UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Dose verification of a stereotactic IMRT treatment planning system Xian, Zheng 2010

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
[if-you-see-this-DO-NOT-CLICK]
ubc_2010_spring_xian_zheng.pdf [ 5.82MB ]
Metadata
JSON: 1.0069932.json
JSON-LD: 1.0069932+ld.json
RDF/XML (Pretty): 1.0069932.xml
RDF/JSON: 1.0069932+rdf.json
Turtle: 1.0069932+rdf-turtle.txt
N-Triples: 1.0069932+rdf-ntriples.txt
Original Record: 1.0069932 +original-record.json
Full Text
1.0069932.txt
Citation
1.0069932.ris

Full Text

Dose Veri cation of aStereotactic IMRT TreatmentPlanning SystembyZheng XianB.Sc., Peking University, 2005A THESIS SUBMITTED IN PARTIAL FULFILMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMaster of ScienceinThe Faculty of Graduate Studies(Physics)The University Of British Columbia(Vancouver)April, 2010© Zheng Xian 2010AbstractIn this project, ion chamber measurement and  lm dosimetry were used toverify dose distributions for a new stereotactic IMRT (Intensity ModulatedRadiation Therapy) treatment planning system. This technique combinesthe principles of stereotactic radiosurgery and IMRT to signi cantly increasethe positioning accuracy compared with conventional IMRT . Ion chambermeasurements reveal that the discrepancy between the measured and thecalculated dose at the isocenter can be up to 2 %. Angular dependence ofion chamber sensitivity and the tissue equivalence of the phantom materialwere determined to be the main sources of this discrepancy. Radiochromic lm was used as the  lm dosimeter in the project. A set of performancetests of Gafchromic EBT  lm indicated that the uncertainty in GafchromicEBT  lm dosimety was expected to be 2.5 %. However, the discrepancies wefound in measurements of clinical cases using the  lm were much larger thanthis. And further investigation into this discrepancy was beyond the scopeof this thesis.iiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . xi1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62.1 Introduction to Radiation Therapy Physics . . . . . . . . . . 62.1.1 Radiation Interactions with Matter . . . . . . . . . . . 62.1.2 Radiation Biology . . . . . . . . . . . . . . . . . . . . 122.1.3 General Principles of Radiation Therapy . . . . . . . . 132.2 Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.2.1 Linear Accelerator . . . . . . . . . . . . . . . . . . . . 142.2.2 Multi-leaf Collimator . . . . . . . . . . . . . . . . . . 172.3 Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182.3.1 SRT and SRS . . . . . . . . . . . . . . . . . . . . . . . 182.3.2 Intensity Modulated Radiation Therapy . . . . . . . . 192.4 Treatment Plan Veri cation . . . . . . . . . . . . . . . . . . . 22iii3 Film Dosimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . 243.1 Introduction to Film Dosimetry . . . . . . . . . . . . . . . . . 243.1.1 General Considerations when Using Film . . . . . . . 263.1.2 Introduction to Radiochromic Film . . . . . . . . . . . 283.2 Film Scanning and Calibration . . . . . . . . . . . . . . . . . 303.2.1 Scanners . . . . . . . . . . . . . . . . . . . . . . . . . 303.2.2 General settings . . . . . . . . . . . . . . . . . . . . . 343.2.3 Position and Orientation Dependence . . . . . . . . . 363.2.4 Calibration . . . . . . . . . . . . . . . . . . . . . . . . 413.3 Performance of Gafchromic EBT Film . . . . . . . . . . . . . 533.3.1 Post-exposure Development . . . . . . . . . . . . . . . 533.3.2 Uniformity . . . . . . . . . . . . . . . . . . . . . . . . 563.3.3 Dose Rate Dependence . . . . . . . . . . . . . . . . . 583.3.4 Overall Accuracy . . . . . . . . . . . . . . . . . . . . . 594 Treatment Planning and Dose Veri cation . . . . . . . . . . 634.1 iPlan Treatment Planning Software . . . . . . . . . . . . . . . 644.1.1 PatXfer . . . . . . . . . . . . . . . . . . . . . . . . . . 644.1.2 iPlan Image . . . . . . . . . . . . . . . . . . . . . . . . 644.1.3 iPlan Dose . . . . . . . . . . . . . . . . . . . . . . . . 674.1.4 Phantom Mapping . . . . . . . . . . . . . . . . . . . . 714.2 Stereotactic IMRT Delivery System . . . . . . . . . . . . . . 734.2.1 Localization and Immobilization . . . . . . . . . . . . 734.2.2 BrainLAB m3 Micro MLC . . . . . . . . . . . . . . . 754.3 Dose Veri cation . . . . . . . . . . . . . . . . . . . . . . . . . 784.3.1 Phantom . . . . . . . . . . . . . . . . . . . . . . . . . 784.3.2 MLC performance Tests . . . . . . . . . . . . . . . . . 824.3.3 Ion Chamber Measurement . . . . . . . . . . . . . . . 875 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . 945.1 MLC Performance Tests . . . . . . . . . . . . . . . . . . . . . 94iv5.1.1 Dynamic MLC Dynalog Statistical Analysis . . . . . . 945.1.2 E ective Leaf Gap Tests . . . . . . . . . . . . . . . . . 975.2 Ion Chamber Measurement . . . . . . . . . . . . . . . . . . . 995.2.1 Calibration . . . . . . . . . . . . . . . . . . . . . . . . 995.2.2 Veri cation of IMRT Cases . . . . . . . . . . . . . . . 1015.3 Film Dosimetry . . . . . . . . . . . . . . . . . . . . . . . . . . 1095.3.1 Case One . . . . . . . . . . . . . . . . . . . . . . . . . 1095.3.2 Case Two . . . . . . . . . . . . . . . . . . . . . . . . . 1205.3.3 Case Three . . . . . . . . . . . . . . . . . . . . . . . . 1206 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1287 Future Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1317.1 Gafchromic EBT2 Film . . . . . . . . . . . . . . . . . . . . . 1317.2 Monte Carlo Simulation . . . . . . . . . . . . . . . . . . . . . 131Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135AppendicesA Source Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140A.1 Matlab Source Code Used to Adjust the Resolution of scannedimages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140A.2 Matlab Source Code Used to Calibrate the Film and Convertthe Dose Distribution into Omni-pro Compatible Format . . . 141vList of Tables3.1 Dose range and saturation level of di erent types of  lm . . . 273.2 Atomic composition and e ective atomic number of GafchromicEBT  lm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303.3 Monitor units and calculated dose for calibration exposures . . 473.4 Results of the overall accuracy test . . . . . . . . . . . . . . . 603.5 Inter-piece variation of  lm calibration . . . . . . . . . . . . . 624.1 Speci cations of BrainLAB m3 micro multi-leaf collimator . . 764.2 Results of the statistic MLC position test . . . . . . . . . . . . 834.3 Dynalog  le format . . . . . . . . . . . . . . . . . . . . . . . . 854.4 Results of the ion chamber angular dependence test (4.2 cm eld size) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 904.5 Results of the ion chamber angular dependence test (1.2 cm eld size) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 915.1 Statistical analysis of the Dynalog  le (Bank A) . . . . . . . . 955.2 Statistical analysis of the Dynalog  le (Bank B) . . . . . . . . 965.3 Dynamic leaf shift measurements . . . . . . . . . . . . . . . . 985.4 Ion chamber cross calibration data measured in solid waterphantom (Sept. 15 2009) . . . . . . . . . . . . . . . . . . . . . 995.5 Ion chamber cross calibration data measured in cube phantom(Sept. 15 2009) . . . . . . . . . . . . . . . . . . . . . . . . . . 1005.6 Ion chamber measurement of stereotactic IMRT plan (CaseOne) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102vi5.7 Dose comparison for individual  elds (Case One) . . . . . . . 1045.8 Ion chamber measurement of stereotactic IMRT plan (CaseTwo) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1055.9 Ion chamber measurement of stereotactic IMRT plan (CaseThree) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108viiList of Figures2.1 The linac gantry head . . . . . . . . . . . . . . . . . . . . . . 152.2 A linac System . . . . . . . . . . . . . . . . . . . . . . . . . . 163.1 AGFA Duoscan® scanner . . . . . . . . . . . . . . . . . . . . 333.2 Scanning resolution issue . . . . . . . . . . . . . . . . . . . . . 353.3 Pro les showing the scattering e ect . . . . . . . . . . . . . . 373.4 Image showing the scattering e ect . . . . . . . . . . . . . . . 383.5 Schematic diagram of the  lm positioning frame . . . . . . . . 403.6 Pro les across the same line of the same piece of  lm (markedc3) scanned in portrait and landscape mode . . . . . . . . . . 423.7 Two cube phantoms used in this project . . . . . . . . . . . . 453.8 Internal struction of  lm cube phantom . . . . . . . . . . . . . 463.9 Pixel value calibration curve . . . . . . . . . . . . . . . . . . . 493.10 Net optical density calibration curve . . . . . . . . . . . . . . 503.11 Dose distribution calculated using pixel value . . . . . . . . . 513.12 Discrepancy between pixel value calibration and net opticaldensity calibration . . . . . . . . . . . . . . . . . . . . . . . . 523.13 Post-exposure development experiment data . . . . . . . . . . 543.14 Consecutive scans of the same  lm . . . . . . . . . . . . . . . 553.15 Longitudinal pro les of consecutive scans of the same piece of lm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573.16 The region of interest . . . . . . . . . . . . . . . . . . . . . . . 603.17 Three calibration curves scanned and generated in one session 61viii4.1 The  ow chart of iPlan . . . . . . . . . . . . . . . . . . . . . . 654.2 Reference box detection . . . . . . . . . . . . . . . . . . . . . 664.3 Screen capture of the Boolean operation interface. . . . . . . . 684.4 Screen capture of the iPlan image after the objects are contoured. 694.5 Di erent constraint points on the DVH a ecting the 90 % dosecoverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 724.6 Thermoplastic mask . . . . . . . . . . . . . . . . . . . . . . . 744.7 BranLAB m3 micro multi-leaf collimator . . . . . . . . . . . . 774.8 The ion chamber cube phantom and inserts . . . . . . . . . . 784.9 Internal structure of the  lm cube phantom . . . . . . . . . . 804.10 Stereotactic head stage used to hold the phantom . . . . . . . 814.11 Static leaf position QA test results . . . . . . . . . . . . . . . 844.12 0.01 cc mini chamber . . . . . . . . . . . . . . . . . . . . . . . 884.13 Schematic diagram of the cylindrical phantom. . . . . . . . . . 894.14 Schematic diagram showing the convention used in the de ni-tion of chamber angle . . . . . . . . . . . . . . . . . . . . . . . 904.15 Plot of the ion chamber angular dependence (4.2 cm) . . . . . 924.16 Plot of the ion chamber angular dependence (1.2 cm) . . . . . 935.1 Dose pro le in L-R direction of Case One . . . . . . . . . . . . 1055.2 Dose pro le in F-H direction of Case One . . . . . . . . . . . . 1065.3 Dose pro le in A-P direction of Case One . . . . . . . . . . . . 1075.4 The calculated dose distribution of Case One (Coronal plane) 1105.5 The isodose line comparison of Case One (Coronal plane) . . . 1115.6 The y-pro le comparison of Case One (Coronal plane) . . . . . 1125.7 The 2D dose di erence map of Case One (Coronal plane) . . . 1135.8 The 2D Gamma test with 3 mm and 3 % criteria of Case One(Coronal plane) . . . . . . . . . . . . . . . . . . . . . . . . . . 1145.9 The calculated dose distribution of Case Two (Coronal plane) 1155.10 The isodose line comparison of Case Two (Coronal plane) . . . 1165.11 The y-pro le comparison of Case Two (Coronal plane) . . . . 117ix5.12 The 2D dose di erence map of Case Two (Coronal plan) . . . 1185.13 The 2D Gamma test with 3 mm and 3 % criteria of Case Two(Coronal plane) . . . . . . . . . . . . . . . . . . . . . . . . . . 1195.14 The calculated dose distribution of Case Three (Coronal plan) 1215.15 The isodose line comparison of Case Three (Coronal plane) . . 1225.16 The y-pro le comparison of Case Three (Coronal plane) . . . . 1235.17 The 2D dose di erence map of Case Three (Coronal plan) . . 1245.18 The 2D GAMMA test with 3 mm and 3 % criteria of CaseThree (Coronal plane) . . . . . . . . . . . . . . . . . . . . . . 125xAcknowledgementsThis project was performed at the Vancouver Cancer Center. I would like toexpress my gratitude to Dr. Ermias Gete and Dr. Cheryl Duzenli for theirsupervision, direction support and advice. Also thanks to all the people inthe VCC physics department for the help provided to operate treatment unitand in treatment planning.xiChapter 1IntroductionRadiation therapy or radiotherapy, is the medical application of radiationmost commonly as part of cancer treatment. External beam radiotherapyis the most frequently used form of radiotherapy, during which, the patientusually lies on a couch and the radiation beam coming from the therapeuticequipment is directed to the patient’s body to kill the malignant cells withinthe target volume. A more detailed introduction to the physics of radiationtherapy is given in Chapter 2, Section 2.1.1. The radiation beam causes dam-age to both malignant cells and healthy tissue, therefore, di erent treatmenttechniques are used to con ne the dose (which is the energy per unit massdeposited by the radiation beam) to the target volume. From the most sim-ple square  eld irradiation to more complex techniques such as RapidArc™[1],radiotherapy techniques have improved along with advances in equipment toincrease the accuracy,  exibility and e ciency of beam delivery. Intensitymodulated radiotherapy or IMRT for short, is one of the modern radiother-apy techniques that can dramatically improve the dose coverage and normaltissue sparing. This technique not only con nes the beam outline but iscapable of modulating the beam intensity across the plane perpendicular tothe beam axis. By projecting multiple beams from di erent angles, threedimensional dose modulation can be achieved. This technique is particularlyuseful and necessary in some situations such as when the target volume isclose to critical organs that must be protected, or the shape of the targetvolume is highly irregular. The hardware that is used to achieve the modula-tion is called a multi-leaf collimator or MLC, which is nowadays a standardbuild-in part of most medical accelerators. The leaves of the MLC travel dy-1namically while the beam is being delivered, blocking di erent parts of the eld at di erent moments. The relation between the leaf sequence and dosedistribution is complicated. (Further discussions about IMRT are found in2.3.2.)Quality control is important for radiation therapy. The ultimate goalof quality control is to ensure the desired dose distribution is delivered asplanned. Quality control for radiation therapy includes: commissioning ofnew equipment, periodic quality assurance (QA) on hardware and softwareand dose veri cation of individual patients’ treatment plans. Dose veri ca-tion is an essential part of both commissioning and patient plan dose veri -cation. For IMRT the dose calculation depends on accurate modeling of theMLC leaves. Most MLC leaves have specially designed rounded leaf tips tofollow the beam divergent line for di erent  eld sizes. The MLC leaf crosssection is also not rectangular, but is designed in a tongue and groove style tominimize the inter-leaf leakage[2]. For most of the time during beam deliverythe MLC opening is small compared to the range of electrons set in motionby the photon beam, therefore electron disequilibrium may be signi cantand this is di cult for the treatment planning software to account for. Thedose delivery is also a ected by the mechanical performance of the MLC.Considering all these factors, the agreement between the dose distributioncalculated from the treatment planning system and the delivered result isusually worse than for simpler techniques. A thorough dose veri cation isthus required for both commissioning purposes and individual patient planveri cation to ensure that any discrepancies are within clinical tolerance.Ideally, a three dimensional dose measurement method would be the mostcomprehensive test, however, currently there is no such three dimensionalmeasurement technique that can meet the requirements of both spatial res-olution and accuracy. A gel dosimeter[3] and an accurate tomography tech-nique may be a solution but gel dosimeters are not routinely used due to anumber of practical issues. The most commonly used methods of IMRT dose2veri cation include ion chamber measurement,  lm dosimetry and MonteCarlo simulation[4]. An ion chamber measurement provides the most reli-able and accurate dose reading but cannot easily provide information aboutthe dose distribution, therefore it is usually used to acquire a single refer-ence point dose reading. Film dosimetry can be used to measure the twodimensional dose distributions. Unfortunately, the absolute dose reading for lm dosimetry is in general inaccurate, and is the result of several factorssuch as limitations of the  lm itself, the performance of the optical scannerand chemical processing. (Detailed discussion of  lm dosimetry is given inChapter 2.) Monte Carlo techniques can simulate particle interactions andtransport from the linac to and within the patient’s body and therefore cancalculate dose distributions accurately[5]. One option for veri cation of thetreatment planning dose calculation is to do an independent simulation ofthe dose using Monte Carlo techniques. Additionally, the performance of thetreatment unit can be incorporated into the simulation using the informationfrom the MLC log  le, which records the leaf position as a function of timeduring treatment delivery. This method is generally more accurate than  lmdosimetry. Also, a three dimensional dose distribution can be acquired. But,Monte Carlo simulation is not a measurement method, but a highly accuratecalculation method. Monte Carlo algorithms therefore need to be veri edby direct dose measurement prior to clinical use. As a result, a direct two-dimensional dose measurement is still necessary. (More discussions aboutdose measurement techniques can be found in Chapter 4.3.)BrainLAB AG (Feldkirchen, Germany) provides iPlan® RT treatmentplanning software which is a stereotactic radiotherapy dose planning soft-ware program capable of doing IMRT treatment planning. This software, isparticularly suitable for stereotactic planning. (Stereotactic radiotherapy isintroduced in Chapter 2, Section 2.3.1.) iPlan comes as part of the Brain-LAB stereotactic radiosurgery system and is designed to be compatible tothe m3® High-Resolution multi-leaf collimator. With the help of this sys-3tem the patient can bene t from both IMRT and stereotactic techniques,which is quite necessary for some di cult intra-cranial and head and neckcases where small irregularly shaped target volumes are very close to criticalstructures[6{10]. As mentioned in previous paragraphs, dose veri cation isan essential part of commissioning and QA processes. Dose veri cation forthe iPlan system has its unique challenges, primarily because the  eld sizeis smaller than for more conventional IMRT techniques. Detailed discussionabout this is given in Chapter 2, Section 2.4.The overall scope of this project is to evaluate a radiochromic  lm dosime-try system and use it together with the ion chamber measurement to achievedose veri cation for the iPlan stereotactic IMRT treatment planning system.Radiochromic  lm is a self- developing and light insensitive alternative to ra-diographic  lm. The Gafchromic® EBT  lm is specially designed for IMRTQA purposes. It does not have to be handled in a dark room thus can beeasily cut and loaded under normal room lighting conditions. The lack ofchemical processing not only reduces the post irradiation work load but alsoremoves one of the more problematic aspects of traditional  lm dosimetry.Gafchromic  lm may be used with a  atbed scanner, which is accessible anda ordable. Finally, the composition of radiochromic  lm is much closer towater or tissue and thus gives a more representative measure of dose in thesematerials compared with silver bromide radiographic  lm. For these ben-e ts, it is of great interest to replace radiographic  lm with radiochromic lm. Before radiochromic  lm can be clinically used, its performance has tobe tested and a proper procedure for using it as a dosimeter has to be estab-lished, because the performance also depends on how the  lm is scanned andanalyzed. Detailed discussions and performance tests of Gafchromic EBT lm are reported in Chapter 3.The EBT  lm technique is tested on three clinical cases following a com-missioning process. In Chapter 4, the treatment planning and delivery systemfor the stereotactic IMRT technique used for the three cases is explained in4detail. The last section of Chapter 4 describes in the methods used in doseveri cation and commissioning, including MLC performance tests , phan-tom structure, ion chamber measurement and the  lm dosimetry technique.Chapter 5 summarizes the experimental data and discussion. Final con-clusions are presented in Chapter 6. Future work is described in the lastchapter.5Chapter 2TheoryThis chapter serves to introduce some theoretical background to provide ageneral understanding of radiation therapy and some speci c radiotherapytechniques. The chapter is divided into three sections. In Section 2.1, phys-ical and biological e ects of ionizing radiation are discussed as well as thegeneral principles of radiation therapy. In Section 2.2, two important ra-diotherapy instruments, the linear accelerator and the multi-leaf collimatorare introduced. Stereotactic radiotherapy and intensity modulated radiationtherapy which are the major techniques used in this project, are discussed inSection 2.3.2.1 Introduction to Radiation TherapyPhysicsRadiation can cause biological damage to living tissue, including acute ra-diation illness at high doses, and cancer induction or genetic damage at lowdoses. Radiation is also widely used in health care for medical imaging andtumor treatment. In order to utilize radiation for therapeutic and diagnosticpurposes, it is important to understand how radiation interacts with livingtissue.2.1.1 Radiation Interactions with MatterIn general the term radiation refers to subatomic particles or photons that aredirected at a target. Typically, radiation can be categorized as non-ionizing6or ionizing. The di erence being whether the energy of the impinging indi-vidual particle is high enough to detach electrons from atoms or molecules.The ability of radiation to ionize depends on the energy, and the type of theindividual particle but not on the number of particles. A beam of intense lowenergy radiation, although possibly possessing a large amount of energy, maynot be able to cause any ionization. Di erent kinds of particles interact witha medium through di erent mechanisms. . Some particles interact with softtissue strongly, losing energy rapidly, thus have limited penetration depth.Some particles, on the other hand, can travel for a long distance withoutundergoing interaction. Detailed descriptions of how particles interact withmatter can be found in many publications[11, 12]. For the purpose of thisthesis, a 6 MV photon beam is used, thus only photon and electron interac-tions in this energy range will be discussed in the following paragraphs.Photons interact with matter primarily through four mechanisms in theenergy range from several hundred keV to several MeV, including Rayleighscattering, the photoelectric e ect, Compton scattering and pair production(or more precisely pair and triplet production). The probability of occurrenceof these interactions depends both on photon energy and atomic number ofthe medium.Rayleigh ScatteringRayleigh scattering, which is sometimes referred to as coherent scattering,occurs when a photon travels through a medium in which the particles aremuch smaller than the wavelength of the incoming photon. In Rayleighscattering, only the direction of incident photons changes. The photon energyis conserved and not transfered to the medium. The probability of occurrencefor Rayleigh scattering decreases with the incident photon energy, but itincreases with the atomic number of the medium. The atomic cross section7for Rayleigh scattering (a R) is approximately given by Equation 2.1,a R/Z2(h )2 (2.1)Rayleigh scattering has more practical importance in the low energy regime,partly because the probability is higher, partly because the scattering angleis greater. Rayleigh scattering does not contribute to dose, since no energyis transferred during this interaction. Rayleigh scattering is thus more im-portant in imaging application than in radiation therapy.The Photoelectric E ectThe photoelectric e ect is a phenomenon where the incident photon is ab-sorbed by the atom and an electron is emitted. The photoelectric e ect ismore dominant in the energy range of several eV to around 0.5 MeV. Inorder for the photoelectric e ect to occur, the incident photon energy hasto be higher than a threshold, which is the binding energy of the electron.The incident photon with enough energy interacts with the bound electronin an atom and gets absorbed. The energy of the photon is transferred to theelectron. Part of the energy is used to overcome the binding energy and freethe electron from the atom. The residual energy is transferred to the kineticenergy of the escaping electron. The probability of occurrence for the photo-electric e ect is roughly proportional to the third power of the reciprocal ofthe photon energy provided it is higher than the binding energy. Similar toRayleigh scattering the probability of occurrence is also strongly dependenton the atomic number of the medium. The atomic interaction cross section,a , is approximately given by Equation 2.2,a /Z4(h )3 (2.2)8Compton ScatteringCompton scattering is the dominant mode of interaction in the energy rangefrom several hundred keV to several MeV. When Compton scattering occurs,the incident photon transfers part of its energy to an electron and changesits direction. The electron which receives the energy and a certain amount ofmomentum is called a recoiling electron. In Compton scattering, the photonsonly interact with loosely bound electron such as the outer shell electrons orfree electrons. During the process, the overall energy and momentum of bothparticles are conserved, and are used to derive the relationship between thescattering angle and energy transferred. Compton scattering is the only typeof interaction that is not highly dependent on the Z of the medium. Theprobability of occurrence of Compton scattering only depends on incidentphoton energy, and decreases with increasing photon energy. The interac-tion cross section follows the Klein-Nishina function[11]. Since all other typesof interaction are highly Z-dependent and the probability of occurrence in-creases with respect of the atomic number, Compton e ect turns out to bemore important for media with low Z, such as carbon and hydrogen. Forsoft tissue, Compton scattering contributes most to the dose deposition formegavoltage photons[13].Pair ProductionPair production refers to the creation of an electron and a positron pair froma photon. In order for this interaction to occur, the photon energy should begreater than the rest energy of the electron-positron pair it is going to create.In order to conserve the momentum, pair production can only occur in thevicinity of the nucleus so that part of the momentum can be absorbed bythe recoiling nucleus. For photon energies close to the threshold, the createdelectron and positron travel almost in opposite directions to each other. Forenergy well above the threshold, the pair could travel more in the forwarddirection. The probability of occurrence of pair production increases rapidly9as the photon energy increases, and it is also strongly dependent on theatomic number. The interaction cross section per unit mass is approximatelyproportional to the atomic number. The positron created in a pair productionprocess slows down by interacting with the surrounding environment, andeventually annihilates with another slow electron creating a pair of 511 keVphotons.Summary of Photon InteractionsThe probability of occurrence for these interactions depends both on thephoton energy and the atomic number of the medium. This is why highatomic number materials have very di erent absorption characteristics com-pared with low atomic number materials. For media that have high atomicnumber, the photoelectric e ect and pair production have relatively higherinteraction cross sections compared to Compton scattering. Compton scat-tering thus is dominant only for a very narrow energy range. At low energy,since the photoelectric e ect is strong, high atomic number media absorb thephoton much more strongly than low atomic number media. For media thathave low atomic number, Compton scattering is the most dominant mode ofinteraction for wider spectrum of a few hundred keV to several MeV. Thismakes it of great importance for radiotherapy, since soft tissue is primar-ily made up of low atomic number media and the photon energy used inradiotherapy is just in that energy range.Electron InteractionsIn all the above interactions, secondary electrons and or positrons are cre-ated, and it is those electrons and or positrons that eventually deposit theenergy in the medium by causing ionization and excitation. Because elec-trons are charged particles, electron interactions di er greatly from photoninteractions. and are described using di erent terminology. For example, theattenuation of photons is logarithmic and the probability of interaction for10photons is often described by cross sections. Due to the Coulomb force, elec-trons are strongly in uenced by the surrounding environment and typicallyinteract and loose energy continuously as they travel through a medium andcome to rest. Stopping power, which is the amount of energy transferred tothe medium per unit thickness of travel, is used to characterize electrons.Electron interactions can be categorized into three classi cations: ioniza-tional, radiative and scattering interactions. In ionizational interactions, theelectron’s kinetic energy is transferred to an atom causing ionization or ex-citation. For non-relativistic electrons the energy transferred in ionizationalinteractions is inversely proportional to the kinetic energy of the incident elec-tron. For high energy electrons, since the velocities are closer to the speed oflight, relativistic e ects must be considered. The mass stopping power is rel-atively constant within the energy range from 0.1 MeV to 100 MeV. Duringradiative interactions, electrons accelerate rapidly within an atomic nucleus’electric  eld and part of the kinetic energy is transferred as a radiated photon(bremsstrahlung). In scattering interaction, the electron is de ected by thestrong Coulomb force without loss of its kinetic energy.Electron EquilibriumNotably, this chain of interactions does not happen locally, thus the dosedistribution is not proportional to the photon  ux at any given point in themedium. The exception is a condition called electronic equilibrium, where thesecondary electrons set in motion inside and outside a small volume elementreach a dynamic balance. This means the same number of electrons withsame energy distribution enter and exit this volume element. When electronicequilibrium is present, the energy deposition by the secondary electrons islocally dependent and is proportional to the photon  ux, which simpli esdose calculation and measurement in most cases. Electronic equilibriumdose not hold close to  eld and medium boundaries. For example, when abeam crosses a discontinuity in the medium or at the edge of a photon beam,11the number of electrons entering and exiting the volume is di erent. This iscalled electronic disequilibrium.2.1.2 Radiation BiologyRadiation causes cell damage in di erent ways. Ionization of DNA moleculescan break chemical bonds, causing the strand to break up. Moreover, ioniza-tion can create highly reactive radicals. Free radicals are highly reactive dueto the presence of unpaired valence shell electrons. They can cause damageto cell structure as well as DNA. Usually DNA damage can lead to threedi erent results. In some cases DNA damage can be repaired by DNA re-pair enzymes with the aid of the opposite strand. Sometimes, DNA damagecannot be repaired, and this leads to cell death. In this situation, the singlecell is lost preventing potential genetic damage from being transferred to thenext generation. In other cases, DNA damage is unable to be repaired butis also non-lethal. In these cases, the genetic mutations will be passed onto subsequent generations of cells. If the mutation happens to be critical,they can cause cancer or other illness. During each phase of the cell cycle,cells have di erent sensitivity to radiation due to the fact that DNA is beingreplicated in some phases and is not paired. Generally, if the cells are ac-tively undergoing cell division they are more vulnerable to radiation damage.Fractionation of dose delivery is bene cial because radiation from di erentsessions has more chance to catch all cancer cells in the most sensitive celldivision phase, thus resulting in better tumor control. Research also showsthat healthy tissue tends to resist more radiation damage compared withcancer cells if dose per fraction is kept below a certain level. For the sus-ceptibility of radiation on an organ level, the volume e ect has to be to takeinto account. Parallel organs, in which the functional units work in parallel,for example the liver, can tolerate high dose if a relatively small portion isirradiated. For serial organs, such as the spinal cord, on the other hand, thetolerance is almost independent of irradiated volume.122.1.3 General Principles of Radiation TherapyThe goal of cancer treatment is to control the malignant cells, either cura-tively or palliatively. This goal is the same for radiation therapy. Radiationtherapy uses ionizing radiation to treat cancer by killing the malignant cellsto achieve local tumor control. Radiation therapy can be used as a primarycurative treatment, or it can be used as an adjuvant treatment. As radiationcan damage both malignant cells and healthy cells, e ort has to be made toprotect the healthy surrounding tissue. Radiation therapy can be categorizedas: External Beam Radiotherapy, where the source of radiation is outside ofthe body; brachytherapy, where a sealed source is implanted into the tu-mor target and radioisotope therapy, where an unsealed radioactive sourceis given by infusion or oral ingestion. The subject of this thesis is related toexternal beam radiotherapy and further discussion will be restricted to thismethod.For external beam radiation therapy, because radiation from the externalsource has to travel through a patient’s body to reach the clinical target inmost cases, dose deposition along the path is inevitable. An external beamtreatment unit not only functions as a radiation source but most importantlyas a guidance system to direct and con ne the radiation damage to clinicaltargets. Usually the radiation is given from several di erent angles so thatunnecessary dose deposition for each  eld to any point in normal tissue isonly a small fraction of the tumor dose. The shape of each  eld can be ad-justed according to the projection of the tumor target(s). The applicationof medical linear accelerators and high tech accessories allows more sophis-ticated techniques to be realized, including Intensity Modulated RadiationTherapy (IMRT), Rapidarc™, Image Guided Radiation Therapy (IGRT) andStereotactic Radiosurgery (SRS). But the purpose of all these improvementsis simple, achieving better tumor control while sparing healthy tissue as muchas possible.132.2 Equipment2.2.1 Linear AcceleratorRadiation therapy has over one hundred years of history. Originally X-raytubes were used to treat super cial malignancies due to a limited penetrationdepth. Most deep target treatments used radioactive isotopes as the radiationsource, until the medical linear accelerator was invented in the 1960’s[14].A medical linear accelerator (Linac) uses a tuned-cavity waveguide, inwhich the radio frequency eletromagnetic  elds are used to accelerate elec-trons to high energy. Usually the wave guide is horizontally mounted so thata magnetic bending device is used to turn the electron beam vertically to-ward the patient. The direct output is a monoenergetic electron beam from4 MeV to 25 MeV. If a photon beam is needed, the electron beam is directedinto a high-density metal target (usually made of tungsten alloy) to create anx-ray beam with a continuous energy spectrum of maximum energy equal tothe electron energy. X-rays are generated by rapid slowing down of the elec-trons inside the target (Bremmstrahlung radiation). A collimation system islocated in the head of the gantry, including two pairs of jaws and usually anintegrated multi-leaf collimator (The integrated MLC can be seen on Figure2.1). The jaws are used to de ne the basic rectangular  eld size, while amulti-leaf collimator de nes the actual  eld shape.Figure 2.2 shows a photograph of a medical linac. Notice the gantry canrotate about a horizontal axis. The collimator can rotate about the centralbeam axis. The patient couch can rotate about a vertical axis. All thesethree axes intersect at one particular point named the isocenter which is inmost cases 100 cm away from the source. Other accessories include position-ing lasers, and immobilization devices that are used to increase positioningaccuracy.The  exibility, accuracy and safety of linacs are the most signi cant ad-vantages over the older generation of therapeutic isotope machines. The linac14Figure 2.1: This is a photograph of a linac gantry head. Inside the windowat the right edge, a line of tungsten leaves is visible. This is the integratedVarian Millennium MLC. The visible leaves are from a single bank and areall fully retracted. At the center of the window, there are two visible blockswhich are the second pair of the jaws.15Figure 2.2: This is a photograph of a linac system. Under the currently setupcondition, the couch angle is in zero position, as well as the gantry angle.The collimator angle is 90 °.16can simply be powered o when not in use. There is no permanent sourcerequiring heavy shielding and there is no need to worry about radiation leak-age when the machine is powered o . There are numerous built-in interlockspreventing accidental operation of the machine. System redundancy is highin monitoring the dose delivery to make sure no excess dose is given. TheLinac is the most frequently used machine for delivery of many advancedtechniques including IMRT[15].2.2.2 Multi-leaf CollimatorAn multi-leaf collimator (MLC) is a collimation system capable of shapinga radiation  eld in an arbitrary shape. MLCs consist of two opposing setsof tungsten leaves which travel across the  eld to shape and modulate theradiation beam. The position and speed of each leaf can be programmedindividually. The number of leaf pairs varies from model to model. Allleaves are parallel and move in a plane perpendicular to the beam axis.Ideally, in the vertical direction the MLC leaf edges follow a divergent linefrom the source to minimize the penumbra. In actual fact, MLC leaveshave round tips to make the size of penumbra independent of leaf position.Tongue and groove design is used to counter inter-leaf leakage. Initially,MLCs were used as static shaping accessories, replacing the hand-craftedlead block, reducing work load. Around the 1980’s, MLC was  rst used asa dynamic modulating device in IMRT delivery. The requirement for realtime response of MLCs has pushed manufacturers to improve the designof MLCs to make them faster and more accurate. Nowadays, MLCs havebecome an essential part of a dynamic delivery system capable of deliveringIMRT, dynamic arc and RapidArc™ radiation therapy. An MLC controllerrecords real time internal monitoring data, providing a way to check theMLC performance. Monte Carlo simulation based on this dynamic log  le isanother important application that is discussed later in this thesis.172.3 Techniques2.3.1 SRT and SRSStereotactic radiosurgery (SRS) is a technique where an extremely high doseis given to a small, highly con ned and precisely localized target usually lo-cated in the brain. The extremely high dose can be given in a single fraction,due to the fact that the targets are small and large volume of healthy tissuecan be avoided by using beams from di erent angles. The bene t of deliver-ing the dose in a single fraction is that it is easier to achieve precise targetlocalization in one session[16, 17]. If the dose is delivered in several fractionsthe technique is called stereotactic radiotherapy (SRT). The intra-cranial en-vironment is challenging for radiation therapy because of the existence ofso many critical structures. In order to spare as much healthy brain tis-sue as possible, a sharp dose gradient outside the  eld edge is demanded.Special immobilization and positioning devices are used. These localizationand positioning applications di erentiate SRT from conventional radiationtherapy.Unlike conventional conformal radiotherapy techniques, a  xed coordi-nate system for SRS is set up by using a stereotactic frame and localizationbox used throughout the planning and treatment processes. The stereotacticframe is also used as an immobilization frame. It is physically screwed intothe patient’s skull and attached to the treatment couch, so that the relativeposition of the patient to the treatment couch does not change during thewhole imaging and treatment process. The localization box can be mountedon the treatment couch and is located outside the patient’s head. Prior tothe treatment, a CT scan of the patient’s cranium is performed with thelocalization box on. Metal rod markers are implanted on three surfaces ofthe box, which can be clearly seen on CT image. The planning system au-tomatically recognizes these markers and sets up a  xed coordinate systembased on these  ducials. Due to the fact that most intracranial tumors that18are treated with SRS are relatively small, conformal treatments are typicallyused.[18] For fractionated SRT a less invasive immobilization mask is used.For larger treatment volumes, IMRT in combination with SRT may providetherapeutic advantages[6{10].2.3.2 Intensity Modulated Radiation TherapyIntensity modulated radiation therapy (IMRT), as the name indicates, is usedto modulate the intensity of incoming beams, adding  exibility (compared toconventional beams of uniform intensity) to achieve a higher degree of spatialconformity of dose distributions. Typically in IMRT, we reduce the intensityof rays that go through sensitive critical structures and increase the intensityof those rays that can primarily see the target volume. This modulation mayresult in highly non-uniform dose distribution associated with each individual eld within the target volume, compared to traditional conformal treatment.For a single  eld, the modulated dose distribution may show unwanted coldspots and hot spots, but this would be compensated for by other  elds. Theoverall e ect is an additional degree of freedom for dose manipulation. Auniversally accepted de nition of IMRT does not exist but an understandingagreed upon by most is that IMRT is a radiation treatment technique withmultiple beams in which at least some of the beams are intensity-modulatedand intentionally deliver a non-uniform intensity to the target. The desireddose distribution in the target is achieved after superimposing such beamsfrom di erent directions. Because the calculation is done in a discrete mannerand the MLC has limited spatial resultion, the modulation of each beamhas  nite resolution. The smallest division used in the optimization processis known as a beamlet. The additional degrees of freedom are utilized toachieve a better target dose conformality and/or better sparing of criticalstructures[19]. Generally, inverse planning techniques are used in IMRT.19History of Intensity Modulated Radiation TherapyThe earliest paper which is recognized as the cornerstone of IMRT devel-opment is published in 1982 by Brahme et al with the name \Solution ofan integral equation encountered in rotation therapy."[20] In this paper, theauthor posed the central question:\Which is the desired lateral dose pro lein the incident beam that produces a desired absorbed dose distribution inthe body after one complete rotation?" Though this paper is mathematicalin nature and the proposed model is rotational delivery rather than for static elds, it provides the most important conclusion:\a highly non-uniform in-tensity pro le is needed to produce a uniform dose distribution in the targetvolume."While the  rst theoretical inverse planning methods were developed in the1980s, it was not at all clear at that time how IMRT would be practicallydelivered[19]. But a dynamic treatment technique called dynamic wedge, inuse since 1978, gave an important indication of a possible solution. Duringthe delivery of dynamic wedge, one leaf bank of an MLC, or at that time asingle jaw, is stationary, while the opposing leaves move. The beam aperturebecomes increasingly narrower as the treatment progresses. The dwell timeof the jaw or leaves varies according to position, achieving a gradually de-creasing dose distribution across the  eld, similar to the result we get froma physical wedge. By extending this idea, if the motion is independentlymodulated for each leaf, i.e. the speed of closing in is variable rather than aconstant, an arbitrary intensity pro le can be acquired. After solving severalmechanical and quality assurance issues this idea of using MLC to deliverIMRT was quickly realized.Inverse Planning and Forward PlanningThe calculation of non-uniform intensity pro les for each individual  eldbegins with the dose prescription to the target volume and dose limits forsurrounding critical structures. Desired intensity pro les are derived by min-20imizing a cost function based on the dose volume constraints to both tumorand normal tissue. The process is known as ‘inverse planning’. Actual leafmotion sequences are calculated based on beam intensity pro les taking intoaccount physical MLC characteristics. The calculation of dose distributionsbased on MLC leaf sequences is called ‘forward calculation’. Improvement inIMRT inverse planning algorithms and dose calculation continue to be made.After some early attempts to apply inversely planned IMRT to certainsites such as lung, it was found that inverse planning based on simple doseobjectives and constraints did not yield satisfactory results. This has ledto or re-activated many discussions about the need for biologically-basedIMRT planning. However, quite satisfactory results have been achieved formany sites such as head and neck and prostate using physical dose-volumebased optimization. As a result, dose-volume limits are implemented in mostcommercial inverse planning systems today including the treatment planningsystem (TPS) used in this thesis. More and more detailed dosimetric issuesare accounted for in modern systems, such as tongue and groove correctionand hot beamlet correction (which is a correction method removing the spikesexceeding a certain threshold from the intensity pro les). Generally speak-ing, many of the recent developments in inverse planning aim at a better,more complete, representation of the clinical prescription and objectives, sothat the resulting IMRT treatment plans become more clinically meaningful,rather than mathematically optimal.Forward planning has also seen improvement due to increases in compu-tational speed. Early pencil beam algorithms are being replaced by moresophisticated and reliable Monte Carlo algorithms, for which the computa-tional e ciency is much better than before.212.4 Treatment Plan Veri cationThe paper ‘Comprehensive quality assurance for the delivery of intensitymodulated radiotherapy with a multi-leaf collimator used in the dynamicmode’ by LoSasso et al, is usually treated as a primary reference for IMRTquality assurance[4]. In this paper, the author proposed that IMRT QAshould involve routine machine QA and veri cation of individual patienttreatments. Routine machine QA focuses on a series of MLC basic perfor-mance tests, such as leaf positioning accuracy, leaf gap accuracy and stability.The veri cation of patient treatment plans includes ion chamber measure-ment,  lm dosimetry and MLC log  le analysis. This thesis generally followsthese recommendations.For dose veri cation purposes, ion chamber measurement and  lm dosime-try are two main measurement methods. Due to the fact that the treatmenttechnique to be tested in this thesis is a combination of IMRT and SRT tech-niques, there are several additional challenges compared to traditional IMRTdose veri cation. The primary cause of these challenges is the small  eld size.For SRT, a typical  eld size is approximately 20 mm in diameter, comparedto a typical body IMRT  eld size which can be as large as 10 cm in diameteror more. The BrainLAB m3 MLC, designed for this purpose, has a maxi-mum  eld opening of only 9.8 cm by 9.8 cm at 100 SAD. The MLC leavesare also narrower to provide a higher spatial resolution. As a result, elec-tron disequilibrium is also more signi cant than for a traditional IMRT. Thesmall  eld size and high degree of dose modulation causes signi cant electrondisequilibrium. For individual  elds, the dose distribution is usually highlymodulated compared with the total accumulated dose. The in uence of suchhigh dose gradients on  lm dosimetry is still unclear. A special small sizeion chamber should be used to measure the dose from such highly modulatedsmall  elds, otherwise the volume averaging e ect will compromise the dosereading in high gradient regions. A direct consequence of using a smaller ionchamber is a relatively lower sensitivity, which is obviously a disadvantage.22Furthermore, as a direct consequence of high dose gradient and spatial reso-lution, the positioning tolerance for stereotactic techniques is much tighter.Special immobilization devices are used and mechanical performance errorshave to be taken into consideration.The detailed introduction of BrainLAB m3 micro MLC is given in Chap-ter 4, Section 4.2.2. The discussion of localization and immobilization devicesis given in Chapter 4, Section 4.2.1. Detailed descriptions of the clinical doseveri cation are given in Chapter 4, Section 4.3. In this section, a highly sus-pected directional dependence for ion chamber response is investigated andruled out. The numerical data of clinical case dose veri cation are presentedin the Chapter 5, where related discussions are also presented.23Chapter 3Film DosimetryFilm dosimetry is a high resolution, two dimensional relative dosimetry method.As the spatial accuracy of a dose distribution is of great importance,  lmdosimetry is a useful tool for commissioning and veri cation of IMRT tech-niques. Conventional radiographic  lm is light sensitive, thus should be han-dled without exposing to light. In addition, the necessary chemical process-ing introduces uncertainty and is ine cient. The more recently developedradiochromic  lm is insensitive to room light and does not need chemicalprocessing. Due to the nature of radiochromic  lm, readout may be conve-niently performed using a  atbed optical scanner. In addition, radiochromic lm is more tissue equivalent than radiographic  lm. In this project, we in-vestigate the performance of radiochromic  lm and apply it to measure thedose distributions as part of the commissioning process for the stereotacticIMRT technique.3.1 Introduction to Film DosimetryThe principle of  lm dosimetry is that the micro particles within the activelayer are sensitive to ionizing radiation. High energy particles deposit energywhen they pass through the  lm interacting with those micro particles, andthe deposited energy consequently changes the physical or chemical proper-ties of the local micro structure. This usually results in a change of darknessor color of the  lm, which can be quantitatively measured by light detectingdevices, such as optical scanners or densitometers. Depending on which scan-ning system was applied, the  nal dose distribution can be acquired either24as a fully continuous two dimensional distribution or dose points on a grid.The traditional medical radiographic  lm relies on similar chemical re-actions as in photographic  lm. Brie y speaking, the exposure to ionizingradiation causes the silver bromide in the  lm emulsion to decompose andcreate silver ions. Post-exposure chemical processing is a necessity to developand  x the exposure pattern on the  lm and stop the  lm from possible futurereaction to light. The darkness of the exposed  lm, or the optical density,is sensitive to the environmental conditions during chemical processing. Theconcentration of the developing solution, the duration of the development andthe temperature all could in uence the optical density. For example, varia-tions of 1uni2103 in the developer bath can a ect the optical density by as muchas 10 %[12]. Because the chemical in the solution is continuously consumedduring the process, no two pieces of  lm can be processed under exactlythe same condition. The inconsistency of processing environment leads to uctuation of  lm response. This drift theoretically restricts the accuracy ofthe dose information we get. Therefore,  lm dosimetry is believed to haverelatively poor accuracy and usually should be used as a relative dosimetrymethod together with some absolute dosimetry techniques such as ion cham-bers. In this case, the ion chamber reading can be treated as an absolutedose reference point. The relative dose ratio among di erent pixels, which isless sensitive to variations in processing, can be used along with the referencepoint dose to calculate dose value for each pixel.Despite the major di culties mentioned above,  lm has several advan-tages that most other dosimetry methods do not have. The most obviousone is the high spatial resolution. The spatial resolution of radiographic orphotographic  lm is theoretically limited by the size of light sensitive par-ticles such as silver bromide grains in the active layer. In principle theseparticles are generally smaller than 0.1 mm. In practice, resolution is limitedby the aperture of the scanning device or densitometer, which can be up to 0.3 mm. However,  lm resolution is still higher than other two dimen-25sional dosimetry methods such as EPID (Electronic Portal Imaging Device),or TLD (Thermoluminescent Dosimetry) grid. Another attractive feature isthe robustness and the  exibility of  lm which can be easily cut into di erentsizes or shapes and  t into interior planes of phantoms. This could not beeasily done by projective imaging modalities such as EPID.3.1.1 General Considerations when Using FilmGenerally, the following characteristics of  lm should be carefully evaluatedsince they will greatly in uence the dose measurement: energy response (re-lated to e ective Z), dose rate dependence, dose range , linearity, directionaldependence and sensitivity to ambient conditions.Energy ResponseEnergy response is how sensitive the  lm is to incident photons of di erentenergy. Ideally, energy response should be as uniform as possible. How-ever, typically radiographic  lm is 10 times as sensitive to 0.1 MV photonsthan to 1 MV photons because of the stronger photoelectric interaction withAgBr below 150 keV[12]. Scattered components often have relatively low en-ergy after the photons undergo several interactions and lose energy. Typicalradiographic  lm thus tends to be much more sensitive to the scattered com-ponents, which is a drawback since the in uence of the scattered componentstends to be ampli ed. In some cases, where the energy spectrum dramaticallychanges with depth, one should expect the sensitivity will change accordingly.Dose Rate DependenceDose rate dependence should be minimal in order for any  lm dosimetry sys-tem to work. If there were to be a strong dose rate dependence, one wouldbe required to know the dose rate at any point in order to determine thedose distribution correctly, which is practically impossible. Most x-ray  lm26Type Range (cGy) Saturation (cGy)Kodak EDR2 25 400 700Kodak XV-2 0 100 200XTL 1 15 30Kodak PPL 0.25 5 10MD55 0 3000 N/AEBT 0 800 N/ATable 3.1: Dose range and saturation level of di erent types of  lmmanufacturers claim their product is safe enough to be treated as dose rateindependent. Several types of radiochromic  lm have been experimentallytested for dose rate dependence. Galante et al have reported that for poly-carbonate, results show up to 10 % dose rate dependence for a dose rate rangefrom 433 cGy/min to 4333 cGy/min. When radiochromic  lm is intended tobe used for real-time dose monitoring, Rink et al have reported up to 4 %dose rate dependence.Dose RangeThe useful dose range determines which type of  lm is suitable for a particularapplication. Some data including dose range and saturation level of di erenttypes of  lm is listed in Table 3.1 [12].Tissue EquivalenceRadiographic  lm usually has di erent e ective Z from normal tissue. Thiscould lead to di erent responses between the  lm and the tissue, and leads toinaccurate representation of dose to tissue. Gafchromic  lms (such as EBT)have e ective Z much closer to normal tissue, within the range 6.0 to 6.5[21],compared with the e ective Z of tissue which is 7.22.27Ambient ConditionsThe  lm response can also be a ected by environmental conditions such ashumidity and temperature or storage conditions, storage time, post-exposuredevelopment time. Even more importantly, how scanning devices interactwith  lm could bring more complexity to the problem. The general strategyis to acquire a calibration  lm each time and try to maintain all the controlfactors the same during the acquisition of both measurement  lms and cali-bration  lms. Some of these aspects of radiochromic  lm dosimetry will bediscussed again in the following sections.3.1.2 Introduction to Radiochromic FilmRadiochromic  lm is a relatively new type of  lm. During irradiation, itsactive component within the transparent layer undergoes a polymerizationprocess and becomes dark blue. The optical absorption peak is at 670 nm forthe darkened region. The color change induced by the radiation is stable andpermanent. Unlike radiographic  lms, radiochromic  lms do not require postchemical processing, which eliminates the complication due to processingvariations. Furthermore, radiochromic  lms have very low sensitivity to roomlight. Hence the change of optical density is negligible if exposed to roomlight for a reasonable duration, which brings convenience to the handling of lm. The  lm can be cut or loaded under normal room light. The strongabsorption peak in red visible light makes it compatible with RGB  atbedscanners, which could potentially save the investment in an expensive highquality medical  lm scanner. Due to these advantages, we decided to applythis new type of radiochromic  lm to IMRT dose veri cation. In order to useradiochromic  lm as a reliable tool to validate our system, the performanceof the  lm was thoroughly tested. A proper and detailed handling procedurewas established. The overall estimate of the uncertainty was determined.The biggest radiochromic product line, which is manufactured by Inter-28national Specialty products Inc. (ISP) is named Gafchromic®  lm. Some-times radiochromic  lm is also referred to as Gafchromic  lm. There is aseries of Gafchromic products suitable for di erent applications. The recentGafchromic EBT®  lm is specially designed for application to IMRT QA.The sensitivity and dose range  t IMRT QA the best. Thus, in our researchwe chose to work with Gafchromic EBT  lm.The basic Gafchromic EBT  lm properties are as follows. The dose rangeis from 1 cGy to 800 cGy. EBT is ten times more sensitive than the previousgeneration Gafchromic HS  lm and MD-55, and is energy independent fromthe keV range into the MeV range. The uniformity is better than 1.5 %. Ithas faster and lower post-exposure density growth and can withstand tem-peratures of up to 70uni2103.Gafchromic EBT  lm has two identical active layers sandwiched betweensurface layers. Two clear polymer layers are on the outside of the two surfacesproviding protection. With this con guration, the  lm could be immersedinto water for a relatively long period. However, because the edge is notsealed, water can di use into the active layer. But the rate of di usion isrelatively low. Experiment shows that after 24 hours of immersion, water canpenetrate as far as 8 mm from the edge. The di used region can be easilyrecognized because it turns milky opaque if water molecules have di used.Cutting could also induce tension at the edge. Reading should be avoided upto several millimeters from the edge. It is also convenient to be able to useink pen to draw marks on either surface of the  lm. Marking is necessary tokeep track of the orientation of the  lm.The tissue equivalence of the new Gafchromic EBT  lm is better thanthe old model (Gafchromic HS and Gafchromic MD55). The e ective Z is6.98. This value is closer to the e ective Z of water (7.3) than the value forGafchromic MD-55 ( 6.5). The details of the atomic composition is shownin Table 3.2.Other properties including post-exposure development, uniformity, and29C H O N Li Cl Ze 42.3 % 39.7 % 16.2 % 1.1 % 0.3 % 0.3 % 6.98Table 3.2: Atomic composition and e ective atomic number of GafchromicEBT  lmdose rate dependence are discussed in the following sections.3.2 Film Scanning and Calibration3.2.1 ScannersIn order to be able to quantitatively analyze the  lm response, the amountof darkening has to be converted into digital signal. This is usually done byscanning the  lm using a scanner or measuring the optical density with adensitometer.Film scanners can be largely categorized into two groups: multiple detec-tor scanners and single detector scanners. The  rst group includes medicalgrade scanners such as the Vidar Diagnostic Pro advantage digitizer andhigh end consumer grade  atbed scanner such as Epson 1000XL. An exam-ple of the latter group includes the Welho er WP 102  lm densitometer andthe AZTEK® Premier drum scanner. It is quite obvious that in order toacquire a two dimensional signal array, the single detector scanner has tomechanically move the scanning head in a sweeping pattern across the  eldof interest, while the multiple detector scanner only has to move the head inone direction (provided the multiple detectors are arranged in a line), andtherefore making it faster than the single detector scanner. By sacri cingthe time e ciency, the single detector scanner eliminates potential interfer-ence across di erent sources and detectors, and eliminates non-uniformitiesbetween detectors. Generally speaking, the quality of the data acquired fromsingle detector scanners is better than multiple detector scanners. However,30in practice, the time consumption is usually intolerable and a multiple de-tector scanner is the preferred option.Drum scanners are used in high end photographic scanning. The materialto be scanned has to be transparent. The  lm is attached to a vacuum glasscylinder, which is spinning at a very high speed. A narrow laser beam passesthrough the  lm and goes into the vacuum tube, and is converted into electricsignal. The spot size is tunable and the scanning speed is relatively high.More importantly, the output quality is very high. However, drum scannersare not commonly used and are extremely expensive. The machine itself ishuge and heavy. Also, the way that the  lm has to be submerged into oiland attached to the drum makes it impractical in a clinical environment.The Wellhofer WP 102  lm scanning densitometer is available in theVCC medical physics department. The light source in this densitometer has awavelength at 950 nm which is signi cantly di erent from the absorption peakof Gafchromic EBT  lm, which is around 635 nm. Test scanning on WellhoferWP 102  lm densitometer with fully exposed and un-exposed GafchromicEBT  lm shows that the densitometer cannot distinguish them. The 950 nmlight is not attenuated by the dark blue of the  lm. Unless we can changethe light source of the Wellhofer WP 102  lm densitometer, it would notwork with the Gafchromic EBT  lm. Replacement of the light source wouldrequire extensive work and the quality of such a modi ed densitometer hasnot been tested, so this densitometer has not been considered in this work.The structure of the Vidar Diagnostic Pro® advantage digitizer is simi-lar to a  atbed scanner. The major di erence is that instead of keeping thetarget still and moving the scanning head across the  eld, the  lm actuallymoves through the scanner in the Vidar scanner design. This could poten-tially induce variation in the longitudinal direction with a unique pattern[22].The Vidar scanner is generally believed to have better signal to noise ratioto  atbed scanners due to the fact that it uses LED as the light source. Theunique LED light source not only eliminates the necessity of warming up but31also dramatically increases the dynamic range. The primary problem withthe Vidar scanner is that the wavelength of the light source does not matchthe absorption peak of the Gafchromic EBT  lm, which could result in rel-atively low sensitivity. In the VCC medical physics department, the Vidarscanner is currently malfunctioning so we could not do any tests on it, butwe have some previous data from scanning radiographic  lm on the Vidarscanner, which can be used as a reference if necessary.As  atbed scanners may be the most suitable options for clinical  lmdosimetry, we have chosen to test two models of  atbed scanners availableto us. As suggested by the  lm manufacturer, the  atbed scanner not onlyworks well but also provides several unique advantages. For example, thescanning can be done in RGB mode and the red channel can be extracted toprovide the highest contrast, or the information contained in the other twochannels could be used to do some additional correction, which is the casefor Gafchromic EBT2  lm. The two available scanner models at VCC arethe Microteck 9800XL and the AGFA Duoscan TMA 1600.The Microteck scanner is relatively newer than the AGFA scanner, how-ever, with the Microteck scanner we found a severe repeatable  uctuation ofmeasured pixel value in the longitudinal direction. The longitudinal pro le ofa uniformly exposed  lm shows a unique \W" shape pattern repeatedly. Theamplitude of the  uctuation is relatively high and could dramatically a ectthe dose measurement. Without being able to  x the scanner we decided tofocus e orts on the older AGFA scanner.Although it is relatively old, the AGFA Duoscan scanner still provides16 bits per color channel and overall 48 bits per pixel. The con gurationis di erent from most modern scanners in that it has a separate slot witha glass tray to support a transparent target. The scanner is connected tothe computer by a SCIS interface. A stand-alone version of software (AGFAFotoLook 3.60.00) is used to provide the software interface for scanning.Figure 3.1 shows the  lm insert’s position on the AGFA Duoscan scanner.32(a) (b)Figure 3.1: These photographs show the AGFA Duoscan scanner we use inthis project. 3.1(b) explicitly shows the position of the transparent insertused to hold the  lm.333.2.2 General settingsWhen scanning a  lm, the scanner should be set to RGB mode and withcolor depth of 48 bits. This gives the red channel a 16 bit color depth, whichis equivalent to 65536 steps from black to red. Using 8 bits per channelis strongly not recommended since that could only leave 256 steps in colordepth and when converted into dose, the dose resolution is low.Spatial resolution should be set to 300 dpi. Most modern scanners in-cluding the AGFA Duoscan scanner we use, can easily go up to very highspatial resolution for example 1800 dpi. Considering that the dose calcula-tion in most clinical treatment planning systems have a resolution of about 1pixel per mm, there is no advantage of scanning the  lm with a much higherresolution. However, the suggested 72 dpi[23] from the manufacturer is notadvised. Scanning in low resolution using a scanner with much higher capa-bility could dramatically increase the noise level. This can be explained inthe following way. In order to get a low resolution output the scanner caneither average the signal from several pixels, or it can simply skip severalpixels and pick one particular pixel as a representative. Most scanners willdo latter in order to increase the speed of scanning. Since the limiting factorto scanning speed is how fast the head can move, skipping several lines ismuch faster than scanning each line and doing the average internally. If onescans the  lm in 72 dpi there is a good chance that the particular pixel willcoincidently be a dirt or a scratch, although the physical dimension of thatscratch could be much smaller than the pixel it represents. So we determinedthat the best technique is to scan the  lm using a moderate resolution anduse software to lower the resolution by using a digital  ltering method. Me-dian  lter is the one used on this step in this project. By doing this, thesignal to noise ratio can be maintained. Figures 3.2 demonstrates this e ect.The  uorescent light needs to warm up to keep the light output stable.This means before each scanning session the lamp of the scanner has to beturned on for as least half an hour. The reproducibility of repeated scans34(a) (b)0 50 100 150 200 2501234567x 104PixelPixel Value(c)0 50 100 150 200 2501234567x 104PixelPixel Value(d)Figure 3.2: Scanned images demonstrating the di erence between scanningthe  lm in low resolution mode (100 dpi) 3.2(a) and scanning the  lm in highresolution mode (300 dpi) and then applying a median  lter to adjust theresolution to 100 dpi 3.2(b). The noise level of the  rst one is signi cantlyhigher than the second one. Di erences can be more easily distinguished in3.2(c) and 3.2(d). 3.2(c) and 3.2(d) are horizontal pro les through the imagein 3.2(a) and 3.2(b).35is indicated on the pro le shown in Figure 3.14. Scanning the  lm withoutwarming the lamp up could result in poor reliability. Although the absoluteamplitude of variation in pixel value seems small, the variation in measureddose could be signi cant.Furthermore, the dynamic range should be set to its maximum, and allcorrections to the image should be turned o . Scanned images should besaved in TIFF format.3.2.3 Position and Orientation DependenceAs mentioned by the  lm manufacturer and many other research groups, lm scanners that have a long di used light source and a CCD detector aresusceptible to artifacts that can a ect the performance. We shall call thisscattering problem. The scattering problem will lead to both position depen-dence and orientation dependence of the scanned image. Flatbed scannershave a bar shaped light source and normally use  uorescent light. Althoughthese scanners are designed to have some collimation above the CCD detec-tor array that only allows incident photons within a certain angle to enter,calibration is still required to make sure all the detectors produce the sameoutput signal when there is nothing being scanned (or when there is abso-lutely no attenuation). This procedure is automatically done each time ascan is performed. However, radiographic  lms and radiochromic  lms arecomposed of particles dispersed in a matrix having a di erent refractive in-dex, and these  lms will scatter light. As a result, some scattered photonswill pass through the collimation system. Even after an internal calibration,if a uniformly irradiated or un-exposed  lm is scanned, the signal acqiuredat the extreme left and right sides will be systematically lower than in themiddle. This non-uniform e ect is shown in Figures 3.3 and 3.4.Because this artifact is due to the shape of the light source, its response isrelated to the position of the  lm relative to the scanning area. For example,putting a smaller  lm in a di erent position within the scanning area will360 500 1000 1500 2000 25001234567x 104PixelPixel Value  UnexposedExposedFigure 3.3: These two pro les are taken from scanned images of an un-exposed un-cut piece of  lm and a uniformly exposed piece of  lm. Pro lesare along the horizontal direction which is perpendicular to the scanningdirection. It is obvious that these pro les do not indicate a uniform responsecross that direction even though the dose delivered is uniform or zero. Atthe edge of the  eld, the pixel value drops signi cantly for both images.37Horizontal DirectionScanning DirectionFigure 3.4: This image is taken from the same piece of  lm correspondingto the second pro le in Figure 3.3. The colormap has been adjusted so thatthe minimum and maximum thresholds are 16500 and 18500 instead of 0and 65535 by default. The black strips at both left and right side re ectthe scattering e ect. Those regions lack scattered photons making themsigni cantly lower in pixel value compared with the central region. The lightblue line in the center indicates where the pro le is taken from Figure 3.338result in di erent pixel values. The lateral pro le of a uniformly exposed lm will be systematically lower in the vicinity of left and right ends of thescanning area.The manufacturer of EBT  lm claims that the scattering e ect of the lm is due to the di erence of refractive index between the substrate and theactive component and that polymerizations will not change the refractiveindex. They believe that a correction curve for the scattering artifact willnot depend on the dose, thus a universal correction curve can be acquiredusing an un-exposed  lm. Many researchers have reached the opposite con-clusion that the correction curve does depend on the optical density[24, 25].Our experiments also veri ed that correction for the scattering e ect indeeddepends on the optical density. If we consider this dependence, the correc-tion curve now becomes a two variable function. If the scattering e ect isassumed to be only locally dependent, we could practically acquire this twovariables relationship and apply it to correct for this artifact[24, 25]. This isindeed how others have dealt with this problem and the reported overall ac-curacy is still around 4 %. For our measurements, we decided to not correctfor this, since our PTV target is small enough that we can put the region ofinterest onto the central region of scanning area. But e orts are also madeto ensure the measurement  lm and the calibration  lm are scanned in thecentral region by building a frame to easily accurately position  lms withdi erent sizes concentrically. The frames are L shape. The o set for theframe for 17 cm  lm pieces is 5 cm in the scanning direction and 2.5 cm inthe longitudinal direction. The o set for the frame used for 6 cm  lm piecesis 10.4 cm in the scanning direction and 7.5 cm in the longitudinal direction.The schematic diagram in Figure 3.5 shows the shape of the frame as well aswhere it was positioned relative to the scanning area and the  lm.Both the manufacturer and many research groups have reported thatGafchromic EBT  lm shows di erent response if the  lm is scanned in dif-ferent directions[23, 24, 26{28]. This has been referred to as the orientation39FrameFilmFigure 3.5: This schematic diagram shows the position of the  lm positioningframe. The grey outer box is the plastic border outside the scanning area.The L shaped dark region is the frame used to localize the  lm at the center.In actual fact the  lm is transparent.40dependence of scanning. An explanation provided by the manufacturer andreferred to in many articles is that the micro structure of the active compo-nent of the Gafchromic EBT  lm is needle shaped polymers. The orientationof these needle shaped polymers tends to be along a particular direction whichdepends on how the  lm is manufactured, in particular, how the protectivelayer is coated. The manufacturer indicates that the needle like micro struc-ture tends to align with the shorter direction of an un-cut  lm which is thecoating direction[24]. They also suggest that the  lm should be scanned inan orientation such that the alignment of the micro structure is parallel tothe scanning direction. In our experiment, we test this by scanning the sameset of  lms in two di erent orientations and compare the scanned results.The results show that the same piece of  lm scanned in two di erent direc-tions produces a signi cant di erence in pixel value. Pro les taken from eachimage also show that one pro le is approximately 10 % higher pixel valuecompared to the other. The results are shown in Figure 3.6.Since pixel values depend on  lm orientation, calibration has to be ac-quired such that all pieces are scanned in the same orientation as the mea-surement  lm. Although scanning in either portrait or landscape mode ispossible, there is a slight di erence in the noise level between the two. Scan-ning in portrait mode gives slightly lower noise, and the scattering e ect isless signi cant than for landscape mode. Hence in our project we choose toscan the  lm in portrait mode.3.2.4 CalibrationFilm response is represented as pixel value when the  lm is scanned. How-ever, pixel value is just an intermediate parameter. Ultimately the pixelvalue is be converted into a dose value for each pixel. Due to the fact that lm response for di erent batches is not expected to be the same, a universalcalibration relationship between pixel value and dose does not exist. In orderto establish a correlation between  lm response and dose reading, calibration410 50 100 150 200 2501234567x 104PixelPixel Value  PortraitLandscapeFigure 3.6: Pro les across the same line of the same piece of  lm (marked c3)scanned in portrait and landscape mode. The green line shows signi cantlyhigher pixel value than the other. It appears landscape mode gives a higherpixel value which con rms the hypothesis that landscape mode scatters morelight. Portrait mode is the recommended one in this thesis.42irradiation is necessary. The calibration irradiation is performed on a set ofsmall pieces of  lm under well de ned conditions. The dose given to thesecalibration pieces has to be well known. Calibration pieces of  lm are fromthe same batch as the one used in the experimental irradiation, acceptingthe assumption that pieces of  lm from the same batch are produced at thesame time and thus have much closer dose response. Although post-exposuredevelopment for Gafchromic  lm is reported to be relatively minimal[23], itcan contribute up to about 5 % of the overall optical density if the wait timeis long enough. To take this into account, the duration between irradiationand scan should be controlled. Either the scan time should be controlled sothat the  lm is only scanned after a certain time post irradiation, or the cal-ibration pieces should be irradiated directly before or after the experimentalirradiation and then scanned in the same manner. Requiring that the  lm al-ways be scanned after exactly a certain time post irradiation is in exible andinconvenient, therefore we choose to produce calibration pieces each time weuse the Gafchromic  lm. Furthermore, the manufacturer suggests that if the lm is used as an absolute dosimeter, calibration pieces should be acquiredevery time the  lm is being used[23].Calibration irradiation can be done in di erent ways, as long as the dosebeing delivered is well de ned and can be calculated with high certainty.However, for di erent calibration setups, practical issues a ect the outcome.One possible setup is to place the  lm strip parallel to the beam axis, so thatthe dose distribution along the center line is essentially a percentage depthdose (PDD) curve. Since PDD data is normally available (either the rawmeasured data, or from the TPS which is essentially interpolated data fromthe raw PDD table), by matching it with the dose response along the centerline, a correlation between dose and optical density can be acquired[25]. Thissetup was investigated in this project and several practical drawbacks werediscovered. First, the calibration is highly sensitive to geometry matchingerrors. Any uncertainty in the placement of the  lm will contribute to error43in the calibration curve. Particularly, if the  lm is not exactly parallel to thebeam axis, a slight rotation can a ect the dose distribution. It is challengingto absolutely restrict the  lm from any random shift or to trace the beamaxis based on the irradiation pattern. Secondly, this setup is more vulnerableto noise. Because the dose response keeps changing along the central beamaxis, it is not possible to average over this direction. In the perpendiculardirection, although it is safe to average over a short distance, it is still notreliable because the dose distribution is not theoretically  at in that direc-tion. Furthermore, a single PDD irradiation spans a limited dose range. Tocover the dose range needed for a normal single fraction prescription, severalcalibration curves should be pieced together. Depending on how the solidwater slabs are setup, single calibration curves sometimes su er from lack ofback scattering at the distal end. Together with the positioning uncertaintythis usually shows up as a discontinuity when trying to match these curvestogether. Finally, the parallel setup, is susceptible to  uence perturbationdue to the gap between the  lm and solid water.Practical experience shows that EBT  lm together with a  at-bad scannershows more noise than radiographic  lm with the Vidar scanner. Withoutclearly knowing what is the source of this  uctuation and how much it a ectsthe dose measurement, it is recommended to do the calibration irradiation inthe most simple, straightforward setup, avoiding any unnecessary potentialissue that can compromise the calibration accuracy. Therefore, the calibra-tion pieces for this project are irradiated perpendicular to the beam axisunder a square  eld.Two identical cube phantoms are used for IMRT measurements. Theouter dimensions are 19.5 cm × 19.5 cm × 19.5 cm. The phantoms areshown in Figure 3.7. One of the phantoms is designed to work with  lms.The  lm is held at the central slice through the geometric isocenter, whichis at a depth of 9.2 cm. Figure 3.8 shows the internal structure of thisphantom and how the  lm is held inside of it. Both calibration irradiation44Figure 3.7: Two cube phantoms used in this project. Both of them have thesame outer dimensions. The left one is designed for  lm dosimetry.45Figure 3.8: Part of the  lm cube phantom is opened showing the internalstructure.46 lm ID Monitor Unit (MU) Dose (cGy)c2 0 0a1 50 34.48a2 100 68.97a3 150 103.45b1 200 137.93b2 250 172.41b3 300 206.90c1 350 241.38c3 450 310.34d1 550 379.31d2 700 482.76d3 800 551.72Table 3.3: Monitor units and calculated dose for calibration exposuresand measurement irradiation are done at this depth. The  lm piece used fordose measurement is cut to 17 cm by 17 cm, while the calibration pieces are6 cm by 6 cm. When they were irradiated they were both positioned at thecenter of the phantom. The irradiation of both calibration and measurement lms is done at 100 cm SAD. For calibration, the  eld size is set to 4.2 cmby 4.2 cm with m3 micro collimator opening also set to 4.2 cm by 4.2 cmat isocenter. The monitor units delivered can be found in Table 3.3. Theactual dose given is calculated using iPlan treatment planning software inconformal treatment mode. The highest dose given is up to 551.7 cGy, whichis su cient for IMRT commissioning purposes. Matlab code is used to createthe calibration curve and convert the measured optical density distribution todose. The scanned  le is saved as tagged image  le format (.tif). Matlab candirectly read the  le into a three dimensional matrix. Matrix elements are16 bit integers. The three dimensions correspond to three color channels inthe order of red, green and blue. A roughly 1 cm by 1 cm square region in thecenter of each calibration piece is sampled pixel by pixel. The average pixelvalue is used as a representative value for each piece. The dose delivered to47each piece can be calculated based on the monitor units given, so that a pixelvalue to dose calibration curve can be acquired. Figure 3.9 shows the pixelvalue calibration curve. A fourth order polynomial  t to the measurementdata is used as the calibration curve. The curve  ts the measurement datawell except for the zero dose point.Another way used by many groups is to convert the pixel value to opticaldensity according to equation 3.1. Because the maximum possible readingwithout any attenuation is always internally calibrated to 65535 in 16 bitmode by the scanner, 65535 is always used as I0.OD =  log10 II0(3.1)ODnet = OD OD0 (3.2)Net optical density is calculated using Equation 3.2. Net optical densitydescribes the change in attenuation due to exposure, whereOD0represents the optical density of unexposed  lm. We have tried both cal-ibration methods and the results show negligible di erence except for theextremely low dose regions. The results are shown in Figures 3.11 and 3.12.The discrepancy is mainly due to interpolation between the zero dose pointand the  rst non-zero dose point. This discrepancy is believed to have nosigni cant clinical importance because the peripheral region of the IMRT eld is of relatively low importance. However, we still chose to perform  lmanalyze using the optical density technique due to this e ect.48Figure 3.9: This is the pixel value to dose calibration curve. The greenline represents the 4th order polynomial  t of the experimental data. Thecurve  ts with the original data well except for the zero dose point. Thisintroduces a discrepancy at the low dose end between pixel value calibrationand optical density calibration. Uncertainty of each point is caused by thenon-uniformity of the  lm and is approximately 500 for each point. They error bars for each point are approximately the same size as the circlesymbols.49Figure 3.10: Net optical density calibration curve50Figure 3.11: Comparison between pixel value and optical density calibrationtechniques. This image indicates the calculated dose distribution and theposition of the x pro le shown in Figure 3.12.51Figure 3.12: In this graph, the red curve is calculated using pixel valuecalibration while the green curve is calculated using the optical density cal-ibration. Both of them are based on the same data set. The comparisonindicates that the discrepancy is only signi cant at the low dose end.523.3 Performance of Gafchromic EBT Film3.3.1 Post-exposure DevelopmentPost-exposure development is measured by scanning the same piece of  lmsequentially at di erent times after exposure. Films are scanned three timesin each session and an average pixel value was plotted against the post ex-posure time. Result are shown in Figure 3.13. The  eld size was 10 cm by10 cm. Depth of solid water build up is 10 cm. The pixel value is a calculatedaverage over a 2.5 cm by 2.5 cm square central region. The dose was 200 cGyand was delivered using a 100 SAD setup. The majority of post-exposuredevelopment occurs in the  rst couple of hours, after which the rate of changeis very small. So, we suggest that the irradiated  lm should be stored at leastovernight before being scanned and if possible scanning of the measurement lm and calibration  lm should always be done together following the sameorder used in irradiation. This is the method applied to all  lm scanning inour project.The manufacturer claims that post-exposure development contributes lessthan 5 % of the overall darkness change. Their data suggests post-exposuredensity growth is not dose dependent. Our experiments show that post-exposure development contributes about 2 % of the optical density for a postexposure time of 15 hours.There is a concern about day-light sensitivity. The manufacturer claimsthe  lm can be exposed to o ce light for several hours without developingnoticeable density change. However, if exposed to sunlight or similar highintensity light environment, the density change can be much greater. Wehave not done any experiments to test this e ect. However, our practiceis to shorten such unnecessary exposure time. After the  lm is cut, it isalways put back into a light-proof envelope. Unused  lms are always kept ina light-proof environment.There is also concern about the density being a ected by the scanning53Figure 3.13: This graph shows the pixel values acquired at di erent post-exposure times of a single piece of  lm. The initial pixel value 34700 corre-sponds to 0.276 in optical density, while the last data point corresponds to0.282. The di erence (0.006) is approximately 2 % of the value of last datapoint, which corresponds to maximum development.54process because the light intensity of the scanner appears to be higher thanroom light. However, we have not found a report on this issue in any publica-tion. We tested the reproducibility of consecutive scans to see if the darknesscan be a ected noticeably. The result is shown in Figure 3.14. There is nonoticeable darkness change between the two scans. We conclude that thein uence of scanning on the optical density is negligible. This result also in-0 500 1000 1500 2000 250011.11.21.31.41.51.61.71.81.92 x 104PixelPixel Value  Scan 1Scan 2Figure 3.14: This graph shows the pro le of consecutive scans of the samepiece of  lm. The  lm is exposed to a uniform square  eld. The  lm isscanned at 300 dpi. A 5 by 5 Wiener  lter is used to reduce the noise. Thereis no obvious di erence between those two curves.dicates that the scanned result is highly reproducible in consecutive scans in55the short term, if the  lm is not taken out of the scanner and placed again.Introducing a delay between scans or removing the  lm from the scannerand placing it could lead to darkness development between scans, changes inalignment, change of distribution of the dirt on the  lm and the sensitivity uctuation of scanners.3.3.2 UniformityThe intrinsic uniformity of a single sheet of  lm is declared to be better than0.8 % by the manufacturer. They also claim that the scanning variationcontributes to 0.4 % and the total variation within a single sheet can be upto 1.2 % in total. The sheet to sheet variation is reported to be less than1.5 % by the manufacturer[29]. The optical density variation is believed tobe primarily due to the variation of the  lm’s active layer thickness[30]. Thisvariation is detected in our experiments. Figure 3.4 explicitly shows thisvariation. The graph shows the variation in red channel pixel value. Bydefault, the pixel value range is from 0 to 65535, with correlates with 16 bitgray scale. On this scale, the image appears uniform to the human eyes.However the pixel value is not the same over the entire region. If the colorwindow is adjusted to from 16500 to 18500, bright and dark bands will bevisible. However, the source of this variation is not clear. It could be theintrinsic variation within the  lm, or it may be caused by the scanner. Todetermine the source of this variation, the same piece of  lm is scanned asecond time with the  lm shifted by a small amount in the scanning direction.In Figure 3.15, pro les along the scanning direction are plotted. Here we seethat the trend of variation appears to be the same for both pro les exceptthat one pro le is shifted relative to the other, which con rms that variationshifts with the shift in  lm position. This is evidence that the variation iscaused by the  lm not the scanner. The amplitude of the variation in thisexample is at most 500 in pixel value which corresponds to a variation inpixel value around 3 % of the mean pixel value. It is more important to560 500 1000 1500 2000 2500 30001.651.71.751.81.85 x 104PixelPixel Value  Scan 1Scan 2Figure 3.15: Longitudinal pro les of consecutive scans of the same piece of lm is show in this graph. The  lm is shifted on purpose between the scans.The green pro les clearly is shifted towards left relative to the blue pro le,which con rms that the variation is caused by the  lm itself not the scanner.57convert this variation in pixel value amplitude to variation in dose reading.This, however, highly depends on how much dose is given, due to the factthat the calibration curve is non linear and the derivative of the calibrationcurve depends on dose itself. In this example, it is at the high dose end ofthe calibration curve, the variation in dose will be greater than 3 %. Thusthe intrinsic variation in  lm optical density is the major factor limitingthe accuracy of applying the Gafchromic EBT  lm for dosimetric purposes.This problem is acknowledged by the manufacturer. A new product namedGafchromic EBT2  lm has been introduced to address this problem. Detailsof the new Gafchromic EBT2 product are discussed in Section 7.1.3.3.3 Dose Rate DependenceDose rate dependence of Gafchromic EBT  lm is theoretically a concern forthe purpose of verifying IMRT treatment, because the dose rate is not con-stant during delivery. Rink et al have investigated the dose rate dependenceof Gafchromic EBT  lm for real-time changes in optical density[31]. Theyclaim that a statistically signi cant di erence between change of optical den-sity of  lms exposed at di erent dose rates occurs within approximately oneorder of magnitude change in the dose rate. The percent standard deviationof the change of optical density values using dose rate from 16 cGy/min to520 cGy/min was <4.5 %. It is important to note the fact that this de-pendence is for real-time optical density change, with is di erent from thesituation in our project, where the  lm is stored at least over night to allowthe post-exposure density to fully developed. The post-exposure density de-velopment could contribute to as much as 2 % from 100 min to 900 min postexposure in our experiments. The rate of change of optical density is muchhigher in the  rst 100 min post irradiation indicated by extrapolation of theexperimental data. It is understandable that read time dose rate depen-dence is related to the post-exposure darkness development. However, thedid not test the dose rate dependence for non real-time reading. Although58the manufacturer declare the dose rate dependence is negligible, this remainsa potential cause for uncertainty in dose reading.3.3.4 Overall AccuracyThe overall accuracy of the  lm dosimetry system was tested by comparingthe dose distribution acquired using the  lm with calculation results. Thisaccuracy test accounted for the performance of  lm, the scanner, the codesand the procedure we use. Six pieces of  lm were irradiated with square elds under a 100 SAD setup condition. The dose value was picked up ina small square around the isocenter and the average value was used as themeasured dose value for each piece. In order to exclude any in uence ofthe TPS calculation error, the setup of the measurement irradiation was thesame as the setup used to expose the calibration pieces, with  eld size of4.2 cm by 4.2 cm and SAD of 100 cm. Therefore the same MU to dose ratiowas used for both the calibration calculation and dose calculation for themeasurement pieces. This ratio, if inaccurate, would a ect the calculationof both the calibration pieces and the measurement pieces in the same wayand would cancel out. Hence, this test excludes any TPS calculation errors.Among the six pieces of  lm, three were given 500 MU. The other threeare given 250 MU, 700 MU and 100 MU respectively. The same Matlabcode was used to convert the  lm response to dose and the dose distributionwas evaluated using OmniPro I’mRT® software. Figure 3.16 shows how theregion of interest was chosen. The data are listed in Table 3.4. They showthat for well de ned square  elds the  lm measurement yields up to 2.5 %error in dose.To determine the reason for the discrepancies, an additional experimentwas done to test the consistency among di erent pieces of  lm. Three cali-bration sets, each containing 12 pieces of  lm and all acquired in the sameway, were rescanned in one session. The calibration curves were recalculatedbased on the new data and a set of optical densities from one of them was  t59Film ID MU Calc. Dose (cGy) Measured Dose (cGy) Di . (%)E1 500 344.8 342.3 -0.73E2 500 344.8 342.8 -0.58E3 500 344.8 350.7 1.71F1 250 172.4 169.4 -1.74F2 700 482.8 495.0 2.53F3 100 69.0 70.8 2.61Table 3.4: Results of the overall accuracy testmFigure 3.16: The region of interest for the  lm data listed in Table 3.4 isshown here.60through all three calibration curves. Calculated dose values were compared.Figure 3.17 shows the three calibration curves. The Table 3.5 shows the dose−1 0 1 2 3 4 5 60.10.150.20.250.30.350.40.450.50.550.6Dose (Gy)Optical Density  Nov 9 Batch 3Oct 7 Batch 3Oct 7 Batch 3Figure 3.17: Three calibration curves scanned and generated in one sessionvalues calculated using three di erent calibration curves. The percentagedi erences are listed. The discrepancies were up to 2.5 %, which is the evi-dence that the discrepancies found in Table 3.4 are primarily caused by thevariation in  lm response di erence from piece to piece. This di erence isexpected to be similar to the variation across a single piece of  lm, althoughthe magnitude of the non uniformity within single pieces may be smaller.This non uniformity e ect contributes at most 2.5 % to the overall discrep-ancies in our experimental results. It is this non uniformity e ect that the61manufacturer intents to address with EBT2®  lm.Dose by Original Dose by Dose by Di . of Di . ofCalibration (Gy) Cali. 2 (Gy) Cali. 3 (Gy) Cali. 2 (%) Cali. 3 (%)0.000 -0.004 -0.005 N/A N/A0.345 0.353 0.347 2.317 0.7250.690 0.694 0.678 0.623 -1.7341.034 1.018 1.034 -1.575 -0.0271.379 1.376 1.373 -0.235 -0.4691.724 1.718 1.739 -0.354 0.8442.069 2.075 2.096 0.305 1.3072.414 2.471 2.419 2.374 0.2223.103 3.130 3.183 0.862 2.5603.793 3.885 3.841 2.429 1.2524.828 4.960 4.846 2.737 0.3825.517 5.569 5.604 0.930 1.564Table 3.5: Inter-piece variation of  lm calibration62Chapter 4Treatment Planning and DoseVeri cationA radiotherapy treatment system consists of a planning system and a deliverysystem. The treatment planning system applied in this work is the iPlan®treatment planning software. The delivery system consists of a Varian linearaccelerator and a BrainLAB m3 high-resolution multi-leaf collimator. Thetreatment planning system generates the IMRT treatment plan, which is es-sentially an optimized leaf sequence  le, based on the dose prescription to thePTV and the dose deposition constraints on the critical organs close to thePTV. The plan is then transferred through the central information systemto the Linac and MLC control consoles. The control consoles control the leafmotion as well as the dose rate during the treatment delivery. Before thesystem can be clinically used, the treatment system has to be commissionedto ensure that the planning system can correctly calculate dose distributionsbased on the leaf sequence  les, as well as to ensure that treatment can be de-livered according to the treatment plan within the tolerance level. It is clearthat dose veri cation is an essential part of the commissioning process. The rst section of this chapter focuses on the treatment planning system. It ex-plains the typical work  ow for generating a stereotactic treatment plan andhow the dose veri cation works from the software end. The second sectionintroduces the delivery system. Localization and immobilization devices aswell as the m3 micro MLC are discussed because they are directly related tothe dose delivery. The last section explains the methods and materials used indose veri cation. It includes detailed information about phantom structure,63MLC performance tests, ion chamber measurement and  lm dosimetry.4.1 iPlan Treatment Planning SoftwareThe iPlan treatment planning software is a multi-purpose planning systemdeveloped by BrainLAB. It can generate treatment plans for di erent kindsof techniques. The reason we choose it as the planning system for our projectis that the m3 high-resolution multi-leaf collimator used in this project is aproduct from the same company. Thus the iPlan treatment planning soft-ware and the m3 MLC should be highly compatible. Moreover, the m3 MLChas already been commissioned for static application for clinical use in iPlan.IPlan has some unique features compared with the Eclipse treatment plan-ning system. To give a clear picture, a brief introduction of the planningwork  ow using the iPlan system is given in the following sections. A  owchart of a typical treatment planning procedure using the iPlan system isshown in Figure 4.1.4.1.1 PatXferThe user interface of the iPlan treatment planning system has a step by stepguided style. Although di erent tasks are done using di erent modules ofthe system, the user interface features of these modules are all similar so thatusers can easily master all of them. A typical treatment plan using the iPlansystem starts with importing CT, MRI or PET images into the planningsystem. This is done using PatXfer, a module of the iPlan software package.4.1.2 iPlan ImageAfter the patient CT data set and MR data set are imported into the iPlansystem, the next step is to deal with the issues of image analysis such assetting up the coordinates, image fusion and contouring the organs and PTV.64PatXfer iPlanImage iPlanDosePhantomMappingTreatmentVerification/QACT, MR or PETiPlanFigure 4.1: The  ow chart of iPlanThis is done in iPlan Image. The software will guide the user through thetypical work sequence.For stereotactic treatment a rigid coordinate system is setup based onthe CT scan. The CT scan of the patient’s head is acquired with a speciallydesigned reference box  xed to the head, so that the box has a  xed spatialrelationship with the patient’s anatomy. There are two Aluminum rods em-bedded on each surface of this box which can be detected on the CT scanand represented as six dots on each slice. The software detects the positionof these six markers and uses the distance between each pair to determine therelative position of each slice to the reference coordinate system. This coordi-nate system is used throughout the remaining process of treatment planningand delivery, serving as a basis for accurate spatial positioning which is theessential di erence from non-stereotactic treatment techniques. Figure 4.2shows a screen capture of this step.MR images are acquired since they provide better soft tissue contrast65Figure 4.2: A screen capture of the iPlan image recognizing the markers onthe reference box which is used to determine the relative position of eachslice to the coordinate system66compared with CT. The MR data set has to be fused with the CT data setbefore it can be used to help with contouring critical organs and PTV. ForiPlan, image fusion is done automatically based on the least square algorithmusing the intensities of both images. The fusion result is checked and if notsatisfactory, users can manually adjust the fusion.The next step is to contour the critical organs as well as the tumor vol-ume. For intracranial treatment, the critical organs include the eyes, thebrainstem, the medulla oblongata, the optic tracts, the optic nerves and theoptic chiasm. The contouring of these structures is automatically done onthe MR images using an anatomic atlas. Tumor target is manually contouredby the oncologist. The PTV is grown from the gross target volume (GTV)by a certain margin, usually 2 mm. If there is no overlap between the PTVand any other critical structure, the next step is to proceed with treatmentplanning. Sometimes the PTV region does overlap with a nearby criticalstructure. Because the algorithm for beam optimization requires the over-lapping volume to be treated di erently, the user must then use the Booleanoperation function to separate the overlapping part from the critical organs.For example, in the situation that the brainstem overlaps with the PTV,two new volumes have to be generated: one named brainstem 1 referringto the overlapping part, the other named brainstem 2 referring to the non-overlapping part. During IMRT planning brainstem 1 will be assigned a lowerpriority (which in iPlan is called ‘guardian number’), so that treatment plan-ning software assures the tumor control of this volume as the  rst priority.The Figure 4.3 shows how the OAR is separated using Boolean operation.Figure 4.4 shows the screen shot of the iPlan image after the OAR and PTVare contoured.4.1.3 iPlan DoseThe iPlan dose module is the core of this planning system. The inverseoptimization and dose calculation is performed in this module. The work  ow67Figure 4.3: Screen capture of the Boolean operation interface.68Figure 4.4: Screen capture of the iPlan image after the objects are contoured.69starts with locating the isocenter, which is typically taken as the geometriccenter of the PTV. Then, treatment beams are associated with the isocenter.The treatment technique can be either conformal, IMRT or arcs. The numberof beams, beam angles and couch angles have to be speci ed and enteredby the planner. The treatment planning system will calculate leaf positionfor conformal treatment, or it will generate the leaf sequence  le for IMRTtreatment. To be able to optimize the leaf sequence  le and calculate thedose distribution correctly, the proper machine data set has to be chosen.These machine data  les are created based on measured data. The measureddata include PDD, RDF and the nominal linac output for a speci c set ofconditions. The machine data  les are energy speci c and machine speci c,although it is possible that several similar machines share the same data  leif the di erence among them is not signi cant. There is a debate on whetherthe IMRT beam arrangement should be coplanar. Body IMRT treatments areusually coplanar based on the fact that there is not much room for the couchto rotate without colliding with the gantry. But for intra-cranial treatment,since the target is located at the end of the couch and is held by a narrowerstage than the couch itself, the available couch angle range is much larger.The advantage of using non-coplanar setup is to avoid primary beam passingthrough the eyes or other critical organs.The next step is beam optimization. This is when the program uses dosevolume constraints set by the user to determine optimal modulation for eachbeam. Users have to  rst specify the grid size, which is the spatial resolutionused in the dose calculation. Smaller grid size gives higher accuracy butobviously costs more in calculation resources. It is advised that the gridsize should be bigger than the pencil beam algorithm beamlet size. Thedose prescription to the PTV and critical structure dose constraints can bespeci ed at this point. Both prescription and constraints are de ned on dosevolume histograms (DVH). BrainLAB suggests that the PTV dose constraintshould be set on the 50 % volume level, because the actual DVH curve70always shows an S shape, setting the constraint on 50 % volume is a goodcompromise to balance the cold spot and hot spot. However, at the VCC thestandard is to cover the PTV with the 90 % isodose surface. Practically, wemove the constraint point closer to the high volume side to force the resultto give better dose coverage by compromising the hot spot dose. Figure 4.5explains this problem.The optimization generates four di erent plans including: PTV only,OAR low, OAR medium and OAR high. As their names indicate, eachplan spares the OAR to di erent degrees. Usually, considering an acceptabledose coverage and reasonable OAR sparing, OAR low or OAR medium aregood candidates. If none of these plans is acceptable, the user can go back tomodify the OAR constraint and dose prescription to hopefully get a betterresult. This optimization process is iteration based and time consuming.Only one of the four plans can be chosen and saved for future access.4.1.4 Phantom MappingAfter the plan is completed a number of veri cation tests need to be madebefore the plan is ready for treating the patient. The dose distribution canbe exported by just clicking the ‘dose export’ button. However, for thepurpose of a phantom test, since the phantom has a completely di erentshape from the patient, the dose distribution from the original beam setupis di erent from the original calculation. The planning system should beable to re-calculate the dose based on a phantom shape so that this canbe used to compare with  lm measurement or ion chamber measurement.The iPlan phantom mapping module allows the user to map any existingIMRT treatment plan onto a phantom and calculate the dose distribution.The user must be careful to specify the isocenter to the right position of thephantom. For fractionated treatment, the total dose or MU is transferred.The program will ask for the number of fractions in order to calculate the dosedistribution for a single fraction. If the user wants to modify the number of71(a)(b)Figure 4.5: Di erent constraint points on the DVH a ecting the 90 % dosecoverage72fractions, they can multiply the prescription by an arbitrary factor. Finally,the dose distribution is exported so that the calculated dose distribution canbe compared to the measurement.The coordinate system used in phantom mapping is determined by thecoordinate system associated with the CT data of the phantom, in this case,a cube phantom. The  lm plane position relative to this coordinate systemdoes not vary in these experiments. To get the dose distribution of the  lmplane of a 17 cm × 17 cm square area with a 1 pt/mm resolution, the rangeshould be L: -83.63, R: 86.37, step: 1.0; F: -59.4, H: 110.6, step: 1.0; A: 5.82.This range is symmetric relative to the geometric center of the phantom.This is how the  lm data is registered with the measurement data.4.2 Stereotactic IMRT Delivery System4.2.1 Localization and ImmobilizationExtra-cranial or body IMRT treatment has limited spatial positioning accu-racy, because of the conventional localization and immobilization techniqueapplied. For body treatment, typical localization and immobilization devicesinclude thermoplastic masks and vacuum molded bags. The vacuum moldedbag is placed between the patient and the linac couch, while the thermoplas-tic is placed over the patients’ body. Both devices’ shape can be  xed tomaintain the consistency of the patient’s position. If the thermoplastic maskis not tight enough, the patient can still move a little bit within the mask.Often the mask does not  t the patient perfectly, so there is an intrinsicfreedom of movement. Figure 4.6 compares the thermoplastic masks used inconventional IMRT and stereotactic treatment. Moreover, unlike the intra-cranial environment, the relative position of the anatomy is more likely tochange for body treatment. Tumor target may shrink or migrate in betweenfractions. Breathing motion can also dramatically change the target positionfor some cases. All these factors contribute to the positioning inaccuracy of73(a)(b)Figure 4.6: The mask in Sub- gure 4.6(a) is the mask used in conventionalIMRT treatment. The mask is soft and not tight enough to fully restrictthe patient’s movement. The mask in Sub- gure 4.6(b) is the one used instereotactic treatment. It is more rigid and can restrict the movement of thejaw.74body IMRT treatment. Hence, body IMRT treatment is suitable for rela-tively large targets. The Varian Millennium MLC is used in these cases witha minimum leaf width of 5 mm projected at 100 cm SAD.For intra-cranial treatment, stereotactic radiosurgery is used to help in-crease the positioning accuracy. Traditional stereotactic radiosurgery usesan invasive head frame to restrain the patient’s motion. For the IMRT ap-plication, an alternative of using an immobilization mask is a much bettercandidate. Compared to the mask used in body IMRT this stereotactic im-mobilization mask is more rigid and has a bite block attached to it. The smallamount of movement within the mask is measured by taking two X-ray pho-tos from two di erent directions and matching the CT data to these photos.Furthermore, the position is also monitored using infra-red markers on thepatient in real time to maintain the accuracy. This frame-less SRS/SRTsystem can achieve comparable spatial accuracy to the technique using theframe system. The BrainLAB m3 MLC has a 3 mm minimum leaf widthprojected at 100 cm SAD surface, which provides higher spatial resolutionthan the Varian Millennium MLC[2, 6]. The combination of stereotactic andIMRT technique will bene t relatively smaller intra-cranial lesions comparedto the normal size that will bene t from conventional IMRT[6{10].4.2.2 BrainLAB m3 Micro MLCAs discussed in Chapter 2.3.2, the MLC is an essential component of theIMRT technique. The BrainLAB m3 micro multi-leaf collimator, as thename indicates, is designed to de ne a high resolution treatment  eld. Forstereotactic treatments, a steep dose fall-o is required to protect normal tis-sue and, in particular, critical structures from high doses. Stereotactic singlefraction treatment should meet special requirements of dose gradient steep-ness which have been de ned by the AAPM. For single fraction radiosurgery,the recommended limit for dose gradient in the beam penumbra (from 80 %to 20 %) should be greater than or equal to 60 %/3 mm[32]. The steepness of75the dose gradient in radiosurgery is a function of beam geometry and beampenumbra. The 5 mm width of the Varian Millennium MLC cannot guar-antee that in all cases the requirement of 60 %/3 mm can be met. The m3with its 3 mm-wide leaves has an e ective penumbra of less than 3 mm forall SRS  eld sizes and meets all SRS requirements[2, 6].Since the development of the m3 was a joint project between BrainLABand Varian, the m3 is designed based upon the architecture of a standardVarian Millennium MLC to be fully compatible with Varian linacs. Figure4.7 is a photograph of the BrainLAB m3 MLC. The M3 has 52 tungstenleaves (26 pairs), which move along a plane that is perpendicular to thebeam central axis. Unlike the Varian Millennium MLC that has a minimumleaf width of 5 mm at isocenter, the m3 has minimum leaf width of 3 mmaround isocenter. Details of leaf speci cation can be found in Table 4.1.Number of leaves 26 PairsLeaf width (at isocenter) 14 × 3.0 mm6 × 4.5 mm6 × 5.5 mmMaximum  eld size 9.8 cm × 9.8 cmMaximum leaf over-travel 5 cmClearance from isocenter 31 cmMaximum leaf speed 1.5 cm/sWeight 30 kgTable 4.1: Speci cations of BrainLAB m3 micro multi-leaf collimatorExcept for leaf width and leaf number, the most distinct di erence istongue and groove design and the leaf tip shape. A more complicated ‘tongueand groove’ leaf cross-section was necessary to allow drive shafts to be in-serted into each leaf. Adjacent leaves have this shaft inserted at verticalincrements to permit optimum positioning of each leaf’s driving motor. Leaftips are milled to three angled straight edges to cover the whole range thatthe leaves can travel. Compared to Varian Millennium MLC this angle is76Figure 4.7: This is a photograph of the BrainLAB m3 micro MLC. Thisdevice is attached to the gantry head when in use. The square window inthe center is where the beam passes through. Noticeably the maximum  eldsize is only 10 cm by 10 cm de ned at 100 cm distance. The leaves can beseen on the right edge inside of the window.77much smaller, about 3°, due to the fact the maximum  eld opening is only9.8 cm by 9.8 cm.4.3 Dose Veri cation4.3.1 PhantomTwo phantoms are constructed for the purpose of dose veri cation. BothFigure 4.8: This  gure shows both the ion chamber cube phantom and theinterchangeable inserts. Notice the di erent internal structure of the inserts.They were designed to  t for di erent types of ion chambers. The locationof the chamber tip is supposed to be at the geometric center of the cubephantom.78of them are plastic cubes with dimension of roughly 18.5 cm by 18.5 cmby 18.5 cm. The outer dimensions of both phantoms are identical, whilethe inner structure is di erent. One of them has a cylindrical hole which ts cylindrical inserts that in turn  t di erent types of chambers (The twomost frequently used chambers are a 0.01 cc chamber and a 0.1 cc chamber).Figure 4.8 shows both cube phantoms and the interchangeable inserts. Thechamber’s sensitive volume is located at the geometric center of the phantomif the chamber is fully inserted into the proper insert. The other phantomis for  lm dosimetry. It was originally designed to contain traditional radio-graphic  lm requiring it to be light proof. The cube has a shell with thicknessof 5 mm. Four identical slabs are placed inside the shell of the cube. The lm is supposed to be located between the second and the third slabs. Theslabs are tightly  t so that the air gap in between the slabs is small. Rotat-ing the phantom will not cause signi cant movement of the slabs within thephantom. Figure 4.9 illuminates the internal structure described above. Thematerial of which the phantom is made is polystyrene. The e ective atomicnumber is 5.74. This phantom can contain a piece of  lm with a maximumsize of 17 cm by 17 cm. For smaller size, four pieces of tape are pasted ontop of the second slab forming a double line cross shape to restrict the smallpiece of  lm inside of it. This is also shown in Figure 4.9. The overall uncer-tainty of  lm positioning is less than 1 mm in both longitudinal and lateraldirections.To simulate the clinical treatment setup, the stereotactic head rest is mod-i ed to support the phantom. Figure 4.10 shows how the head stage assemblyis used to hold the phantom. Two plastic bars are pasted on two adjacentedges of the plastic plate on the head rest. Pushing the phantom against thiscorner will assure the position of the geometric center of the phantom is at areproducible location. By switching between the two phantoms, ion chambermeasurement and  lm dosimetry experiments can be performed in the samecon guration. Ignoring minor di erence between the two phantoms, the ion79Figure 4.9: This photograph shows the internal structure of the  lm dosime-try phantom. The surface with tape on it is located at the central depth.The total area of that surface is 17 cm by 17 cm. When 6 cm by 6 cm piecesare used, they were restricted at the center by the cross of the tape.80(a) (b)(c) (d)Figure 4.10: These photographs indicate how the stereotactic head stage isattached to the linac couch and how the cube phantom is held on it. 4.10(a)shows how the adjustable stage is mounted on the linac couch. 4.10(b) showshow a modi ed plastic plate design for the phantom is connected on the stage.4.10(c) shows how the cube phantom is located on the plate. 4.10(d) showshow the reference box is connected to this assembly and used to help alignthe geometric center of the box to the isocenter.81chamber reading corresponds to the center point of the  lm. .4.3.2 MLC performance TestsMLC is the component that most likely could a ect the accuracy of beamdelivery. Therefore, MLC performance tests have to be performed prior todose veri cation. MLC performance tests for this project involve three parts:static MLC positioning test, dynamic MLC statistic test and the e ective leafgap test. These three tests focus on di erent aspects of MLC performance.The  rst one inspects the absolute leaf positioning accuracy for static beamdelivery; the second one measures the dynamic response of the MLC. Thelast one insures that the TPS correctly models the distance between actualleaf tip position and the divergence line of the beam edge.Static MLC Positioning TestBecause the m3 micro MLC is currently used in clinical treatment, the staticMLC positioning test is routinely performed during the monthly QA on it. Aset of MLC patterns include three simple rectangular openings with sizes of2, 5 and 9 cm, and a special pattern called \Multiport" is used. The actualMLC position is detected using EPID. A small amount of dose is deliveredwith each pattern of MLC. Edge position of each MLC leaf is measured onthe EPID acquired image and is compared with the planned position. Thewhole measurement is repeated for four times at di erent gantry positions(0, 90, 180, 270°) in order to take the e ect of gravity into account. Sampledata of the 5 cm by 10 cm pattern at 0° gantry position is shown in Table4.2. Figure 4.11 illuminate the data in Table 4.2. The last column of Table4.2 represents the actual distance between opposite leaves which is supposedto be around 50 mm. The tolerance level is set such that the discrepancyshould be less than 1 mm. Maintenance is performed when the discrepancyis larger than 2 mm. Since this static MLC positioning test is part of the82Lbank (mm) Di (mm) Lbank (mm) Di (mm) Sum (mm)Leaf 01 25.2 0.2 25.0 0.0 50.2Leaf 02 25.0 0.0 25.1 0.1 50.1Leaf 03 25.1 0.1 25.2 0.2 50.2Leaf 04 25.0 0.0 25.2 0.2 50.2Leaf 05 25.4 0.4 25.3 0.3 50.7Leaf 06 25.1 0.1 25.2 0.2 50.3Leaf 07 25.1 0.1 25.2 0.2 50.3Leaf 08 25.1 0.1 25.2 0.2 50.4Leaf 09 25.1 0.1 25.3 0.3 50.4Leaf 10 25.1 0.1 25.3 0.3 50.4Leaf 11 25.1 0.1 25.3 0.3 50.4Leaf 12 24.8 -0.2 25.3 0.3 50.1Leaf 13 25.2 0.2 25.3 0.3 50.5Leaf 14 24.8 -0.2 25.4 0.4 50.2Leaf 15 24.8 -0.2 25.5 0.5 50.2Leaf 16 25.2 0.2 25.4 0.4 50.6Leaf 17 25.2 0.2 25.4 0.4 50.6Leaf 18 24.9 -0.1 25.4 0.4 50.3Leaf 19 24.9 -0.1 25.4 0.4 50.3Leaf 20 24.9 -0.1 25.4 0.4 50.3Leaf 21 24.9 -0.1 25.5 0.5 50.4Leaf 22 24.9 -0.1 25.5 0.5 50.4Leaf 23 24.9 -0.1 25.5 0.5 50.4Leaf 24 24.6 -0.4 25.4 0.4 50.0Leaf 25 24.6 -0.4 25.5 0.5 50.1Leaf 26 24.7 -0.3 25.4 0.4 50.1Table 4.2: Results of the statistic MLC position test for the m3 MLC. Thedata was taken on Aug 26, 2009. The gap size was 5 cm. Gantry angle was0°. All distance are measured at 100 cm SAD.83                                   LBank (mm)         Diff (mm)         RBank (mm)         Diff (mm)         Sum (mm)Leaf 01                   25.2                  0.2                   25.0                  0.0                  50.2Leaf 02                   25.0                  0.0                   25.1                  0.1                  50.1Leaf 03                   25.1                  0.1                   25.2                  0.2                  50.2Leaf 04                   25.0                  0.0                   25.2                  0.2                  50.2Leaf 05                   25.4                  0.4                   25.3                  0.3                  50.7Leaf 06                   25.1                  0.1                   25.2                  0.2                  50.3Leaf 07                   25.1                  0.1                   25.2                  0.2                  50.3Leaf 08                   25.1                  0.1                   25.2                  0.2                  50.4Leaf 09                   25.1                  0.1                   25.3                  0.3                  50.4Leaf 10                   25.1                  0.1                   25.3                  0.3                  50.4Leaf 11                   25.1                  0.1                   25.3                  0.3                  50.4Leaf 12                   24.8                  -0.2                   25.3                  0.3                  50.1Leaf 13                   25.2                  0.2                   25.3                  0.3                  50.5Leaf 14                   24.8                  -0.2                   25.4                  0.4                  50.2Leaf 15                   24.8                  -0.2                   25.5                  0.5                  50.2Leaf 16                   25.2                  0.2                   25.4                  0.4                  50.6Leaf 17                   25.2                  0.2                   25.4                  0.4                  50.6Leaf 18                   24.9                  -0.1                   25.4                  0.4                  50.3Leaf 19                   24.9                  -0.1                   25.4                  0.4                  50.3Leaf 20                   24.9                  -0.1                   25.4                  0.4                  50.3Leaf 21                   24.9                  -0.1                   25.5                  0.5                  50.4Leaf 22                   24.9                  -0.1                   25.5                  0.5                  50.4Leaf 23                   24.9                  -0.1                   25.5                  0.5                  50.4Leaf 24                   24.6                  -0.4                   25.4                  0.4                  50.0Leaf 25                   24.6                  -0.4                   25.5                  0.5                  50.1Leaf 26                   24.7                  -0.3                   25.4                  0.4                  50.1Leaf Position Quality Assurance Test Results                                         Aug 26, 2009Beta version only.  NOT FOR CLINICAL USE   Right BankLeft Bank0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27Leaf #2324252627Leaf Position (mm)Figure 4.11: Static leaf position QA test resultsroutine monthly QA, there is a large amount of historical data available. Allthese data suggest that the average discrepancy is within the tolerance level.Occasionally, discrepancies can reach around 1 mm, but they are always wellbelow 2 mm.Dynamic Log  les Statistical AnalysisThe m3 micro MLC creates a log  le after each dynamic beam delivery. Thislog  le is called Dynalog  le. The dynalog  le contains information about thestates of all leaves recorded at each approximately 50 ms interval throughoutthe beam delivery. The format of the Dynalog  le is illustrated in Table4.3. The number on the very top-left corner is the tolerance level. The unitassociated with this number is MLC driving motor counts. This numberde nes the tolerable discrepancy between the actual leaf position and wherethe leaf should be. If the discrepancy goes beyond this value, a beam hold-o signal will be triggered. Starting from the next entry, the  rst columnis called dose fraction index, which ranges from 0 to approximately 25000.The dose index re ects the dose fraction of each entry. The second columnis the previous segment number. The MLC works as follows. In the DVA  le(leaf sequence  le), which de nes the leaf movement over the delivery time,leaf position is prede ned at certain points in time. These prede ned points84A#200B C D E F G H I J K L# # # # # # # # # # #0 0 0 1 0 25 -3070 -3069 -3070 -3070    13 0 1 1 0 25 -3070 -3069 -3070 -3070    69 1 1 0 25 832 -3070 -3069 -3070 -3070    69 1 1 0 25 832 -3070 -3069 -3070 -3070    ......24990 30 1 1 24167 25000 -3070 -3069 -3070 -3070    "MKey DescriptionA Plan toleranceB Current dose fraction. Range is from 0 (0 %) to 25000 (100 %)C Previous segment numberD Beam Hold- O stateE Beam ON stateF Previous segment’s dose index (out of 25000)G Next segment’s dose index (out of 25000)H Leaf A1 interpolated plan positionI Leaf A1 actual positionJ Leaf A1 plan position for previous dose indexK Leaf A1 plan position for next dose indexL Represents the repeating data for other leavesM Represents a completed  eld deliveryTable 4.3: Dynalog  le format85are called control points. The period in between two adjacent control pointsis called a segment, during which the leaf position is linearly interpolated.Usually, for one beam delivery, the number of segments is approximately30. The third column is the beam hold-o state, if a beam hold-o signal isissued, the entry will record as 1 for this column. The forth column is thebeam-on state. Theoretically, the beam on state and the beam hold-o stateshould be opposite. In reality, there is a one entry lag of the beam-on staterelative to the beam hold-o state. The  fth and sixth column representsthe previous and next control points’ dose indices. From the next column on,the columns contain the information for each leaf. Four columns as a groupare for each leaf. The  rst one represents where the leaf should be base oninterpolated data from the control points. The second one represents wherethe leaf actually is. The third and the forth columns are the leaf position ofprevious and next control points.Because the actual Dynalog  le is too long to be presented here, onlystatistical results are shown. The detailed discussion of the statistical analysisof Dynalog  le is continued in Section 5.1.1.E ective Leaf Gap TestBrainLAB refers to the e ective leaf gap test as the \Measurement of dynamicleaf shift". The dynamic leaf shift describes an e ective leaf shift due to theleaf end design. BrainLAB provides a standard procedure to measure thedynamic leaf shift which is essentially an ion chamber dose measurement ofsliding windows with di erent widths. The determined dynamic leaf shift isto be entered into the beam pro le  le at the \Radiologic Field/Leaf shiftdynamic"  eld of the iPlan system. The following steps are the detailedprocedure provided by BrainLAB:• Use a large detector (ionization chamber) and position itin the water phantom in such a way, that the detector axisis perpendicular to the leaf direction. Set the water sur-86face level to an SSD = 980 mm and adjust the detector atisocenter (depth = 20 mm).• Set the jaws to form a square  eld of 100 × 100 mm2.• Set the dose rate of the LINAC to 300 MU/min and deliver300 MU for each  eld.• Successively irradiate the dynamic mMLC  le \M3 1.d01",\M3 5.d01", . . . ,\M3 100.d01" (gap sizes: 1, 5, 10, 20, 50,100 mm) and note the dose values.• Close the mMLC and measure the leakage dose using thesame setting as above (asymmetric gap setting; intra-leafgap should be 50 mm o the isocenter).• Set the mMLC to a square  eld of 100 × 100 mm2 andmeasure the open  eld dose using the same setting as above.• The measured dose D can approximately be describe by alinear function D Dleak = b(gap+ 2 ) = b gap+a, wheregap is the nominal gap width (1, 5, . . . , 100 mm), Dleak is themeasured mMLC leakage and  is the e ective dynamic shiftper leaf. After determination of a and b by linear regression, is calculated by  = a=2b.The results of this dynamic shift measurement can be found in Section 5.1.2.4.3.3 Ion Chamber MeasurementThe ion chamber measurement is one of the most important parts of thedose veri cation, mostly because of its proven reliability. Film dosimetry isusually used as a relative dosimetry method; therefore a reference point isrequired and often measured by the ion chamber. Even if other dosimetrymethods are intended to be used as absolute dosimetry methods it is alwaysa good idea to cross check the data with ion chamber measurement, which is87of extreme importance if the reliability of these dosimetry methods are stillunder investigation.The ion chamber used in this project is a NAC009 TRIAX miniatureionization chamber. The chamber and the cable are shown in Figure 4.12.The electrometer used is a Victoreen Model 530 precision electrometer. AVarian linac is used to irradiate the phantom and is referred as Unit 2 inVancouver Cancer Center.(a) (b)Figure 4.12: These photographs show the mini chamber used in this project.Because the ion chamber is not calibrated for absolute dosimetry, anymeasurement using this ion chamber is relative. A calibration measurementis performed under a 10 cm by 10 cm  eld size at depth of 5 cm in solidwater. In order to be more e cient and check the machine output andambient condition variation, a consistency measurement is also performedjust before any clinical dose veri cation using the cube phantom. Details ofthis procedure are described in Section 3.2.4.Because most of the stereotactic treatment plans use a non-coplanar beamarrangement, it was also important to establish whether a signi cant angulardependence exists for ion chamber in the cube phantom.A cylindrical phantom was built for this purpose. The schematic diagramof the cylindrical phantom is shown in Figure 4.13. The cylinder is 9 inchhigh and with a diameter of 20.8 cm. A cylindrical hole is drilled from the88Figure 4.13: Schematic diagram of the cylindrical phantom.center of the side so that the same insert of the cube phantom can be  ttedinto this cylinder, holding the ion chamber at the geometric center. Thecylinder is build with acrylic resin which is the same material used to buildthe insert. During the experiment, the cylindrical phantom is sitting on thecouch in a way that the ion chamber axis is along the couch axis. Linacgantry is rotated to 90° or 270° so that the beam is coming from the side ofthe phantom. By rotating the couch, the beam incident angle with respectto the ion chamber axis is changed. The convention used in de ning the ionchamber angle is shown in Figure 4.14. A square  eld is delivered for every15° increment. Two  eld sizes are used. The results are presented as the ratioof the ion chamber reading at a given angle, normalized to the reading for theperpendicular case and are shown in Table 4.4 and Table 4.5. Figure 4.15and Figure 4.16 are graphical representations of the data in Table 4.4 andTable 4.5. In general, the ion chamber sensitivity increases when the incidentdirection approaches the ion chamber axis. The amplitude of variation is atmost 1.5 %. The detailed relationship relative to the angle is di erent forthe two  eld sizes and there is a relatively large uncertainty in the shapeof the curve. The relatively large uncertainty in the angular dependencecurve is likely due to gantry sag and wobble in couch rotation axis. Thesedata are not suitable for correcting IMRT veri cation measurements, but do89Figure 4.14: Schematic diagram showing the convention used in the de nitionof chamber angleLeft side Left side Right side Right side AverageAngle rdg. (nC) ratio rdg. (nC) ratio ratio90 -0.3924 1.0000 -0.3872 1.0000 1.000075 -0.3937 1.0033 -0.3878 1.0015 1.002460 -0.3937 1.0033 -0.3890 1.0046 1.004045 -0.3944 1.0051 -0.3899 1.0070 1.006030 -0.3952 1.0071 -0.3924 1.0134 1.010315 -0.3940 1.0041 -0.3904 1.0083 1.00620 -0.3950 1.0066 -0.3964 1.0238 1.0152Table 4.4: Ion chamber reading of a 4.2 cm by 4.2 cm square  eld withdi erent incident angle relative to the ion chamber axis. 200 MU is deliveredper each  eld.90Left side Left side Right side Right side AverageAngle rdg. (nC) ratio rdg. (nC) ratio ratio90 -0.3164 1.0000 -0.2965 1.0000 1.000075 -0.3137 0.9915 -0.2974 1.0030 0.997360 -0.3180 1.0051 -0.2979 1.0047 1.004945 -0.3188 1.0076 -0.2984 1.0064 1.007030 -0.3198 1.0107 -0.2991 1.0088 1.009815 -0.3199 1.0111 -0.2995 1.0101 1.01060 -0.3222 1.0183 -0.3010 1.0152 1.0168Table 4.5: Ion chamber reading of a 1.2 cm by 1.2 cm square  eld withdi erent incoming direction relative to the ion chamber axis. 200 MU isdelivered per each  eld.establish the magnitude of the directional dependence of the ion chambermeasurements. The stem irradiation theory predicts that when irradiate theion chamber along the long axis, more volume of the ion chamber stem will beirradiated and will contribute to the collected charge. This theory supportsthe result that approximating the ion chamber axis yields larger readings. Itis also expected that di erent  eld sizes or  eld shapes would yield di erentcurves. For IMRT dose veri cation,  eld sizes do dynamically change andthe cumulative dose is the average of several beams coming from di erentangles. So, that the expected e ect should be even smaller than the 1.5 %we acquired.910 10 20 30 40 50 60 70 80 9011.0051.011.0151.021.025Chamber Angle (Degree)Ratio  Left sideRight sideAverageFigure 4.15: Plot of the ion chamber angular dependence (4.2 cm)920 10 20 30 40 50 60 70 80 900.990.99511.0051.011.0151.021.025Chamber Angle (Degree)Ratio  Left sideRight sideAverageFigure 4.16: Plot of the ion chamber angular dependence (1.2 cm)93Chapter 5Results and DiscussionThis chapter contains results and discussion of the MLC performance tests,ion chamber measurements and  lm dosimetry measurements described inprevious chapters. MLC performance tests are presented in the  rst section.Because these experiments assess the general performance of the most criticalcomponent of the delivery hardware, the results of these tests assist with thedetermination of reasonable expectations for dose veri cation of clinical cases.After the MLC performance tests, the results and discussion of ion chambermeasurements are presented in the section 5.1.1. Ion chamber calibrationis discussed  rst, followed by ion chamber dose veri cation of three clinicalcases. The third section contains the dose veri cation data for  lm dosimetryfor the same clinical cases.5.1 MLC Performance Tests5.1.1 Dynamic MLC Dynalog Statistical AnalysisAs previously introduced in Section 4.3.2, the statistical analysis of the Dy-nalog  le of one particular beam of an IMRT treatment is presented in Table5.1 and Table 5.2 as an example. The tolerance level for MLC leaf positionis 200 motor counts. The analysis only includes those entries for which thebeam state is on. Sometimes a particular leaf may have to travel a relativelylong distance to reach the designated position required by the leaf sequence les. This usually happens at control points where a relatively sudden changeof the leaf position may be required. Under such a condition, the leaf usually94Max. of j j Ave. of j j Std. of j j Num. exceedingthe thresholdLeaf 01 1 1 0 0Leaf 02 1 1 0 0Leaf 03 1 1 0 0Leaf 04 1 1 0 0Leaf 05 1 1 0 0Leaf 06 1 1 0 0Leaf 07 1 1 0 0Leaf 08 547 5.1176 46.8149 1Leaf 09 592 21.2721 77.0172 4Leaf 10 302 11.7868 42.8664 3Leaf 11 744 31.4412 96.4892 7Leaf 12 770 32.0735 102.0655 5Leaf 13 795 8.0882 68.5996 1Leaf 14 815 9.4191 71.3447 1Leaf 15 815 9.6912 71.6380 1Leaf 16 798 8.8750 69.5089 1Leaf 17 753 6.6324 64.4779 1Leaf 18 649 5.8750 55.5605 1Leaf 19 1 1 0 0Leaf 20 1 1 0 0Leaf 21 1 1 0 0Leaf 22 1 1 0 0Leaf 23 1 1 0 0Leaf 24 1 1 0 0Leaf 25 1 1 0 0Leaf 26 1 1 0 0Table 5.1: This is the statistical analysis of the bank A Dynalog  le. The datais considered only when the beam is on. The  rst column is the maximumdiscrepancy between the planned leaf position and the actual leaf position.The second column is the average of the discrepancy. The third column isthe mean. The last column is the number of entries, of which the discrepancyexceeds the tolerance.95Max. of j j Ave. of j j Std. of j j Num. exceedingthe thresholdLeaf 01 1 1 0 0Leaf 02 1 1 0 0Leaf 03 1 1 0 0Leaf 04 1 1 0 0Leaf 05 1 1 0 0Leaf 06 1 1 0 0Leaf 07 1 1 0 0Leaf 08 481 4.5294 41.1597 1Leaf 09 398 14.7868 61.5429 4Leaf 10 124 4.1250 17.0609 0Leaf 11 443 30.5735 86.8110 7Leaf 12 507 31.1324 90.8176 6Leaf 13 254 2.8603 21.6946 1Leaf 14 252 2.8456 21.5231 1Leaf 15 252 2.8456 21.5231 1Leaf 16 277 3.0294 23.6668 1Leaf 17 329 3.4118 28.1258 1Leaf 18 375 3.7500 32.0702 1Leaf 19 1 1 0 0Leaf 20 1 1 0 0Leaf 21 1 1 0 0Leaf 22 1 1 0 0Leaf 23 1 1 0 0Leaf 24 1 1 0 0Leaf 25 1 1 0 0Leaf 26 1 1 0 0Table 5.2: This is the statistical analysis of the bank B Dynalog  le of thesame  eld as in Table 5.1.96needs more than 50 ms to reach the designated point and thus triggers aseries of beam-hold o signal. Because the beam is o during such hold-o the discrepancy during that period does not have any clinical signi cance.Note that the  rst entry of these beam hold-o series is recorded, because ofthe lag time between beam hold-o and beam o state.As mentioned in the paragraph above, a discrepancy that is large enoughto trigger a beam hold-o is in most cases triggered at the control points.The number of the beam hold-o series triggered for each leaf is listed in thelast column of each table. In this column the number for each leaf is under10 and the sum of them is up to approximately 30. Noticeably the averagediscrepancy listed in the second column is well under 200 motor counts.The nominal dose rate is 300 MU/min which corresponds to 5 MU/s. Sofor each entry, which corresponds to approximately 50 ms, approximately0.25 MU is given. This is the estimated MU leakage due to the lag in beamresponse, as one entry is the lag time. Considering the number of beamhold-o series triggered may be up to 30, the upper limit of the MU leakageduring a single beam is estimated to be 7.5 MU. The total MU given ina single beam is approximately 200 MU, the worst estimated discrepancycaused by this lag is 3.75 %. In practice, it is observed that when the beamhold-o is removed, there is still a one entry lag before the beam is turnedon. So this e ect is largely cancelled. The 3.75 % is therefore an upper limit.5.1.2 E ective Leaf Gap TestsTable 5.3 summarizes the results of the e ective leaf gap test. The man-ufacturer provides a calculation sheet to calculate the dynamic leaf shiftparameter which is then input into the planning system. The number ob-tained for this measurement is 0.29 mm, which is the same as the numberacquired during the acceptance and obviously the zero di erence is withinthe tolerance of 0.01 mm[6]. This indicates the e ective leaf gap parameteris stable and accurate.97Doserate (MU/min) 300Total Dose 300Field width (cm) 10Leaf speed (cm/min) 20MLC Leakage 0.019367MLC  le Gap size (mm) Measuredionization signal (nC)M3 1.D01 1 1.5598M3 5.D01 5 2.692M3 10.D01 10 4.121M3 20.D01 20 6.953M3 50.D01 50 15.448M3 100.D01 100 29.61Static Field Square Size Measuredionization signal (nC)Open Field 100x100 57.76Closed Field (Leakage) 0x0 1.1186Dynamic leaf shift (mm) 0.29Table 5.3: Dynamic leaf shift measurements985.2 Ion Chamber Measurement5.2.1 CalibrationIn order to establish absolute dose for validation of treatment plans, crosscalibration of the ion chamber in the phantom used to validate the TPScalculation is performed by delivery of a known dose to the phantom. Theion chamber reading per unit dose is then established. The reference dosecondition at VCC is that under a 10 cm by 10 cm  eld at SAD 100 cm, if1 MU is delivered 1 cGy dose will be given at dmax in water. For the crosscalibration, solid water is used and 200 MU is delivered. The ion chamber isset up at depth of 5 cm. (Detailed information is listed in Table 5.4.) UsingPhantom Solid waterSAD 100 cmDepth 5 cmField size 10 cm × 10 cmMillennium MLC Retractedm3 MLC opening O Linac energy 6 MVDose rate 300 MU/minMonitor unit 200MUAve. elec. rdg. -0.5723 nCTable 5.4: Ion chamber cross calibration data measured in solid water phan-tom (Sept. 15 2009)the TMR data for depth of 5 cm, the dose given to the solid water phantomat the isocenter can be calculated in following way.Dose = MU TMR(depth = 5;10 10) (5.1)= 200 0:915 (5.2)= 183(cGy) (5.3)99Phantom Cube phantomSAD 100 cmDepth  9.2 cmField size 4.2 cm × 4.2 cmMillennium MLC Retractedm3 MLC opening 4.2 cm × 4.2 cmLinac energy 6 MVDose rate 300 MU/minMonitor unit 200MUAve. elec. rdg. -0.4310 nCTable 5.5: Ion chamber cross calibration data measured in cube phantom(Sept. 15 2009)Since the electrometer reading acquired is -0.5723 nC, the dose to electrom-eter reading relationship for this ion chamber is determined.Dose=E:Rdg: = 183cGy=( 0:5723nC) = 319:8cGy=nC (5.4)This dose to electrometer reading ratio is essentially based on the assumptionthat the dose delivered at the reference condition is accurate. However, thelinac output does vary about 1 % on a daily basis. Therefore, we cannotguarantee that this dose to electrometer reading ratio is correct in terms ofabsolute dosimetry.During the calibration, the same  eld (Detailed information of the beamsetup is listed in Table 5.5.) is also delivered to the cube phantom. Theelectrometer reading acquired is -0.4310 nC. Using the dose to electrometerreading ratio acquired above, the dose at the isocenter of the cube phantomcan be calculated.Dose =  319:8cGy=nC  0:4310nC (5.5)= 137:8cGy (5.6)Therefore the monitor unit to dose ratio for the cube phantom under this100 eld setup condition is,MU=dose = 200MU=137:8cGy = 1:451MU=cGy (5.7)Compared to the TPS calculated result that by delivering 145 MU, 100 cGywill be given to the isocenter, this measurement agrees with the predictionwithin 0.1 %. This monitor unit to dose relationship holds when the linacoutput is stable. Since the linac output drifts on a daily basis, the MU to doseratio changes accordingly. A consistency veri cation is performed prior tothe dose measurement for clinical cases, during which this MU to dose ratiomay change. However, the purpose of the dose veri cation of these clinicalcases is to compare the measured dose (under the commissioned deliverycondition) with the TPS predicted value. Machine output drift should thusbe excluded. For this purpose, the monitor unit to dose ratio is  xed at thevalue we measured during the cross calibration as 1.451 MU/cGy.5.2.2 Veri cation of IMRT CasesIon chamber dose veri cation of three clinical IMRT cases is described in thissubsection. The detailed information is listed for Case One only and for theother two cases only the  nal results are shown. A cube phantom is usedfor the measurements of clinical cases as well as the dose calibration, anddose calibration of the ion chamber is performed as described in previousparagraphs. In order to test the machine output consistency a square  eldis delivered to the cube phantom.Case OneThe  rst case is based on a previously treated right sided meningioma. Theoriginal treatment is a seven beam stereotactic technique. For the purposeof this study, the case was replanned using stereotactic IMRT techniqueretaining the original seven beam angle arrangement. The treatment plan101was optimized by the iPlan treatment planning software. In the originalplan, a dose of 1.60 cGy per fraction was prescribed and optimized to theisocenter. In order to be consistent with the  lm dosimetry measurement,twice the dose was delivered for dose validation. In other words, 3.20 Gywas delivered to the phantom. Detailed beam parameters for Case One arelisted in Table 5.6. From the result of the previous subsection, the MUPhantom Cube phantomSAD 100 cmMLC name BrainLAB m3Linac energy 6 mVDose rate 300 MU/minLeaf sequencing DynamicTaG optimization onIsocenter dose 3.20 Gyx1 20.0 mmx2 20.0 mmy1 28.0 mmy2 29.0 mmTable Gantry Coll. Depth Dose to MU Ave. elec.Name angle angle angle equiv. isoc. rdg.(mm) (cGy) (nC)Beam 1 90 20 90 91.8 1 34 -0.0021Beam 2 90 100 90 88.1 66 190 -0.1921Beam 3 90 130 90 111.0 54 142 -0.1624Beam 4 55 60 90 120.9 57 166 -0.1678Beam 5 55 95 90 106.0 45 156 -0.1381Beam 6 330 275 90 98.6 38 156 -0.1155Beam 7 305 250 90 112.0 60 204 -0.1687Total 320 1048 -0.9467Table 5.6: Ion chamber measurement of stereotactic IMRT plan (Case One)to dose relationship is measured as 1.451 MU/cGy and is assumed to be xed. During the consistency test on the same day of verifying Case One,102the electrometer reading acquired is -0.3101 nC. Also, because 145 MU isdelivered, the dose given to the isocenter of cube phantom is,Dose = 145MU 1:451MU=cGy = 99:93cGy (5.8)Therefore the dose to electrometer reading relationship for this day is,Dose=E:Rdg: = 99:93cGy= 0:3101nC = 322:3cGy=nC (5.9)Comparing to the value acquired during calibration, which is -319.8cGy/nC,the di erence is 0.78 %. This discrepancy includes machine output variationand is within the tolerance.The total acumulated eletrometer reading for this stereotactic IMRTtreatment plan is -0.9467 nC, the total dose delivered to the isocenter is,Dose = 322:3cGy=nC  0:9467nC = 305:1cGy (5.10)The TPS calculation results showed that the dose delivered to the isocentershould be 311 cGy. So, the discrepancy is -1.9 % for this case. For individual elds, the calculated and measured dose is listed in Table 5.7. Chamber angleis also calculated and listed as the last column of this table. The chamberangle is de ned as the acute angle between the incident beam axis and theion chamber axis. The chamber angle e ect is discussed in 4.3.3.Although a mini ion chamber was used for the dose measurement, theaveraging e ect over a  nite volume could contribute to the discrepancies.To illustrate how signi cant the averaging e ect could be, pro les along threeperpendicular directions of Case One are plotted in Figure 5.1, Figure 5.2and Figure 5.3. O all the  gures the blue vertical line indicates where theion chamber measurement is. The pro les indicate that for two directions,there is no signi cant dose gradient. For the other direction, dose gradientis noticeable. Considering the fact that the stereotactic IMRT is supposed103C. Dose M. Dose Di erenc. ChamberBeam ID (cGy) (cGy) (%) Angle (°)Beam 1 1 0.6 -32.3 70.0Beam 2 62 61.9 -0.1 10.0Beam 3 53 52.3 -1.2 40.0Beam 4 55 54.0 -1.7 44.8Beam 5 44 44.5 1.2 35.3Beam 6 37 37.2 0.6 60.1Beam 7 59 54.3 -7.8 39.7Table 5.7: Dose comparison for individual  elds (Case One)to be used in situations that the target volume has a irregular shape, thepresence of dose gradients around the isocenter is to be expected.Case TwoThe second case is based on a previously treated right trigeminal schwan-noma. This patient was originally treated with an eight beam stereotactictreatment. For the purpose of this study, the case was re-planned usingstereotactic IMRT technique retaining the original beam angle arrangement.Similar to the  rst case, twice the dose is delivered in order to be consistentwith  lm dosimetry measurement. 3.21 Gy is delivered to the isocenter. Theconsistency test indicated that the variation in dose due to ambient condi-tions is 0.41 %, which is within the tolerance level. The cumulative dosemeasured is 318.2 cGy, compared with the TPS calculated result of 321 cGy,the discrepancy is -0.9 %. The comparison for individual  elds is listed inTable 5.8.Case ThreeThe third case is based on a previously treated pituitary adenoma. The pa-tient was previously treated with a seven beam stereotactic treatment. Thecase was re-planned using stereotactic IMRT technique and retaining the104Figure 5.1: Dose pro le in L-R direction of Case OneTable Gantry Coll. Depth Dose to MU Ave. elec.Name angle angle angle equiv. isoc. rdg.(mm) (cGy) (nC)Beam 1 47 50 90 132.3 46 174 -0.1614Beam 2 47 84 90 118.3 23 140 -0.0590Beam 3 47 117 90 130.9 10 124 -0.0365Beam 4 350 222 90 138.2 72 158 -0.2214Beam 5 350 305 90 105.6 21 144 -0.0696Beam 6 310 240 90 129.9 78 174 -0.2373Beam 7 310 310 90 133.5 12 160 -0.0415Beam 8 300 275 90 97.4 66 154 -0.1644Total 321 1228 -0.9911Table 5.8: Ion chamber measurement of stereotactic IMRT plan (Case Two)105Figure 5.2: Dose pro le in F-H direction of Case One106Figure 5.3: Dose pro le in A-P direction of Case One107same beam angle arrangement. 1.66 Gy is prescribed to the isocenter perfraction. During the consistency test, the variation in dose rate due to am-bient conditions is measured to be 0.66 %, which is within the 1 % tolerancelevel. The cumulative dose measured is 167.0 cGy, compared with the TPScalculated result of 166 cGy, the discrepancy is 0.6 %. The comparison forindividual  elds can be found in Table 5.9.Table Gantry Coll. Depth Dose to MU Ave. elec.Name angle angle angle equiv. isoc. rdg.(mm) (cGy) (nC)Beam 1 60 85 90 99.4 25 64 -0.0776Beam 2 45 120 90 137.7 21 70 -0.0641Beam 3 10 75 90 92.6 26 61 -0.0798Beam 4 350 290 90 93.6 25 64 -0.0785Beam 5 270 330 90 100.4 25 57 -0.0777Beam 6 310 240 90 129.9 22 62 -0.0688Beam 7 310 320 90 112.4 24 58 -0.0722Total 166 436 -0.5187Table 5.9: Ion chamber measurement of stereotactic IMRT plan (Case Three)For the three cases measured, the ion chamber measurement and the TPScalculation of the total dose to the isocenter agreed within 1.9 %, which iswithin standard clinical tolerance. However, the discrepancy for individual elds can be quite large. The reasons for this are as yet undetermined.Angular dependence of the ion chamber has been ruled out (See Subsection4.3.3). Noticeably, the third case has signi cantly smaller discrepancy forindividual  elds compared with the second case. The distinguishing propertyof the third case is that the shape of the treated volume is close to a sphereand relatively regular. In Case Three, the individual  elds are not heavilymodulated and the dose from each  eld is similar. This may be an indicationthat the larger discrepancy for the individual  elds in Case Two is related toa high degree of intensity modulation, which leads to a relatively high dosegradient for individual  elds in the vicinity of the measurement point. A108slight shift of the chamber position could result a fairly large dose discrepancyon a  eld-by- eld basis. However, the net e ect could be quite small, whenthe dose from all  elds is taken together[33{35].5.3 Film DosimetryAll  lms were calibrated and analyzed using the techniques described inSection 3.2.4. A  lm calibration was done for each case. The order of thecases is the same as in the ion chamber measurement.5.3.1 Case OneFigure 5.4 shows the TPS calculated dose distribution for this case, withwhich the measured dose distribution is to be compared with. For this casethe calculated dose to the isocenter is 320 cGy. Figure 5.5 shows the isodoseline comparison between the two datasets. In general, the isodose lines showgood agreement except for the 100 % isodose line. The measured  lm dosein the center region is systematically lower than the TPS calculated dose.This is clearly seen in Figure 5.6 which shows the pro le comparison alongthe y-direction. Only the central area shows a noticeable discrepancy whichis roughly 5 %. Figure 5.7 shows the 2D dose di erence map. A cold spotis found in the central area, while some hot spots are found in high dosegradient areas. It is noted that the ion chamber measurement in the centralarea reported -1.9 % lower dose than the TPS calculated result. Figure 5.8shows the results of 2D Gamma test for this case. The 2D Gamma test[36] isa frequently used test incorporating both the distance to agreement and dosedi erence between two dose distributions. Thus the 2D Gamma test workswell in both high dose gradient regions and low dose gradient regions[36].109Figure 5.4: The calculated dose distribution of Case One (Coronal plane)110Figure 5.5: The isodose line comparison of Case One. The solid lines are forthe TPS calculation. The dash lines are for the  lm measurement. (Coronalplane)111Figure 5.6: The y-pro le comparison of Case One. The red line is for theTPS calculation. The green line is for the  lm measurement. (Coronal plane)112Figure 5.7: The 2D dose di erence map of Case One (Coronal plane)113Figure 5.8: The 2D Gamma test with 3 mm and 3 % criteria of Case One(Coronal plane)114Figure 5.9: The calculated dose distribution of Case Two (Coronal plane)115Figure 5.10: The isodose line comparison of Case Two. The solid lines are forthe TPS calculation. The dash lines are for the  lm measurement. (Coronalplane)116Figure 5.11: The y-pro le comparison of Case Two. The red line is for theTPS calculation. The green line is for the  lm measurement. (Coronal plane)117Figure 5.12: The 2D dose di erence map of Case Two (Coronal plan)118Figure 5.13: The 2D Gamma test with 3 mm and 3 % criteria of Case Two(Coronal plane)1195.3.2 Case TwoFigure 5.9 shows the calculated dose distribution for Case Two. The TPScalculated dose to the isocenter is 329 cGy. Figure 5.10 shows the comparisonbetween isodose lines from the TPS and  lm measurement. Figure 5.11 is thepro le comparison along the y-direction. Figure 5.12 is the 2D dose di erencemap. The discrepancy found in the region near the isocenter is about 6 %lower as measured by  lm, compared with the TPS calculated result. Forcomparison the ion chamber measurement was -0.9 % lower than TPS in thisregion. Figure 5.13 shows the 2D Gamma test of this case.5.3.3 Case ThreeFigure 5.14 shows the calculated dose distribution for Case Three. The cal-culated dose to the isocenter is 168 cGy. Figure 5.15 is the isodose line com-parison. Figure 5.16 is the pro le comparison along the y-direction. Figure5.17 is the 2D dose di erence map. The discrepancy found in the isocenterdose is about 3 % lower compared with TPS. In comparison, the dose mea-sured with ion chamber was 0.6 % higher than the TPS calculation. Figure5.18 shows the 2D Gamma test of this case.In general, the  lm dosimetry results for the three cases show the sametrend. The measured doses for the central region are all systematically lowerthan the calculated doses. The discrepancy is up to approximately 6 to 8 %,which is signi cant and beyond our expectation based on ion chamber re-sults. This discrepancy is signi cantly larger than the uncertainty that hadbeen ascribed to the  lm thickness variation described in Chapter 3, Section3.3.4, which is expected to be approximately 2.5 %. The same phantom wasused to verify the dose distribution for the static square  eld. Also the sameprocedure was used to scan the  lm and calibrate the  lm. Thus, theoreti-cally, the additional discrepancy is not caused by either the phantom or thehandling procedure. There is a concern that this additional discrepancy is120Figure 5.14: The calculated dose distribution of Case Three (Coronal plan)121Figure 5.15: The isodose line comparison of Case Three. The solid linesare for the TPS calculation. The dash lines are for the  lm measurement.(Coronal plane)122Figure 5.16: The y-pro le comparison of Case Three. The red line is forthe TPS calculation. The green line is for the  lm measurement. (Coronalplane)123Figure 5.17: The 2D dose di erence map of Case Three (Coronal plan)124Figure 5.18: The 2D GAMMA test with 3 mm and 3 % criteria of Case Three(Coronal plane)125due to the uncertainty of  lm positioning, however, in terms of the threedimensional dose distribution, the accumulated dose is fairly smooth aroundthe isocenter, because the isocenter is within the PTV and the shape of thePTV is not highly irregular. Without obvious high gradient in three dimen-sion, the possibility that a small displacement of the  lm can cause such bigdiscrepancy is very small. The possibility that the discrepancy is due to themiscalculation of the treatment planning software is also unlikely, becausethat the same clinical cases are also veri ed by ion chamber and results con-clude that the discrepancy between the measurement and the calculation isat most 2 %. The ion chamber measurement also ruled out the possibilitythat the treatment plan was not properly delivered due to a malfunction ofthe MLC. Considering that the dose distribution of clinical IMRT cases areaccumulated dose distribution of several  elds incident from di erent angles,the discrepancy may because of di erent response to di erent beam incidentangle or consequently due to di erent energy. Although this is not consis-tent with the designed purpose of the  lm and has never been reported inthe literature, this angular dependence have never been tested in our work.It remains a potential cause. Another possibility is that the radiochromic lm responds di erently for the dynamic delivery, which would indicate adose rate dependence. However, a dose rate dependence for non-real timeapplication of Gafchromic EBT  lm is not reported. But since the dose ratedependence is not tested in our work, we cannot ruled out this possibility. Ingeneral, we have ruled out most of the possible reasons which can cause theadditional discrepancy. Further investigation of the source of discrepancy in lm data is beyond the scope of this work.Based on this analysis, it is concluded that Gafchromic  lm dosimetrygives results with discrepancies which are clinically unacceptable. This isbecause in this project we intended to use the Gafchromic  lm for absolutedosimetry. This is also the reason 2D Gamma tests of all three cases failedin the central region. We concluded that the Gafchromic  lm should only be126used for relative dosimetry at this time.127Chapter 6ConclusionThe  rst objective of this thesis was to evaluate the performance of GafchromicEBT  lm and to establish a relative dosimetry system based on this type of lm. While the apparent ease of use of Gafchromic EBT  lms is appealing,we discovered the following limitations during the course of this work:1. The darkness of the exposed  lm, and hence the Optical Densityincreases with post-exposure time. This increase has been measured, and itis about 2 % of the optical density for the  rst 15 hours post exposure. Thechange in Optical Density is negligible after the  rst 15 hours. Although themagnitude of this change is much smaller when compared with the previousversion of Gafchromic  lm, it is still a factor that needs to be controlled.(See Chapter 3, Section 3.3.1.)2. The image acquired from a uniformly irradiated  lm can show large-scale  uctuations in Optical Density that could be as large as 2.5 %. Thise ect is found to be due to the non-uniformity of the  lm thickness. (SeeChapter 3, Section 3.3.2.)3. Unlike radiographic  lm, the signal from radiochromic  lm is de-pendent on its orientation during the scanning process. All pieces of  lmincluding the calibration  lms should be scanned in the same orientation.4. When  atbed scanners are used, the pixel values acquired from the lm are also dependent on the location of the  lm inside the scanner. Ingeneral, the peripheral region of the scanning area gives lower pixel valuescompared to the central region due to the light scattering e ect and due tothe  nite size of the light source. (See Chapter 3, Section 3.2.3.)In order to determine the overall accuracy of the Gafchromic  lm dosime-128try system, we measured well de ned simple square  elds, with well-knowndose outputs. Several pieces of  lm that were cut from a single piece wereirradiated using the same  eld parameters and irradiation condition to dif-ferent dose levels. The measured doses were compared with the calculateddose. The result indicates the overall uncertainty in dose measurement forthe square  elds were approximately 2.5 %. This was an indication that theintra-piece variation of the  lm response contributes approximately 2.5 % er-ror in the dose reading. Inter-piece variation of the  lm response was testedby comparing three sets of calibration  lms that were irradiated to the samedose levels. The three separate calibration curves obtained from the threesets of  lms were compared. The results show that variation in responsebetween di erent pieces has approximately the same magnitude as within asingle piece. (See Chapter 3, Section 3.3.4.)The second objective of this work was to apply this  lm dosimetry systemtogether with an ion chamber dosimetry technique for dosimetric validation ofa stereotactic IMRT treatment planning system. The following summarizesthe results of this work in this respect:1. Film dosimetry using Gafchromic  lm showed that in the peripheralregion of the irradiated volume, the agreement between the measured doseand the calculated dose is clinically acceptable. However, in the central regionwhere the dose level is above 90 % of the prescribed dose, the measured dosewas signi cantly lower in the three cases we measured. The magnitude of thediscrepancy in the central region for one of the cases was as high as 8 %. Thisis signi cantly larger than the discrepancy observed during measurements ofregular static  elds. We have ruled out a number of potential reasons thatmay cause this e ect. (See discussion in Chapter 5, Section 5.3.3.) Theactual causes of this e ect is yet unclear.2. From the ion chamber measurement, the dose veri cation of threeclinical cases indicated that for a point dose, the calculated dose agreed withthe measured values within 2 %. For individual  elds, the discrepancy could129be much larger than this. This is believed to be due to the presence of highdose gradient around the isocenter for individual  eld. Due to the nature ofIMRT each individual  eld delivers a highly modulated dose but the totaldose from all the  elds is relatively uniform across the irradiated volume.3. The angular dependence of the ion chamber response with respect tothe incident beam was systematically investigated. Results indicated thatthe magnitude of this e ect was less than 1.0 % in most cases, the largestobserved di erence being 1.5 %. This was very relevant to the accuracy ofstereotactic  eld veri cation measurements, as it is often the case that thedirection of the incident beam is not perpendicular to the long axis of theion chamber.130Chapter 7Future Study7.1 Gafchromic EBT2 FilmThe  lm manufacturer acknowledges that the intrinsic variation of the thick-ness of the active layer of the  lm is in most cases the major contributor tothe uncertainty of Gafchromic EBT  lm dosimetry. Thus a new GafchromicEBT2  lm was introduced with the aim of solving this problem. GafchromicEBT2  lm is manufactured by adding a radiation insensitive yellow color dyeto the active layer of the  lm. The attenuation caused by this yellow dyecan be detected primarily in the blue channel and is approximately propor-tional to the active layer thickness of the  lm. Because both the red channelsignal and the blue channel signal are thickness and dose dependent, if thered channel and blue channel signal can be read together, it is, in principle,possible that the variation in the active layer thickness can be corrected for.Although the basic idea is clear, practical implementation of the yellow dyeremains to be done. The manufacturer provides a procedure to make thecorrection, but further investigation is necessary to verify this method beforeit can be used. If the thickness variation can be measured and corrected for,the accuracy of Gafchromic  lm dosimetry may be signi cantly increased,which is desired for dose veri cation purpose.7.2 Monte Carlo SimulationAs mentioned in Chapter 1, Monte Carlo simulation is a useful tool for thepurpose of patient-speci c dose veri cation of IMRT treatment plans calcu-131lated by commercial treatment planning systems. This is because the MonteCarlo method calculates the dose from  rst principles, unlike model-baseddose calculation algorithms used in commercial treatment planning systems.For this reason, Monte Carlo simulation has the potential to be the most ac-curate dose calculation method. Compared with other simpli ed algorithms,Monte Carlo simulation gives much more accurate results in situations wheresigni cant electron disequilibrium exists, for example the presence of air cav-ities or rapidly changing tissue density.Compared with measurement-based veri cation methods, the Monte Carlosimulation method has the following advantages: it produces a three di-mensional dose distribution with high resolution and it is much less labor-intensive.At the Vancouver Cancer Center, Monte Carlo simulation is clinicallyused for conventional IMRT treatment plan veri cation. The calculation iscarried out in three main steps. The program modules used for the simula-tion are: BEAMnrc[37], the Virginia code (a simpli ed MLC simulation codereferred here as the Virginia Code as the author is from Medical College ofVirginia Hospitals)[38] and DOSXYZnrc[39]. Among them, BEAMnrc andDOSXYZnrc are both EGSnrc[40] based general purpose Monte Carlo simu-lation programs. BEAMnrc is designed for simulating radiotherapy sourcesbased on the modeling of coupled electron and photon transport. More specif-ically, BEAMnrc is used for simulating particle interactions inside linac head.The simulation starts from electrons exiting the accelerator vacuum exit win-dow and stops before the particles enter a scoring plane which is located abovethe MLC. Electron and photon interactions with all components within thelinac gantry head are simulated. When the simulation is completed, informa-tion including each particle’s position, velocity and kinetic energy is saved toa  le called a phase space  le. The simulation of the MLC is carried out withthe Virginia code. BEAMnrc itself is also capable of simulating the MLC,however, because conventional IMRT treatments are dynamically delivered,132a simulation of the dynamic MLC requires extensive calculation and is thustime consuming. The Virginia code, simpli es the simulation. For photonsit only considers attenuation and  rst Compton scattering. Pair productionand electron interactions (scattering and Bremsstrahlung) within the MLCare ignored. The simpli ed modeling signi cantly increases the calculationspeed, while tests show that the accuracy is not signi cantly reduced. Thesimulation continues from the phase space  le created by BEAMnrc in the rst step. The information about the MLC position is (rather than extractedfrom the treatment plan  le) imported from a so-called dynamic log  le fromthe MLC control console, which is a record of the actual MLC performancemonitored by build-in sensors (See Chapter 4, Section 4.3.2 for detailed de-scriptions.) The result of the MLC simulation is saved to another phasespace  le in the same format. For the last step, the dose distribution is cal-culated by DOSXYZnrc, which is a general purpose Monte Carlo simulationprogram for dose distribution calculation. The  nal three dimensional dosedistribution is exported in DICOM format, which is then imported to VarianEclipse software for comparison with the TPS calculation. The simulation isperformed using a 30-node dedicated cluster of 64-bit 2 GHz AMD Opteronprocessors. The typical calculation time for a single case is approximately2 hours.To modify the current Monte Carlo simulation so that it can be appliedfor dose veri cation of the stereotactic IMRT technique, the code used inthe second step of the Monte Carlo simulation needs to be modi ed. Thisis because the BrainLAB m3 micro MLC is used in the stereotactic IMRTbeam delivery and the physical dimensions of this MLC is di erent from theVarian Millennium MLC. The Virginia code should be modi ed to representthe m3 micro MLC. Noticeably the di erence between the m3 MLC andVarian Millennium MLC is not only the size but also the shape and detailedstructure of the leaves.If the Monte Carlo system can be successfully modi ed and used for133stereotactic IMRT treatment veri cation, it will be an excellent tool for eval-uating dose distributions, considering the accuracy of the current GafchromicEBT  lm dosimetry method is not high enough for clinical application.134BibliographyBibliography[1] K. Otto, \Volumetric modulated arc therapy: IMRT in a single gantryarc," Medical Physics, vol. 35, pp. 310{317, January 2008.[2] B. G. Clark, T. Teke, and K. Otto, \Penumbra evaluation of the VarianMillennium and BrainLAB m3 multileaf collimators," Radiation Oncol-ogy, vol. 66, pp. S71{S75, Nov 2005.[3] M. McJury, M. Oldham, V. P. Cosgrove, P. S. Murphy, S. Doran, M. O.Leach, and S. Webb, \Radiation dosimetry using polymer gels: methodsand applications," The British Journal of Radiology, vol. 73, no. 2000,pp. 919{929, 2000.[4] T. LoSasso, C. S. Chui, and C. C. Ling, \Comprehensive quality assur-ance for the delivery of intensity modulated radiotherapy with a mul-tileaf collimator used in the dynamic mode," Medical Physics, vol. 28,pp. 2209{2219, November 2001.[5] A. Leal, F. Snchez-Doblado, R. Arrns, J. Rosell, E. C. Pavn, and J. I.Lagares, \Routine IMRT veri cation by means of an automated MonteCarlo simulation system," International Journal of Radiation OncologyBiology Physics, vol. 56, pp. 58{68, May 2003.[6] T. Bortfeld, W. Schlegel, K. H. Hover, and D. Schulz-Ertner, \Miniand Micro Multileaf Collimators," tech. rep., German Cancer ResearchCenter, 1999.135[7] S. Benedict, R. Cardinale, Q. Wu, R. Zwicker, W. Broaddus, andR. Mohan, \Intensity-modulated stereotactic radiosurgery using dy-namic micro-multileaf collimation," International Journal of RadiationOncology Biology Physics, vol. 50, pp. 751{758, July 2001.[8] S. Espinoza, M. Saboori, S. Forman, C. R. Moorthy, and D. L. Ben-zil, \Use of stereotactic intensity-modulated radiotherapy in thyroid-related ophthalmopathy," Journal of Neurosurgery, vol. 101, pp. 396{401, November 2004.[9] B. C. Ferreira, M. C. Lopes, and M. Capela, World Congress on Medi-cal Physics and Biomedical Engineering, vol. 25/1. Munich, Germany:Springer Berlin Heidelberg, September 2009.[10] B. Clark, C. Candish, E. Vollans, E. Gete, R. Lee, M. Martin, R. Ma,and M. McKenzie, \Optimization of stereotactic radiotherapy treatmentdelivery technique for base-of-skull meningiomas," Medical Dosimetry,vol. 33, pp. 239{247, April 2008.[11] F. H. Attix, Introduction to Radiological Physics and Radiation Dosime-try. Wiley-Interscience, 1 edition ed., September 1986.[12] P. Mayles, A. E. Nahum, and J.-C. Rosenwald, Handbook of Radiother-apy Physics. Taylor and Francis,  rst ed., 2007.[13] H. E. Johns and J. R. Cunningham, Physics of Radiology. Charles C.Thomas Publisher, 4th ed., February 1983.[14] D. I. Thwaites and J. B. Tuohy, \Back to the future : the history anddevelopment of the clinical linear accelerator," Physics in medicine andbiology, vol. 51, no. 13, pp. R343{R362, 2006.[15] P. Metcalfe, T. Kron, and P. Hoban, The Physics of Radiotherapy X-Rays from Linear Accelerators. Medical Physics Publishing Corporation,1st ed., June 1997.136[16] E. B. Podgorsak, \Stereotactic Irradiation: Historical Perspective andCurrent Trend," Journal of Medical Physics, vol. 26, no. 4, pp. 298{316,2001.[17] D. Verellen, N. Linthout, A. Bel, G. Soete, D. V. den Berge, J. D.Haens, and G. Storme, \Assessment of the uncertainties in dose deliv-ery of a commercial system for linac-based stereotactic radiosurgery,"International Journal of Radiation Oncology Biology Physics, vol. 44,pp. 421{433, May 1999.[18] Y. C. Lo, C. Ling, and D. A. Larson, \The e ect of setup uncertaintieson the radiobiological advantage of fractionation in stereotaxic radio-therapy," International Journal of Radiation Oncology Biology Physics,vol. 34, pp. 1113{1119, March 1996.[19] T. Bortfeld, \IMRT: a review and preview," Physics in Medicine andBiology, pp. R363{R379, June 2006.[20] A. Brahme, J. E. Roos, and I. Lax, \Solution of an integral equation en-countered in rotation therapy," Physics in medicine and biology, vol. 27,pp. 1221{1229, 1982.[21] A. Niroomand-Rad, C. R. Blackwell, B. M. Coursey, K. P. Gall, J. M.Galvin, W. L. McLaughlin, A. S. Meigooni, R. Nath, J. E. Rodgers, andC. G. Soares, \AAPM Report No.63 Radiochromic Film Dosimetry,"tech. rep., AAPM Radiation Therapy Committee Task Group No. 55,Dec 1998.[22] International Specialty Products Inc., E ects of Light Scattering byFilms on the Performance of CCD Scanners.[23] International Specialty Products Inc., Suggested Procedure QAI 366Convert the scanner-space values to absorbed dose values, rev 0.3 ed.137[24] S. Saur and J. Frengen, \GafChromic EBT  lm dosimetry with  atbedCCD scanner: A novel background correction method and full dose un-certainty analysis," Medical Physics, vol. 35, pp. 3094{3101, July 2008.[25] B. C. Ferreira, M. C. Lopes, and M. Capela, \Evaluation of an Epson atbed scanner to read Gafchromic EBT  lms for radiation dosimetry,"Physics in Medicine and Biology, vol. 54, pp. 1073{1085, Jan 2009.[26] L. J. van Battum, D. Ho mans, H. Piersma, and S. Heukelom, \Accuratedosimetry with GafChromic EBT  lm of a 6 MV photon beam in water:What level is achievable?," Medical Physics, vol. 35, Feb 2008.[27] O. A. Zeidan, S. A. L. Stephenson, S. L. Meeks, T. H. Wagner, T. R.Willoughby, P. A. Kupelian, and K. M. Langen, \Characterization anduse of EBT radiochromic  lm for IMRT dose veri cation," MedicalPhysics, vol. 33, pp. 4064{4072, 2006.[28] B. D. Lynch, J. Kozelka, M. K. Ranade, J. G. Li, W. E. Simon, andJ. F. Dempsey, \Important considerations for radiochromic  lm dosime-try with  atbed CCD scanners and EBT GafChromic  lm," MedicalPhysics, vol. 33, pp. 4551{4556, 2006.[29] I. S. P. Inc., \EBT Whitepaper," tech. rep., International SpecialtyProducts Inc., August 2007.[30] I. S. P. Inc., \GafChromic Technical Brief," tech. rep., InternationalSpecialty Products Inc., February 2009.[31] A. Rink, A. Vitkin, I, and D. Ja ray, A, \Intra-irradiation changes inthe signal of polymer-based dosimeter (GAFCHROMIC EBT) due todose rate variations," Phys. Med. Biol., vol. 52, pp. N523{N529, 2007.[32] \AAPM Report No. 54 Stereotactic Radiosurgery," tech. rep., AmericanInstitute of Physics Task Group 42, June 1995.138[33] R. K. Rice, J. J. Hansen, G. K. Svensson, and R. L. Siddon, \Measure-ments of dose distributions in small beams of 6 MV X-rays," Physics inMedicine and Biology, vol. 32, pp. 1087{1099, September 1987.[34] W. U. Laub and T. Wong, \The volume e ect of detectors in the dosime-try of small  elds used in IMRT," Medical Physics, vol. 30, pp. 341{347,March 2003.[35] E. Pappas, T. G. Maris, A. Papadakis, F. Zacharopoulou, J. Damilakis,N. Papanikolaou, and N. Gourtsoyiannis, \Experimental determinationof the e ect of detector size on pro le measurements in narrow photonbeams," Medical Physics, vol. 33, pp. 3700{3710, October 2006.[36] D. A. Low and J. F. Dempsey, \Evaluation of the gamma dose distribu-tion comparison method," Med. Phys., vol. 30, pp. 2455{2464, Septem-ber 2003.[37] D. Rogers, B. Walters, and I. Kawrakow, \BEAMnrc Users Manual,"tech. rep., Ionizing Radiation Standards, National Research Council ofCanada, July 2009.[38] J. V. Siebers, P. J. Keall, J. O. Kim, and R. Mohan, \A method for pho-ton beam Monte Carlo multileaf collimator particle transport," Physicsin Medicine and Biology, vol. 47, no. 17, pp. 3225{3249, 2002.[39] B. Walters, I. Kawrakow, and D. Rogers, \DOSXYZnrc Users Manual,"tech. rep., Ionizing Radiation Standards, National Research Council ofCanada, July 2009.[40] I. Kawrakow, E. Mainegra-Hing, D. Rogers, F. Tessier, and B. Walters,\The EGSnrc Code System: Monte Carlo Simulation of Election andPhoton Transport," tech. rep., Ionizing Radiation Standards, NationalResearch Council of Canada, Carleton University, Feb 2010.139Appendix ASource CodeA.1 Matlab Source Code Used to Adjustthe Resolution of scanned images%This progam i s used to reduce the r e s o l u t i o n%Adjust the parameters i f necessary[ fname pname]= uigetfile ( ’ ∗. t i f ’ , ’ S e l e c t the 300 dpiinput TIFF f i l e ’ ) ;%Change the r e s o l u t i o n i f neededtg=imread ( [ pname fname ] ) ;[ y x z]= size ( tg ) ;xx=floor ( x/3) ;%Reduced 3 timesyy=floor ( y/3) ;%Reduced 3 timesoutput=uint16 ( zeros ( yy , xx , z ) ) ;for i =1:yyfor j =1:xxfor k=1:zave=uint16 (median(median( double ( tg ( ( ( i 1)∗3+1) : i ∗3 ,(( j 1)∗3+1) : j ∗3 , k ) ) ) ) ) ;output ( i , j , k )=ave ;endendend140wname=[ ’ 100 dpi ’ fname ] ;%Change the r e s o l u t i o n i fneededimwrite ( output , [ pname wname ] ) ;A.2 Matlab Source Code Used to Calibratethe Film and Convert the DoseDistribution into Omni-pro CompatibleFormat%Use 6 6x6cm square f i e l d to generate c a l i b r a t i o n curve%Apply on a measuremtn f i l m and convert i t i n t o Omni pro compatible format%P o t i e n t i a l m o n i f i c a t i o n i n c l u e d e s r e s u l o t i o n and headinformation%This i s a corronal plane v e r s i o n and used raw p i x e lv a l u e to c a l i b r a t esq rd=imread ( ’ 100 dpi c2 . t i f ’ ) ;sq ave=mean(mean( sq rd (102:134 ,102:134 ,1) ) ) ;netod=sq ave ;dose =0/145;sq rd=imread ( ’ 100 dpi a1 . t i f ’ ) ;sq ave=mean(mean( sq rd (102:134 ,102:134 ,1) ) ) ;netod =[ netod sq ave ] ;dose =[ dose 50/145];141sq rd=imread ( ’ 100 dpi a2 . t i f ’ ) ;sq ave=mean(mean( sq rd (102:134 ,102:134 ,1) ) ) ;netod =[ netod sq ave ] ;dose =[ dose 100/145];sq rd=imread ( ’ 100 dpi a3 . t i f ’ ) ;sq ave=mean(mean( sq rd (102:134 ,102:134 ,1) ) ) ;netod =[ netod sq ave ] ;dose =[ dose 150/145];sq rd=imread ( ’ 100 dpi b1 . t i f ’ ) ;sq ave=mean(mean( sq rd (102:134 ,102:134 ,1) ) ) ;netod =[ netod sq ave ] ;dose =[ dose 200/145];sq rd=imread ( ’ 100 dpi b2 . t i f ’ ) ;sq ave=mean(mean( sq rd (102:134 ,102:134 ,1) ) ) ;netod =[ netod sq ave ] ;dose =[ dose 250/145];sq rd=imread ( ’ 100 dpi b3 . t i f ’ ) ;sq ave=mean(mean( sq rd (102:134 ,102:134 ,1) ) ) ;netod =[ netod sq ave ] ;dose =[ dose 300/145];sq rd=imread ( ’ 100 dpi c1 . t i f ’ ) ;sq ave=mean(mean( sq rd (102:134 ,102:134 ,1) ) ) ;netod =[ netod sq ave ] ;dose =[ dose 350/145];142sq rd=imread ( ’ 100 dpi c3 . t i f ’ ) ;sq ave=mean(mean( sq rd (102:134 ,102:134 ,1) ) ) ;netod =[ netod sq ave ] ;dose =[ dose 450/145];sq rd=imread ( ’ 100 dpi d1 . t i f ’ ) ;sq ave=mean(mean( sq rd (102:134 ,102:134 ,1) ) ) ;netod =[ netod sq ave ] ;dose =[ dose 550/145];sq rd=imread ( ’ 100 dpi d2 . t i f ’ ) ;sq ave=mean(mean( sq rd (102:134 ,102:134 ,1) ) ) ;netod =[ netod sq ave ] ;dose =[ dose 700/145];sq rd=imread ( ’ 100 dpi d3 . t i f ’ ) ;sq ave=mean(mean( sq rd (102:134 ,102:134 ,1) ) ) ;netod =[ netod sq ave ] ;dose =[ dose 800/145];netod2=netod ;c o e f=polyfit ( netod2 , dose , 4 ) ;n e t o d h i r e s=netod2 (1) : 0.01: netod2 (12) ;plot ( dose , netod2 , polyval ( coef , n e t o d h i r e s ) , n e t o d h i r e s );xlabel ( ’ Dose (Gy) ’ ) ;143ylabel ( ’ Net Optical Density ’ ) ;[ cvfname , cvpname ] = uigetfile ( ’ ∗. t i f ’ , ’ S e l e c t the TO BECONVERTED f i l m ’ ) ;cv=imread ( [ cvpname cvfname ] ) ;cv netod=double ( cv ( : , : , 1 ) ) ;cv dose=polyval ( coef , cv netod ) ;cv dose3=flipud ( cv dose ) ’ ;cv dose2=wiener2 ( cv dose3 , [ 5 5 ] ) ;cvfname aux=cvfname ( 1 : length ( cvfname ) 4) ;fidA=fopen ( [ [ cvpname cvfname aux ] ’ . opg ’ ] , ’wt ’ ) ;% ∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗% Part 1 Create the head ( Modify i f needed )% ∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗fprintf ( fidA , ’<opimrtascii >nn ’ ) ;fprintf ( fidA , ’nn ’ ) ;fprintf ( fidA , ’<a s ci i h e a de r >nn ’ ) ;fprintf ( fidA , ’ F i l e Version : 3nn ’ ) ;fprintf ( fidA , ’ Separator : " ,"nn ’ ) ;fprintf ( fidA , ’ Workspace Name: nn ’ ) ;fprintf ( fidA , ’ F i l e Name: nn ’ ) ;fprintf ( fidA , ’ Image Name: %snn ’ , cvfname ) ;fprintf ( fidA , ’ Radiation Type : Photonsnn ’ ) ;fprintf ( fidA , ’ Energy : 6.0 MVnn ’ ) ;144fprintf ( fidA , ’SSD : nn ’ ) ;fprintf ( fidA , ’SID : 100.0 cmnn ’ ) ;fprintf ( fidA , ’ Field Size Cr : nn ’ ) ;fprintf ( fidA , ’ Field Size In : nn ’ ) ;fprintf ( fidA , ’ Data Type : Abs . Dosenn ’ ) ;fprintf ( fidA , ’ Data Factor : 1.000nn ’ ) ;fprintf ( fidA , ’ Data Unit : mGynn ’ ) ;fprintf ( fidA , ’ Length Unit : cmnn ’ ) ;fprintf ( fidA , ’ Plane : XYnn ’ ) ;%Important! ! ! ! ! ![ x , y]= size ( cv dose2 ) ;xres =0.0254;%=100 dpiyres =0.0254;%=100 dpid e l t a x=xres∗x ;d e l t a y=yres∗y ;s t a r t x = 8.5;s t a r t y = 8.5;%∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗% Changing no . rows and columns%∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗fprintf ( fidA , ’No . of Columns : %dnn ’ , x ) ;fprintf ( fidA , ’No . of Rows : %dnn ’ , y ) ;% ∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗145% Part 2 Create the head ( Modify i f needed )% ∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗∗fprintf ( fidA , ’Number of Bodies : 1nn ’ ) ;fprintf ( fidA , ’ Operators Note : Corr . : Unif . Calib . ,Bkgnd , Calib Output ,Temp. , Press . ,nn ’ ) ;fprintf ( fidA , ’</a sc i i h e ad e r >nn ’ ) ;fprintf ( fidA , ’nn ’ ) ;fprintf ( fidA , ’<asciibody>nn ’ ) ;fprintf ( fidA , ’ Plane Position : 0.0 cmnn ’ ) ;fprintf ( fidA , ’X[cm] , ’ ) ;%Important ! ! ! ! ! !for i =1:xfprintf ( fidA , ’ %.3 f , ’ , s t a r t x +(i 1)∗xres ) ;endfprintf ( fidA , ’nnY[cm]nn ’ ) ;%Important ! ! ! ! ! !for j =1:yfprintf ( fidA , ’ %.3 f , ’ , s t a r t y +(j 1)∗yres ) ;for i =1:xfprintf ( fidA , ’ %d , ’ , uint16 ( cv dose2 ( i , j ) ∗1000)) ;endfprintf ( fidA , ’nn ’ ) ;endfprintf ( fidA , ’</asciibody>nnnnnn ’ ) ;fprintf ( fidA , ’</opimrtascii >nn ’ ) ;fclose ( fidA ) ;146

Cite

Citation Scheme:

    

Usage Statistics

Country Views Downloads
China 24 11
United States 6 25
France 3 0
Brazil 2 0
Germany 1 0
Luxembourg 1 0
India 1 0
City Views Downloads
Beijing 22 0
Unknown 7 1
Guangzhou 2 0
Ashburn 2 0
Redmond 2 0
Framingham 1 0
Sunnyvale 1 0
Bangalore 1 0

{[{ mDataHeader[type] }]} {[{ month[type] }]} {[{ tData[type] }]}
Download Stats

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0069932/manifest

Comment

Related Items