R.L., Taylor IV, G.J. J. Magn. Reson. (1978) 2 9 , 319. 101. Bocian, D.F., Chan, S.I. Annu. Rev. Phys. Chem. (1978) 2 9 , 307. 102. Jacobs, R.E., Oldfield, E. Prog. NMR Spec. (1981) 1 4 , 113. 103. Wyrwicz, A .M. , Schofield, J.C., Burt, C.T. Noninvasive Probes of Tissue Metabolism, Cohen, J.S. Ed., Wiley & Sons: New York, 1982, p 149. 104. Chance, B., Nakase, Y., Bond, M., Leigh Jr., J.S., McDonald, G. Proc. Natl. Acad. Sci. USA. (1978) 7 5 , 4925. 105. Shulman, R.G., Brown, T.R., Ugurbil, K., Ogawa, S., Cohen, S.M., den Hollander, J.A. Science (1979) 2 0 5 . 160. 106. Behar, K.L., den Hollander, J.A., Stromski, M.E., Ogino, T., Shulman, R.G., Petroff, O.A.C., Prichard, J.W. Proc. Natl. Acad. Sci. USA. (1983) 80, 4945. 107. Shaw, D. J. Comput. Assist. Tomogr. (1981) 5, 299. 204 108. Ackerman, J.J., Grove, T.H., Wong, G.G., Gadian, D.G., Radda, G.K. Nature (1980) 283, 167. 109. Nunnally, R.L., Bottomley, P A . Science (1981) 211. 177. 110. Gordon, R.E., Hanley, P.E., Shaw, D., Gadian, D.G., Radda, G.K., Styles, P., Bore, D.J., Chan, L. Nature (1980) 287, 736. 111. Rosen, B.R., Pykett, I.L., Brady, T.J. Nuclear Magnetic Resonance Chemical Shift Imaging, Chapter 17 and references therein. 112. Lauterbur, P.C., Kramer, D.M., House JR., W.V., Chen, C - N . J. Am. Chem. Soc. (1975) 97, 6866. 113. Cox, S.J., Styles, P. J. Magn. Reson. (1980) 40, 209. 114. Pykett, I.L., Rosen, B.R. Radiology (1983) 149, 197. 115. Haselgrove, J.C., Subramanian, V.H., Leigh Jr., J.S., Gyulai, L., Chance, B. Science (1983) 220, 1170. 116. Maudsley, A.A. , Hi Ial, S.K., Perman, W.H., Simon, H.E. J. Magn. Reson. (1983) 51, 147. 117. Hall, L.D., Sukumar, S. J. Magn. Reson. (1984) 56, 314. 118. Mareci, T.H., Brooker, H.R. J. Magn. Reson. (1984) 57, 157. 119. Hall, L.D., Rajanayagam, V., Sukumar, S. J. Magn. Reson. (1985) 6_1, 188. 120. Lauterbur, P.C., Levin, D.N., Marr, R.B. J. Magn. Reson. (1984) 59, 536. 121. Axel, L., Glover, G., Pelc, N. Magn. Reson. Med. (1985) 2, 428. 122. Dixon, W.T. Radiology (1984) 153. 189. 123. Sepponen, R.E., Sipponen, J.T., Tanttu, J.I. J. Comput. Assist. Tomogr. (1984) 8, 585. 124. Hall, L.D., Sukumar, S., Talagala, S.L. J. Magn. Reson. (1984) 56, 275. 125. Haase, A . , Frahm, J . , Hanicke, W., Matthaei, D. Phys. Med. Biol. (1985) 30, 341. 126. Bottomley, P A . , Foster, T.H., Leue, W.M. Proc. Natl. Acad. Sci. USA. (1984) 8_1, 6856. 127. Rosen, B.R., Wedeen, V.J., Brady, T.J. J. Comput. Assist. Tomogr. (1984) 8, 813. 128. Martin, J.F., Wade, C.G. J. Magn. Reson. (1985) 61, 153. 129. Haase, A. , Frahm, J . J. Magn. Reson. (1985) 64, 94. 130. Vo id , R.L., Waugh, J.S., Klein, M.P.,Phelps, D.E. J. Chem. Phys. (1968) 48, 3831. 131. Johnson, G., Hutchison, J .M.S. , Redpath, T., Eastwood, L.M. J. Magn. Reson. (1983) 54, 374. 132. Mansf ie ld , P., Pykett, I.L. J. Magn. Reson. (1978) 29, 355. 133. Haase, A. , Frahm, J . , Matthaei, D., Hanicke, W., Merboldt, K.D. J. Magn. Reson. (1986) 67, 217. 134. Frahm, J . , Haase, A. , Matthaei, D. Magn. Reson. Med. (1986) 3, 321. 135. Hricak, H., Wi l l iams, R.D., Moon Jr., K.L., M o s s , A A . , A lpers , C., Crooks, L.E., Kaufman, L. Radiolgy (1983) 147. 765. 136. Dwyer, A . J . , Knop, R.H., Hoult, D.i. J. Comput. Assist. Tomogr. (1985) 9, 16. 137. Babcock, E.E., Brateman, L., Weinreb, J.C., Horner, S.D., Nunnally, R.L. J. Comput. Assist. Tomogr. (1985) 9, 252. 138. Horman, I. Analysis of Food and Beverages: Modern Techniques, Charalambous, G. Ed., Academic Press: Flor ida, 1984. 206 139. Schultz, H.D. Ed., Coal Liquifaction Products, Volume 1, NMR Spectroscopic Characterization and Product Processes, Wi ley : New York, 1983. 140. Shulman, R.G. Ed., Biological Applications of Magnetic Resonance, Academic Press: New York, 1979. 141. Cohen, J .S . Ed., Magnetic Resonance in Biology, Wi ley & Sons : New York, 1980, Volume 1. 142. Mansf ie ld , P., Maudsley, A . A . Phys. Med. Biol. (1976) 2J., 847. 143. Lauterbur, P.C. NMR in Biology, Dwek, R.A., Campbell , I.D., Richards, R.E., Wi l l iams, R.J.P. Eds., Academic Press: New York, 1977, p 323. 144. Mansf ie ld , P., Maudsley, A .A. , Baines, T. J. Phys. E: Sci. Instrum. (1976) 9, 271. 145. Suryan, G, Proc. Indian Natl. Sci. Acad. Part A. (1951) 15, 53. 146. Singer, J.R. Science (1959) 130. 1652. 147. Kumar, J . , Kumar, V. Science (1972) V75, 794. 148. Stejskal , E.O. J. Chem. Phys. (1965) 43, 3597. 149. Hayward, R.J., Packer, K.J., Tomunson, D. J. Mol. Phys. (1972) 23, 1083. 150. Grover, T., Singer, J.R. J. Appl. Phys. (1971) 42, 938. 151. Watanabe, N., Niki , E. Proc. Jpn. Acad. (1978) 54, 194. 152. Dorn, H.C. Anal. Chem. (1984) 56, 747 A. , and references therein. 153. Haw, J.F., G lass , T.E., Hausler, D.W., Mote l l , E., Dorn, H.C. Anal. Chem. (1980) 52, 1135. 154. Haw, J.F., G lass , T.E., Dorn, H.C. ibid. (1981) 53, 2327. 155. Haw, J.F., Glass , T.E., Dorn, H.C. ibid. (1983) 55, 22. 156. Hall , L.D., Rajanayagam, V. J. Chem. Soc. Chem. Commun. (1985) 499. 157. Porath, J . , Car lsson, J . , O lsson, I., Belfrage, G. Nature (1975) 258. 598. 158. Chelating Sepharose 6B, Pharmacia Fine Chemicals Brochure, Uppsala, 1982., and references therein. 159. Walters, R.R. Anal. Chem. (1985) 57, 1099 A . 160. Hubert, P., Porath, J . J. Chromatogr. (1980) 198. 247. 161. Merboldt, K.D., Hanicke, W., Frahm, J . J. Magn. Reson. (1986) 67, 336., and references therein. 162. Schumacher, M.M. Ed., Enhanced Oil Recovery: Secondary and Tertiary Methods, Noyes Data Corporation, New Jersey, 1978. 163. Col l ins , A .G . Oil Field Subsurface Injection of Water, Wright, C .C , Cross, D., Ostroff , A .G. , Stanford, J.R. Eds., American Society for Testing and Materials, 1977, p 2., and references therein. 164. Dykstra, H., Parsons, R.L. The Prediction of Oil Recovery by Water Flood: Secondary Recovery of Oil in the United States, American Petroleum Institute, New York, 1950. 165. Gogarty, W.B. J. Pet. Tech. (1978) 30, 1089. 166. Shah, D.O., Schechter, R.S. Eds., Improved Oil Recovery by Surfactant and Polymer Flooding, Academic Press; New York, 1977. 167. S o o , H., Radke, C.J. J. Colloid. Interface Sci. (1984) 102, 462. 168. Scheidegger, A.E. The Physics of Flow through Porous Media, University of Toronto Press: Toronto, 1974. 169. M c A u l i f f e , C.D. J. Pet. Tech. (1973) 25, 727. 170. Hall , L.D., Rajanayagam, V., Hall , C. J. Magn. Reson. (1986) 68, 185. 171. Baldwin, B.A., Yamanashi, W.S., Lester, P.D. Magn. Reson. I mag. (1985) 3, 180. 172. Blackband, S., Mansfield, P., Barnes, J.R., Clague, A.D.H., Rice, S.A. J. Pet. Tech. (1986) 26, 101. 173. Maerefat, N.L., Palmer, I., Yamanashi, W.S., Lester, P.D. Magn. Reson. I mag. (1986) 4, 122. 174. Rothwell, W.P., Vinegar, H.J. Appl. Optics (1985) 24, 3969. 175. Vinegar, H.J. J. Pet. Tech. (1986) 38, 257. 176. Blackband, S., Mansfield, P. J. Phys. C: Solid State Phys. (1986) 19, L49. 177. Glasel, J.A. Nature (1970) 227. 704. 178. Hansen, J.R., Lawson, K.D. ibid. (1970) 225, 542. 179. Packer, K.J. J. Magn. Reson. (1973) 9, 438. 180. Finch, E.D., Homer, L.D. Biophys. J. (1974) 14, 907. 181. Glasel, J.A., Lee, K.H. J. Am. Chem. Soc. (1974) 96, 970. 182. Koeriig, S.H., Bryant, R.G., Hallenga, K., Jacob, G.S. Biochemistry (1978) 17, 4348. 183. Woessner, D.E. J. Magn. Reson. (1980) 39, 297. 184. Fung, B.M., McGaughy, T.W. J. Magn. Reson. (1981) 43, 316. 185. Hanus, F., Gillis, P. J. Magn. Reson. (1984) 59, 437. 186. Zimmerman, J.R., Brittin, W.E. J. Phys. Chem. (1957) CM, 1328. 187. Senturia, S.D., Robinson, J.D. Soc. Pet. Eng. J. (1970) 249. 237. 188. Loren, J.D., Robinson, J.D. ibid. (1970) 249, 268. 189. Timur, A . J. Pet. Tech. (1969) 246, 775. 190. Garroway, A.N., Grannell, P.K., Mansfield, P. J. Phys. Chem. (1974) 7 L457. 191. Mansfield, P., Maudsley, A A . J. Magn. Reson. (1977) 27, 101. 192. Singer, J.R. J. Phys. Eng. (1978) U, 281. 193. Panshin, A.J., de Zeeuw, C. Textbook of Wood Technology, McGraw-Hi l l : New York, 1980. 194. Core, H.A., C6t6, W A . , Day, A.C. Eds., Wood Structure and Identification, Syracuse University Press: New York, 1979. 195. Browning, B.L. The Chemistry of Wood, R.E.Krieger Publishing Co.: New York, 1975. 196. Swanson, T., Stejskal, E.O., Tarkow, H. Tappi (1962) 45, 929. 197. Stone, J.E., Scallan, A . M . ibid. (1967) 50, 496. 198. Pittman, R.A., Tripp, V.W. J. Po/ym. Sci. (1970) 8, 969. 199. Nanassy, A.J . Wood Sci. (1973) 5, 187. 200. Carles, J.E., Scallan, A . M . J. Appl. Po/ym. Sci. (1973) 17, 1855. 201. Nanassy, A.J . Wood Sci. (1974) 7, 61. 202. Nanassy, A. ibid. (1976) 9, 104. 203. Nanassy, A . ibid. (1978) I T , 86. 204. Sharp, A.R., Riggin, M.T., Kaiser, R., Schneider, M.H. Wood and Fiber (1978) 10, 74. 205. Riggin, M.T., Sharp, A.R., Kaiser, R., Schneider, M.H. J. Appl. Polym. Sci. (1979) 23, 3147. 206. Peemoeller, H., Hale, M.E., Schneider, M.H., Sharp, A.R., Kydon, D.W. Wood and Fiber Sci. (1985) 17, 110. 207. Kolodzlejski, W., Frye, J.S., Maciel, G.E. Anal. Chem. (1982) 54, 1419. 208. Dudley, R.L., Fyfe, C.A., Stephenson, P.J., Deslandes, Y., Hamer, G.K., Marchessault, R.H. J. Am. Chem. Soc. (1983) 105, 2469. 209. Haw, J.F., Maciel, G.E., Schroeder, H A . Anal. Chem. (1984) 56, 1323., and references therein. 210. Gray, R.L., Parham, R.A. Chemtech (1982) 12, 232. 211. Skaar, C. Water in Wood, Syracuse University Press: New York, 1972. 212. Part 1, Sections 8 - 9 in reference 57. 213. Szymani, R., McDonald, K.A. Forest Prod. J. (1981) 31, 34. 214. McMill in, C.W. Wood Sci. (1984) 14, 97. 215. McMill in, C.W., Conners, R.W., Huber, H.A. For. Prod. J. (1984) 34, 13. 216. Asplund, T., Johansson, L.G. Wood Technology Report, Swedish Forest Products Research Laboratory, April 1984, Nr 53, pp 30-35. 217. Burgess, A.E. Forintek Canada Corp.: Special Publication, (1984) 21., 5. 218. Funt, B.V., Bryant, E.C. Forintek Canada Corp.: Special Publ ication, (1984) 21, 16. 219. Taylor, F.W., Wagner Jr., F.G., McMill in, C.W., Morgan, IJ.., Hopkins, F.F.. For. Prod. J. (1984) 34, 42. 220. Hall, L.D., Rajanayagam, V., Stewart, W A . , Steiner, P.R. Can. J. For. Res. (1986) 16, 423. 221. Hall, L.D., Rajanayagam, V., Stewart, W.A., Steiner, P.R., Chow, S. Can. J. For. Res. (1986) 16, 684. 222. Hall, L.D., Rajanayagam, V. Wood Sci. Techno/. (1986) , in press. 223. page 57 in reference 212. 224. Wang, P.C., Chang, S.J. Wood Fiber Sci. (1986) 18, 308. 225. Shigo, A.L. Sci. Am. (1985) 252, 96. 226. page 353 in reference 194. 227. Basham, J.T. Can. J. Bot. (1958) 36, 491. 228. Rayner, A.D.M., Todd, N.K. Adv. Bot. Res. (1979) 7, 343. 229. Part 1, section 10 in reference 194. 230. Boyce, J.S. Forest Pathology, McGraw-Hi l l : New York, 1961. 231. Liese, W. Ed. Biological Transformations of Wood by Microorganisms, Springer-Verlag: Berlin/Heidelberg, 1975. 232. page 371 of reference 194. 233. page 381 of reference 194. 234. Hailey, J.R.T., Menon, R.S., Mackay, A., Burgess, A.E., Swanson, J.S. Fifth Symposium on Non-Destructive Testing of Wood, Pullman, Washington, September 1985. 235. Wood, A.D. Plywoods of the World, Morrison and Gibb Ltd.: London, 1963. 236. Perkins, N.S. Plywood, DFPA: Tacoma, 1962. 237. NMC - 1280 Manual, Nicolet Magnetics Corporation, Fremont, California, June 1982, page 155. 238. Conover, W.W. Relaxation Times (1981) 2, 8. 239. Levitt, M.H., Freeman, R. J. Magn. Reson. (1981) 43, 65. APPENDIX 1 212 Program to sum/project incomplete 2D data sets /MODIFICATION OF "PR" COMMAND 120460 •120460 120460 326322 "V"R /CREATION OF "VR" COMMAND 120461 130000 MODPR 130000 *130000 /STARTING ADDRESS 130000 0 MODPR, 0 130001 1217022 MEHA NUINS1 /GET NEU INSTRUCTION 130002 2413023 ACCrl 6ADDR1 /STORE IN LOCATION 34073 130003 1217024 MEHA NUINS2 /GET NEU INSTRUCTION 130004 2413025 ACCrl 8ADDR2 /STORE IN LOCATION 34071 130005 1010030 JMP gPFRCMD /DO PR COMMAND AT LOCATION 34020 130006 3000021 ENTER, JMS BPUBKBUF /ENTRY TO MODIFIED REGION 130007 130035 MSBPR /"[BLOCKS TO PROJECT" 130010 3000043 JMS 0PFIXIN 130011 130031 NOBLKS 130012 2413034 ACCM e P P R F C N /STORE « IN PRFCNT 130013 1217026 MEMA 0LINS1 /RESTORE OLD I N S T R U C T I O N 130014 2413023 ACCM BABDR1 /AT LOCATION 34070 130015 1217027 MEMA 0LINS2 /RESTORE OLD INSTRUCTION 130016 2413025 ACCM PAHHR2 /AT LOCATION 34071 130017 1217032 MEMA PEXIT /STORE 2162 AT LOCATION 34017 130020 2413033 ACCM 8PEXAB /TO GET OUT OF PR 130021 1010023 JMP BADDR1 /CONTINUE PR CMD AT LOCATION 34070 130022 1014071 NWINS1. 1014071 130023 34070 ADDR1 , 34070 130024 130006 NUINS2, 130006 130025 34071 ADHR2, 34071 130026 2012326 0LINS1, 2012326 130027 1217006 0LINS2, 1217006 130030 34020 P P R C M D , 34020 130031 0 NOBLKS, 0 130032 2162 PEXIT, 2162 130»33 34017 PEXAD. 34017 130034 34014 PPRFCN, PRFCNT 34014 PRFCNT = 34014 120043 PFIXIN =120043 120021 PUBKBUF =120021 130035 330214 MSBPR, .TEXT "[BLOCKS TO PROJECT" 130036 170313 130037 234024 130040 174020 130041 221712 130042 50324 130043 370000 APPENDIX 2 NOMENCLATURE Two-Dimens ional Imaging : A technique by which an image is displayed in two frequency dimensions. It involves the acquisition of a data matrix as a function of two time variables and when subjected to Fourier transformation yields two orthogonal frequency axes each of which can either represent spatial or a combined spatial and spectroscopic (chemical shift) information. Evolution time : The time after the initial 90° rf excitation pulse during which the spins are made to evolve, for example, under the influence of magnetic f ie ld gradients. These gradients phase encode spatial information along their respective axes. Flip angle : This is the amount of rotation of the macroscopic magnetization produced by the applied rf pulse, with respect to the B 0 f ie ld . Dwell time : The time computer spends in each of its memory locations while sampling a time domain signal . Magnetic Field Gradients : These are capable of producing a magnetic f ield that changes in strength along a certain given direction. The magnitude of these are measured in G/cm. Sl ice Selection : This technique enables the selection of planes within the object and is usually achieved by a combination of magnetic f ield gradient and rf pulses. Contrast : This is the perceived image intensity differences between adjacent regions and it arises because of the nature of the NMR signal which depends on three parameters p, J1 and T2. Phantom : An object of known dimensions and properties and is used to test and evaluate imaging systems. Chemical Shift Imaging A technique usually by which spatial information corresponding to individual chemical species is obtained. Here the chemical shift information is incorporated as another dimension in the image data matrix. In some instances, depending on the imaging sequence used this information occurs together with spatial information along the same dimension. Apodizat ion : A computer manipulation technique by which the time domain signal is modif ied to enhance certain features such as resolution enhancement or S/N improvement. Usually trigonometric functions are multiplied with the time domain signal prior to Fourier transformation. S ine -be l l function : This is an apodization function which consists of the first half cycle of a sine function and is frequently used to enhance the symmetry of a spin echo s igna l - in the time domain. Trapezoidal function : This is an apodization routine which consists of a trapezoidal function. The rising and falling portions of the trapezoid can be altered to suit the desired needs. Ze ro - f i l l i ng : This is a computer manipulation technique by which the number of data points in the current block is doubled with an equal number of data points by addition of zeros. Magnitude calculation : This computer routine replaces the real part of the Fourier transform with a square root of the sum of the squares of the real and imaginary parts. Transposit ion : A n operation which interchanges the rows and columns of a data matrix. For example, a data set S^^tj ) becomes S ^ . t j ) on transposit ion. Spatial resolution : This is the theoretically calculated smallest element of the object f rom which data has been sampled. 215 Fourier Zeugmatography : This is an experiment in which magnetic f ield gradients are applied for incremented periods of the evolution time so as to build up a mult i -d imensional data set of the form S(ti,t 2,t 3). Spin Warp Imaging : Instead of applying fixed gradients for varying time periods as in the Fourier Zeugmatography experiment, varying amplitude gradients are applied for a fixed period of time to build up the image data matrix. Surface Coi l NMR : A simple flat rf receiver coi l placed over a region of interest wi l l have an effect ive selectivity for a volume approximately subtended by the coi l circumference and one radius deep from the coi l center. Such a coi l can be used for simple localization of sites for measurement of high resolution NMR spectra i n - v i v o . Pixel : Acronym for a picture element; the smallest discrete part of a digital image display. Phase : In a periodic function (such as rotational or sinusoidal motion), the posit ion relative to a particular part of the cycle . Homogeneity : Uniformity . In NMR, the homogeneity of the static magnetic f ield is an important criterion of the quality of the magnet. Homogeneity requirements for NMR imaging are generally lower than those for NMR spectroscopy, but for most imaging techniques the homogeneity must be maintained over a larger region. Helmholtz Coil : A pair of current carrying coi ls used to create uniform magnetic f ield in the space between them. Resonator Coil : A rf probe made of thin sheets of copper and in this particular study, the probes were inductively coupled to the transmitter. The axial and transverse resonators mean the orientation of the probes within the magnet; former along the magnet axis and the latter across the axis respectively. Pseudo time domain : In Spin Warp imaging, since the information in the second time domain is obtained by modulating the magnitude of the gradients rather than the by real time increments, this domain is referred to as the pseudo time domain. Observe dimension : In an NMR experiment this represents the acquisition dimension during which the signal is observed or read by the computer. Usually in a 2D data matrix, S ^ . t j ) , it is coventional to say that t 2 is the observe dimension. Phase encoding dimension : In a 2D NMR experiment the second time domain during which information (spatial or spectroscopic) is encoded as phase differences in the observed NMR signal , is called the phase encoding dimension. Two-d imensional x,8 imaging sequence : A Chemical Shift imaging experiment in which the observe dimension contains chemical shift information (6) and the phase encoding dimension contains spatial information (x). Two-d imens ional y , ( x $) imaging sequence : A Chemical Shift imaging experiment in which the observe dimension contains, both chemical shift (5) and spatial information (x) and the phase encoding dimension contains only spatial information (y). Three-dimensional x,y,5 imaging sequence : A Chemical Shift imaging experiment in which the observe dimension contains chemical shift information (6) and the two phase encoding dimensions contain spatial information respectively. The images are usually displayed as 2D spatial maps corresponding to each chemically shifted resonance. Four -dimensional x,y,z,8 imaging sequence : A Chemical Shift imaging experiment in which the observe dimension contains chemical shift information (6) and the three phase encoding dimensions contain spatial information respectively. The images of different chemical species are usually displayed as 2D spatial maps corresponding to a particular spatial sl ice along the third spatial dimension. Two-d imens iona l x,y imaging sequence : This experiment generates a 2D spatial map for a single resonance. Both the observe and the phase encoding dimension represent spatial information. Three-dimensional x,y,z imaging sequence : This experiment generates a 3D data set in which the observed and the two phase encoding dimension represent spatial information. The images are displayed as 2D spatial maps corresponding to a sl ice along the observe dimension. invers ion-recovery sequence : This experiment refers to the inversion recovery sequence used for sp in - lat t ice relaxation time measurements. The initial 180° rf pulse inverts the magnetization which is then allowed to relax longitudinally. The magnetization is sampled later by a 90° rf pulse at different time intervals. This sequence can be combined with most imaging sequences to generate Tj contrasted images. PUBLICATIONS 1. Adaptation of High-Resolution NMR Spectrometers for Chemical Microscopy: Evaluation of Gradient Magnitudes and B] Homogeneity. Laurance D. Hall, Vasanthan Rajanayagam and Subramaniam Sukumar, L. Maen. Reson.. 60, 199-204 (1984). 2. Chemical-Shift-Resolved Tomography Using Four-Dimensional FT Imaging. Laurance D. Hall, Vasanthan Rajanayagam and Subramaniam Sukumar, J, Maen. Reson.. 61, 188-191 (1985). 3. Visualization of Chromatography Columns by NMR Imaging. Laurance D. Hall and Vasanthan Rajanayagam, J. Chem. Soc. Chem. Commun.. 499-501 (1985). 4. Construction of a High resolution NMR Probe for Imaging with Sub-millimeter Spatial Resolution. Laurance D. Hall, Stanley Luck and Vasanthan Rajanayagam, J. Maen. Reson.. 66, 349-351 (1986). 5. Chemical Shift Imaging of Water and n-Dodecane in Sedimentary Rocks. Laurance D. Hall, Vasanthan Rajanayagam and Christopher Hall. L Maen. Reson.. 68, 185-188 (1986). 6. Magnetic Resonance Imaging of Wood. Laurance D. Hall, Vasanthan Rajanayagam, Wendy A. Stewart and Paul R. Steiner, Can. J. For. Res.. 16, 423-426 (1986). 7. Detection of Hidden Morphology of Wood by Magnetic Resonance Imaging, Laurance D. Hall, Vasanthan Rajanayagam, Wendy A. Stewart, Paul R. Steiner and Suezone Chow. Can. J. For. Res.. 16, 684-687 (1986). 8. Evaluation of the Distribution of Water in Wood by Use of Three Dimensional Proton NMR Volume Imaging. Laurance D. Hall and Vasanthan Rajanayagam, Wood Sci. Technol.. 2Q 3 2 3 - 3 3 3