UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Investigation of Parkinson's disease related pattern and altered dopamine release pattern in treatment-induced… Fu, Jessie FangLu 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2016_september_fu_fanglujessie.pdf [ 2.15MB ]
JSON: 24-1.0308595.json
JSON-LD: 24-1.0308595-ld.json
RDF/XML (Pretty): 24-1.0308595-rdf.xml
RDF/JSON: 24-1.0308595-rdf.json
Turtle: 24-1.0308595-turtle.txt
N-Triples: 24-1.0308595-rdf-ntriples.txt
Original Record: 24-1.0308595-source.json
Full Text

Full Text

Investigation of SerotonergicParkinson’s Disease Related Patternand Altered Dopamine ReleasePattern in Treatment-InducedComplications and Non-MotorSymptoms of Parkinson’s DiseasebyJessie FangLu FuB.Sc., The University of British Columbia, 2014A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinThe Faculty of Graduate and Postdoctoral Studies(Physics)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)August 2016c© Jessie FangLu Fu 2016AbstractParkinson’s Disease is the second most common neurodegenerative disorder. Apartfrom motor symptoms, cognitive deficits are also common. Treatments, mainly inthe form of dopamine (DA) replacement therapy, although reduce motor symptomsat first, can lead to treatment-induced complications. Abnormal spatial covariancemetabolic pattern linked to the motor and cognitive symptoms of Parkinson’s Dis-ease (PD) have previously been defined using Fludeoxyglucose Positron EmissionTomography (PET). In contrast, little is known about the functional networks in theserotonergic system, which is known to be closely related to cognitive dysfunctionsof the disease.In this thesis work, we want to investigate the interactions between the dopamin-ergic and serotonergic pathways in presymptomatic and early stages of the disease,and their contributions to treatment-induced complications and non-motor symp-toms in PD subjects.In the first part of this project, we investigated the PD and LRRK2 mutation re-lated patterns in the serotonergic system by studying 12 asymptomatic LRRK2 mu-tation carriers (LRRK2-AMC), 9 healthy controls (HC), and 18 PD subjects using[11C]-3-amino-4-(2-dimethylaminomethylphenylsulfanyl)-benzonitrile (DASB) PETand a principal component analysis (PCA) based regional covariance model withbootstrap resampling. The serotonergic PD-related pattern (SPDRP) significantlyseparated PD subjects from HC subjects (p< 0.0001). A distinct asymptomaticLRRK2 mutation-related pattern (LRRK2-AMRP) significantly separated LRRK2-AMC with HC subjects (p< 0.0001).In the second part of the project, we analyzed the medication-induced DA releasepattern for 10 early PD subjects using double [11C]-Raclopride scans. We found asignificant negative correlation between DA release and age of onset in the striatum.These findings, although obtained with a small number of subjects, suggest thatthe serotonergic system may be affected by PD in a specific pattern and regionsrelatively preserved binding may contribute to cognitive dysfunctions related toiiAbstractPD. LRRK2-AMC subjects showed a distinct pattern, which indicates that eithersuch increase is of compensatory nature or is a characteristic of this specific mu-tation. The combination of abnormal medication-induced DA release pattern andupregulation of the serotonergic system may be able to explain the occurrence oftreatment-induced complications and non-motor symptoms in PD patients, and actas a potential early marker for the disease.iiiPrefacePart of this thesis work (Chapter 4) was presented at 11th International Sympo-sium on Functional NeuroReceptor Mapping of the Living Brain. Full reference canbe found at J.Fu, N.Vafai, E.Shahinfard, N.Heffernan, J.McKenzie, R.Mabrouk,I.Klyuzhin, A.J.Stoessl, V.Sossi, 2016, Investigation of Parkinsons Disease RelatedCovariance Pattern in the Serotonergic System using [11C]-DASB/PET, 11th In-ternational Symposium on Functional NeuroReceptor Mapping of the Living Brain,Boston, USA, July 13-16.This study was approved by UBC Research Human Ethics Board, in particularthe Clinical Research Ethics Board, under ’The Evolution of PD’ (certificate number:H12-00843).ivTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Positron Emission Tomography . . . . . . . . . . . . . . . . . . . . . 11.1.1 Radiotracer . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.1.2 Radioisotope Decay . . . . . . . . . . . . . . . . . . . . . . . 41.1.3 Detection System . . . . . . . . . . . . . . . . . . . . . . . . 61.1.4 Image Reconstruction . . . . . . . . . . . . . . . . . . . . . . 111.2 Kinetic Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121.2.1 Time Activity Curves . . . . . . . . . . . . . . . . . . . . . . 131.2.2 Reference Tissue Model . . . . . . . . . . . . . . . . . . . . . 131.2.3 Simplified Reference Tissue Model . . . . . . . . . . . . . . . 171.2.4 Two-Step Simplified Reference Tissue Model . . . . . . . . . 181.2.5 Logan Plot Method . . . . . . . . . . . . . . . . . . . . . . . 181.3 Network Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191.3.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191.3.2 Principal Component Analysis . . . . . . . . . . . . . . . . . 201.3.3 SSM/PCA Analysis . . . . . . . . . . . . . . . . . . . . . . . 201.4 Parkinson’s Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . 241.4.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . 241.4.2 Symptoms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25vTable of Contents1.4.3 Dopaminergic System . . . . . . . . . . . . . . . . . . . . . . 261.4.4 DA release . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281.4.5 Serotonergic System . . . . . . . . . . . . . . . . . . . . . . . 301.5 LRRK2 Mutation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311.6 Research Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . 322 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342.1 Subjects and Clinical Information . . . . . . . . . . . . . . . . . . . 342.2 Scanning Protocol and Image Processing . . . . . . . . . . . . . . . 352.2.1 Scanning Protocol . . . . . . . . . . . . . . . . . . . . . . . . 352.2.2 Image Processing . . . . . . . . . . . . . . . . . . . . . . . . 353 DASB Parametric Validation . . . . . . . . . . . . . . . . . . . . . . 373.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373.2.1 k2’ Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373.2.2 BPND Values . . . . . . . . . . . . . . . . . . . . . . . . . . 383.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384 SSM Pattern Analysis in Serotonergic System . . . . . . . . . . . 414.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 414.1.1 Pattern Identification . . . . . . . . . . . . . . . . . . . . . . 414.1.2 Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434.2.1 Absolute BPND Values . . . . . . . . . . . . . . . . . . . . . 434.2.2 PD vs HC . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444.2.3 Asymptomatic LRRK2 Mutation Carrier vs HC . . . . . . . 454.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494.3.1 Use of PCA to Identify Disease/Mutation-Related Topogra-phies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494.3.2 ROI-based Network Analysis . . . . . . . . . . . . . . . . . . 494.3.3 PC1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 504.3.4 Pattern Identification and Validation . . . . . . . . . . . . . 504.3.5 Comparison with Other Network Analysis . . . . . . . . . . 504.3.6 Possible Functional Basis for SPDRP and LRRK2-AMRP To-pography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51viTable of Contents4.3.7 Correlation with Clinical Measurements . . . . . . . . . . . . 535 Levodopa-Induced Dopamine Release in PD . . . . . . . . . . . . . 545.1 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545.2.1 Subjects and Quantitative Measurements . . . . . . . . . . . 545.2.2 Regression Models . . . . . . . . . . . . . . . . . . . . . . . . 555.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565.3.1 Correlations in Striatum Regions . . . . . . . . . . . . . . . 565.3.2 Disease Severity . . . . . . . . . . . . . . . . . . . . . . . . . 585.3.3 Correlation with Other Tracers . . . . . . . . . . . . . . . . . 595.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 595.4.1 Correlations with Disease Severity . . . . . . . . . . . . . . . 595.4.2 DA Release Correlation with Age of Onset and Change inUPDRS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 595.4.3 Relationship with Serotonergic System . . . . . . . . . . . . 606 Conclusions and Future Work . . . . . . . . . . . . . . . . . . . . . . 61Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62viiList of Tables1.1 Comparison between commonly used imaging modalities. [1] . . . . 21.2 Characteristics of commonly used PET radioisotopes. The range inwater is the total distance traveled by the positron before it annihi-lates with an electron. [2] . . . . . . . . . . . . . . . . . . . . . . . . 31.3 Symbols used in the reference tissue model, as shown in Figure1.6.[3][4] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162.1 Participants clinical information: values reported as mean±standarddeviation; Disease duration has been accounted from the time ofPD motor symptoms initiation (not from time of clinical diagnosis);M=Males; F=Females; UPDRS=Unified Parkinson’s Disease Rat-ing Scale; UPDRS off=UPDRS without drug (LDOPA) intervention;UPDRS on=UPDRS when on drug medication. . . . . . . . . . . . 345.1 Clinical information for 10 early sporadic PD subjects . . . . . . . . 55viiiList of Figures1.1 Nucleus decays to emit positron (β+) which travels to the site ofannihilation where it annihilates with an electron (e−) producing two511 keV gamma rays (γ) in opposite directions. r is the displacementfrom the parent nucleus to the site of annihilation, whereas p is theactual path of β+ [4] . . . . . . . . . . . . . . . . . . . . . . . . . . . 51.2 A schematic of PET block detectors which is based on the first com-mercial human PET scanner built in 1974. The scanner was made by48 NaI(T1) detectors. [4][5] . . . . . . . . . . . . . . . . . . . . . . . 61.3 a) two annihilation: X is detected by detector elements D3 and D67along line of response (LOR), Y is undetected because photon pathdoes not interact with detector ring. b) top view of annihilation of X[6] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81.4 True coincidence detection vs scatter and random coincidence detec-tion in PET [7] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81.5 Time activity curve for [11C]-DASB tracer with SRTM model fit inthe left caudate region. The estimated parameters from SRTM areshown. Error bars are estimated from the scanner count-rates [4] . . 141.6 Systematic diagram of the reference tissue model [8]. Using a refer-ence region, parameter Cp can be eliminated. . . . . . . . . . . . . . 151.7 Overflow of the Scaled Subprofile Model. . . . . . . . . . . . . . . . 231.8 Chemical structure of dopamine molecule . . . . . . . . . . . . . . . 261.9 Neuronal projections of four DA systems in human brain [9] . . . . . 271.10 Systematic diagram of DA synapse showing the production, release,reuptake, break down and diffusion of DA. [10] . . . . . . . . . . . . 291.11 Chemical structure of serotonin . . . . . . . . . . . . . . . . . . . . . 311.12 Serotonin projections in the brain. Serotonin is produced in nucleusraphe and projects onto other brain regions [11] . . . . . . . . . . . . 32ixList of Figures3.1 estimated k2’ values for each subject compared to reported k2’ valueas shown with red line . . . . . . . . . . . . . . . . . . . . . . . . . . 383.2 Averaged BPND values from SRTM and SRTM2 in each ROI . . . 393.3 Averaged BPND values from SRTM and SRTM2 in each subject . . 403.4 Correlation between SRTM and SRTM2 BPND values in each ROI.Equation of the linear regression model is shown in the Figure . . . 404.1 Frequency Histogram of the included PCs. Only 7 PCs account forgreater than 5% of total variance in the data. The first 3 PCs withthe highest frequency were entered into the logistic regression modelto obtain the combined disease-related pattern. . . . . . . . . . . . . 424.2 Frequency Histogram of the model parameters of the included PCs inthe logistic regression model. The mean of each frequency histogramwas used as the coefficient for each corresponding PC in the logisticregression model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434.3 Serotonergic Parkinson’s disease-related pattern (SPDRP) identifiedby spatial covariance analysis of DASB PET scans from 17 PD pa-tients. This pattern was characterized by a relative decreased SERTbinding in caudate and putamen, and a covarying increased SERTbinding in hypothalamus, hippocampus, anterior cingulate (ACC),amygdala, and medulla. Only regions that significantly contributedto the network at Z>1. Regions with positive weights (increasedbinding) are colour-coded red; those with negative weights (decreasedbinding) are colour-coded blue. . . . . . . . . . . . . . . . . . . . . . 444.4 SPDRP expression in HC, PD and LRRK2-AMC subjects. Therewas a significant separation between HC (blue) and PD (red) groups(p¡0.0001). LRRK2-AMC subjects did not show an elevation of thisdisease-specific pattern. . . . . . . . . . . . . . . . . . . . . . . . . . 454.5 SPDRP expression vs age of disease onset in PD subjects. SPDRPexpression in PD subjects showed an almost significant correlationwith age of onset (p=0.0579). . . . . . . . . . . . . . . . . . . . . . . 464.6 SPDRP expression vs UPDRS motor score in PD subjects. SPDRPexpression in PD subjects showed an almost significant correlationwith UPDRS (p=0.0558). . . . . . . . . . . . . . . . . . . . . . . . . 46xList of Figures4.7 Serotonergic asymptomatic LRRK2 mutation-related spatial covari-ance pattern (LRRK2-AMRP) identified by spatial covariance analy-sis of DASB PET scans from 9 LRRK2-AMC subjects. This patternwas characterized by a relative decreased SERT binding in putamen,and a covarying increased SERT binding in hypothalamus, amygdala,midbrain, PPN, SN and medulla. Only regions that significantly con-tributed to the network at Z>1. Regions with positive weights (in-creased binding) are colour-coded red; those with negative weights(decreased binding) are colour-coded blue. . . . . . . . . . . . . . . . 474.8 LRRK2-AMRP expression in HC, PD and LRRK2-AMC subjects.There was a significant separation between HC (blue) and PD (red)groups (p¡0.0001). LRRK2-AMC subjects did not show an elevationof this disease-specific pattern. . . . . . . . . . . . . . . . . . . . . . 474.9 ROC for training set for 1000 iterations. . . . . . . . . . . . . . . . . 484.10 ROC for testing set for 1000 iterations . . . . . . . . . . . . . . . . . 485.1 Correlation between DA release and age of onset for sPD subjects inmost affected putamen region. . . . . . . . . . . . . . . . . . . . . . . 565.2 Correlation between DA release and change in UPDRS for sPD sub-jects in most affected putamen region. . . . . . . . . . . . . . . . . . 575.3 Correlation between DA release and RAC baseline BPND values aftercorrecting for age of onset for sPD subjects in least affected putamenregion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 575.4 Correlation between DA release and age of onset after correcting forRAC baseline BPND values for sPD subjects in least affected puta-men region. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585.5 Correlation between RAC baseline BPND values and age of onset forsPD subjects in caudate region. . . . . . . . . . . . . . . . . . . . . . 58xiChapter 1Introduction1.1 Positron Emission TomographyPositron Emission Tomography (PET) has been widely used to provide informationabout numerous neural pathways and abnormal patterns of neural activity in pa-tients with neurodegenerative diseases. In this section, I will briefly discuss basicprinciples behind PET radiotracers, radioisotope decays, signal detection and imagereconstruction.OverviewPositron emission tomography (PET) is a nuclear imaging technique which uses ra-diotracers to construct a 3D image based on the spatial and temporal distributionof the tracers and provides functional information of the tissues of interest in-vivo.PET enables the monitoring of molecular or cellular processes for varies diagnosticor therapeutic applications. Since patients with brain disorders often show distinctmetabolic patterns under PET scan, PET imaging offers the possibility to determinein-vivo multiple aspects of physiological processes for the study of varies neurode-generative diseases.The radiotracer tagged with a radioisotope is introduced into the patients body.The radioisotope decays to produce positrons. Annihilation between a positron andan electron produces two 511 keV gamma rays flying off in opposite directions alonga random orientated line. The two gamma rays are detected in coincidence by a pairof scintillation detector elements. The imaginary line that joins the location wherethe two gamma rays are detected is used to assign those matching gamma rays toa specific line of response (LOR). The measured counts in each LOR are used toproduce a 3D image of the object being studied via various image reconstructionalgorithms. Details of the process will be discussed the following sections.Table1.1 summarizes characteristics of common imaging modalities. Comparedto other imaging modalities, PET has the highest sensitivity and specificity, but rel-11.1. Positron Emission TomographySpatial Resolution Temporal Resolution Sensitivity (mol l−1)PET 4-6 mm 5-10 s 10−11 − 10−12SPECT y 8-11 mm 5-10 s 10−10 − 10−11MRI 0.05-1.5 mm 0.1-5 s 10−3 − 10−5CT 0.05-0.8 mm 0.1-0.5 s not well definedUltrasound 0.05-0.5 mm 0.005-1 s single microbubblesTable 1.1: Comparison between commonly used imaging modalities. [1]atively low temporal and spatial resolution. On the other hand, magnetic resonance(MR) imaging and computed tomography (CT) can provide high spatial resolutionand high soft tissue contrast. Hybrid imaging such as PET/CT and PET/MR havebeen adopted to combine anatomical and metabolic information. In the field ofbrain research, PET/MRI provides more comprehensive investigation of brain orga-nization and physiology by looking at metabolic and functional information at thesame time. For example, in the study of brain connectivity, structural or functionalconnectivity from MRI can be combined with metabolic connectivity from PET toprovide more specific, sensitive and quantitative measurements, and therefore yieldnew insights into the brain [12][1].1.1.1 RadiotracerA radiotracer is a biological molecule (tracer) tagged with a radioisotope. The choiceof tracer and molecule depends on the tissue to be studied and questions of interest.In oncology, fluorodexoyglucose (FDG), which is a glucose analog that is takenup by glucose-using cells, is often used as the tracer. FDG molecules are tagged withradioisotope fluorine-18 (F-18), which has a half-life of 110 mins. These radioactive[18F]-FDG molecules are trapped in cells until decay. Since cancerous cells take upmore glucose than normal cells, FDG/PET can be used for diagnosis, staging andmonitoring treatment of various cancers. In neuroimaging, regions with higher brainactivity have higher radioactivity. Since brain has a high uptake of glucose and manyneurodegenerative diseases, such as Parkinson’s disease and Alzheimer’s disease,decrease the brain metabolism in certain brain regions, FDG/PET can be also usedfor neuroimaging [13]. Other tracers, such as raclopride, dihydrotetrabenazine andfluorodopa, are used to target specific physiological pathways related to variousdiseases. Details about a few radiotracers relevant to this project will be discussedin later sections.21.1. Positron Emission TomographyOnce inside the body, the radio-labeled molecule can ’tracer’ out its path. Ideallya tracer should have the following properties:1. specific to the process or site of interest, which can be difficult since tracer inbloodstream can be carried away from the site of interest2. no metabolism after injection. Metabolites can carry labeled isotopes awayfrom the site of interest and cause problems in data interpretation3. easily synthesized from available precursor. The synthesis process should takea relatively short-time before radioisotope activity decays, and have a goodyieldIsotope Half-Life (mins) Maximum Energy (MeV) Range in Water (mm)18F 109.7 0.635 1.0311C 20.4 0.96 1.8613N 9.96 1.19 2.53150 2.07 1.72 4.14Table 1.2: Characteristics of commonly used PET radioisotopes. The range in wateris the total distance traveled by the positron before it annihilates with an electron.[2]PET imaging depends on positron-emitting isotopes. Radioisotopes can be pro-duced by a cyclotron, or as bi-products of a nuclear reactor, or in a generator system.Short-lived isotopes, such as 11C and 150, have to be produced by an in-house ornearby cyclotron. Some commonly used isotopes are listed in Table1.1.1. Isotopessuitable for PET imaging have the following characteristics:1. half-lives are long enough for the duration of the scan, but not too long toavoid unnecessary patient exposure to radiation2. the decayed positrons have relatively low energy and short range before anni-hilating with electrons in the body to avoid inherent error in the data3. targets are easily available for isotope production4. do not change the biochemical properties of the labeled tracer moleculesThe labeled radiotracers have the same physiological properties and hence sametracer kinetics as the unlabeled biological molecules. When the radioisotope decays31.1. Positron Emission Tomographyby emitting gamma rays, distribution of the radiotracer is mapped as a function oftime and space by the detected gamma rays. The spatial and temporal distributionof the tracer provides functional or metabolic information of the tissue dependson its biochemical or metabolic properties without disturbing the normal tissuefunction.The low amount of radiotracer administered to the patient does not induceany pharmacological effect nor affect the biological process under observation [14].1.1.2 Radioisotope DecayRadioactive decay is based on the unstable nucleus with too many neutrons orprotons which disrupts the balance between attractive and repulsive forces in thenucleus. the Coulomb force (repulsive) and the strong force (attractive). Unlikestable nuclei, unstable nuclei do not have enough attractive force to hold the nucleipermanently together, and are therefore radioactive.Positron emission is a particular type of radioactive decay, in which a proton ina proton-rich nucleus (X) is converted to a neutron while releasing a positron (β+)and a neutrino (υ):AZX =AZ−1 Y + e+ + υ (1.1)After converting proton to neutron, the nucleus decays to its stable form (Y).Positron is emitted to conserve electric charge.Annihilation occurs when a subatomic particle collides with its antiparticle, inthis case, when a positron collides with an electron. Due to energy and momentumconservation, low-energy annihilated particles are replaced with two gamma rayphotons. Since both electron and positron have a rest energy of 511k electron volts(eV), this energy is given off equally to two gamma rays. Energy and momentumare conserved with 1.022 MeV of gamma rays traveling at opposite directions, whichare detected by PET detection system [4].Emitted positron undergoes many scattering events along its path in the mediumbefore annihilates with an electron, so the actual path length the positron travelsbefore annihilation (p) is greater than the positron range (r) as shown in Figure1.1.Corrections are made to account for the degrading effects of positron range, espe-cially for isotopes with high positron range. Many have used Monte Carlo simula-tions or high resolution optical methods to model the positron range distribution.[2]41.1. Positron Emission TomographyFigure 1.1: Nucleus decays to emit positron (β+) which travels to the site of anni-hilation where it annihilates with an electron (e−) producing two 511 keV gammarays (γ) in opposite directions. r is the displacement from the parent nucleus to thesite of annihilation, whereas p is the actual path of β+ [4]Radioactive Decay RatesFor a given radioactive nucleus, the exponential probability of decay is describedusing decay constant λ, the probability that a nucleus will decay per unit time. Theradioactivity (A) of a sample is the number of decays per unit time in the units ofBecquerel (decays/second). N is the number of radioactive atom in the sample. Thehalf-life (T1/2)is the time when activity of sample has halved.N(t) = N0e−λtA(t) =−dN(t)dt= λN(t)T1/2 =ln(2)λ(1.2)51.1. Positron Emission Tomography1.1.3 Detection SystemDetection system is a key component to obtain quantitative information from theimaging system. PET scanner detects the two gamma rays originating from positronannihilation using scintillation detectors. The annihilation photons (511 keV) trav-eling in opposite directions form a line of response (LOR) are detected in coincidence(i.e. by searching for light signal at this energy within a very short time windowwithin a few nanoseconds) by a pair of detector elements surrounding the part ofthe body being scanned (in our case, the brain).Figure 1.2: A schematic of PET block detectors which is based on the first com-mercial human PET scanner built in 1974. The scanner was made by 48 NaI(T1)detectors. [4][5]61.1. Positron Emission TomographyScintillationScintillating crystals detect the gamma rays and convert them to scintillation lightvia the following steps [5]:1. Incident photon on the scintillating crystal creates an energetic electron byCompton scatter or photoelectric effect2. Electron loses its energy as it passes through the scintillator, and excites otherelectrons along the way3. Excited electrons return to their ground state, releasing energy in the form ofvisible lightSome photons may scatter off the detector and only deposit portion of theirenergy in the scintillator, especially in small detectors, but increasing detector sizereduces the spatial resolution of the system. So choosing the ideal scintillator ma-terial based on the following criterion is important to have optimal scintillator per-formance:• high effective atomic number (Z). High Z materials have high linear attenua-tion coefficients, which increase the proportion of photons undergoing photo-electric absorption, thus increasing sensitivity of the scintillator• high light yield, meaning that incident photons should produce a large numberof scintillation photons• low self-absorption for the scintillation light• index of refraction close to glass, which improves the optical coupling betweenthe scintillator and photomultiplier tubesPhotomultiplier TubesThe scintillator is coupled with the photomultiplier tubes (PMTs) (as shown in Fig-ure1.2), which then generate electric signal in response to the light incidence andsend the signal to computers. The scintillating photons incident on the surface ofPMTs result in a short electrical pulse, which is then further amplified by electronicsand coincidence circuitry. When two signals from opposing detectors arrive in coin-cidence along a line of response (LOR) and sends this information to a computer.71.1. Positron Emission TomographyIn contrast to single photon emission computed tomography (SPECT), PET doesnot need physical collimator to determine the direction of the incident photons [6].Figure 1.3: a) two annihilation: X is detected by detector elements D3 and D67 alongline of response (LOR), Y is undetected because photon path does not interact withdetector ring. b) top view of annihilation of X [6]Figure 1.4: True coincidence detection vs scatter and random coincidence detectionin PET [7]Photon Interaction and Attenuation CorrectionWhen the two photon beams travel through the medium, they can interact withhuman tissues via Compton scatter or photoelectric absorption.In Compton scattering, which is the most dominant interaction for 511 keVphotons, a photon interacts with an electron, which results in decrease in photon81.1. Positron Emission Tomographyenergy (increase in wavelength) and change in the direction of the photon. This lostenergy is transferred to the recoil electron, and the energy of the scattered photonafter interaction is given by [15]:E′ =E1 + (E/m0c2)(1− cos θ) (1.3)where E’ is the energy of the scattered photon, E is the energy of incident photon,m0c2 is the rest mass of an electron, and θ is the scattering angle.In photoelectric absorption, the incident photon is completely absorbed and anenergetic electron is ejected from the outer bound shell of the atom [15]. In humantissues, the probability of photoelectric absorption is low for photon with 511 keVenergy [16].Compton scattering and other interactions lead to attenuation of the two 511 keVphotons. The number of photons pass through an attenuating material decreasesexponentially with increasing length of the material. For 511 keV photons, about7cm thickness of tissue is needed to reduce the number of photons to half.The probability that a photon will reach the detector is given by:P = exp−∫ x0 µ(x)dx (1.4)where P is the probability a photon will reach the detector at distance x throughsome attenuating material, and µ is the linear attenuation coefficient.Because the interaction probabilities for the two photons are independent ofeach other, the total probability that both photons will reach the detector and beenrecorded as a coincidence event is given by [16]:Pc = exp− ∫ L0 µ(x)dx (1.5)where L is the distance between two detectors.Coincidences Detection and CorrectionAs a result of annihilation, we expect the two photons to arrive at the detectorsat approximately the same time. Temporal mismatches (photon detection not oc-curring at the same time) may occur due to the finite timing resolution of thescintillation crystal and the processing time of the PMT. These timing uncertaintiesare taken into account using the coincidence time-window, usually in the order of91.1. Positron Emission Tomography6-10ns [17]. When annihilation occurs at a location closer to one detector than theother, there will be a slight delay from one photon than the other. This can becorrected with time-of-flight PET imaging, which uses the relative time difference(∆t) between detection of two photons to estimate the most likely location of theannihilation event along the LOR.There are 4 different kinds of coincidence events in PET: true, scatter and ran-dom (as shown in Figure1.4).As mentioned before, true coincidence event occurs when two 511 keV photonsfrom annihilation are detected by scintillators at the same time. Photons do notundergo any interaction before detection and no other event is detected within thiscoincidence time-window [17].Random coincidences occur when two detected photons are actually originatedfro two separate annihilation events, which can be corrected using a delayed coinci-dence circuit. Scattered coincidences are caused by scattered photons within patientbody as a result of Compton scattering. Even though two photons are originatedfrom the same annihilation event, since the direction of the photon is changed, theLOR does not cover the true location of this event. This incorrect LOR assignmentcan be corrected using complex simulation methods [7]. These scattered and randomcoincidences add noise to the signal and decrease image contrast.Attenuation Correction Attenuation due to interaction between photons andtissues can be corrected using attenuation correction (AC). Because tissues withdifferent densities have different attenuation abilities, less dense regions (e.g. lungs)will appear darker (more photon emissions) than more dense regions (e.g. bones)without AC, which can lead to inaccurate estimation of tracer uptake. To performAC, we need to obtain the attenuation map from all LORs. On stand-alone PETscanners, a transmission scan is usually performed, in which an external positronsource is rotated around the patient to determine the attenuation of this transmissionphoton beam [18]. In PET/CT scanners, CT images can be used for PET AC [19].The collected data are the counts for the number of coincidences [n*(l), . . . ,n*(D)] where n*(d) is the total number of coincidences counted by the dth detectorpair and D is the total number of detector pairs. Note that not all photons reachthe detector due to attenuation inside the body and detector. Considering differentfactors affecting the counts of coincidence events, the following equation is oftenused to estimate loss of counts (or attenuation)101.1. Positron Emission TomographyYdθ = γdθ[ηtdθPdθMdθ + ηrdθrdθ + ηsdθsdθ]Pdθ = exp−∫µ(x)dxwhere µ(x) is the linear attenuation coefficient at position x, M is the number ofannihilation along LOR specified by (d, θ), P is the survival probability (probabilityof a photon not interacting along LOR), r is the number of accidental coincidences,s is the number of scattered events, η is the probability of each corresponding event,γ is the probability of event not being lost due to deadtime [20][21].1.1.4 Image ReconstructionAfter corrections for attenuation, scatter and random effects, the number of countsalong each LOR is proportional to the line integral of the activity along that LOR,which is known as projection. Radiotracer distribution inside the body is modeledusing 3D volume elements (voxels), and f is the true image which can be representedas the number of β+ decays at each spatial location corresponding to each voxel.If the number of coincidence events along each LOR is vector p, the system matrixH relates radioactive decay inside the body to coincidence events recorded by thedetector. The imaging system is described as:p = Hf (1.6)Given p and H, we can estimate f through varies image reconstruction algorithms.The reconstruction algorithm of choice depends on the question of interest, andthere is always a trade-off between the signal-to-noise ratio (SNR), contrast, biasand resolution.There are two common types of PET image reconstruction algorithms, the it-erative and analytic algorithms. Iterative algorithms are very flexible and requireno constraints on the system model, but can be very computationally expensive.Analytic algorithms, on the other hand, are much faster, but have higher noise andlimited quantification accuracy [4][22].Analytical ReconstructionAnalytic reconstruction, such as the filtered back projection (FBP) algorithm, re-quires the system matrix H to be simplified. In the analytic approach, a finitenumber of projections is applied back to the image to obtain a rough estimation111.2. Kinetic Modelingof the true radiotracer distribution. Star-like artifacts resulting from the limitednumber of projections can be improved using a ramp filter. The combination of theback projection and the ramp filter is the FBP method. In the analytic approach,nostatistical model is included, which lowers the SNR [23].Iterative ReconstructionCompared to analytic reconstruction algorithms, iterative reconstruction approxi-mates the real solution of the object using multiple iterative steps, which allows usto reconstruct better images but at a higher cost of computational time. Iterativereconstruction algorithm has the advantage of improved noise insensitivity, which isparticular interest for images with poor noise statistics like PET [23][22].In iterative reconstruction approach, we need to define the parameters to esti-mate (represent radiotracer concentration) and the system model which relates theradiotracer distribution and the mean of the measured data.The system matrix Hij, which is the probability that an emission from voxel jis detected in projection i, characterizes the imaging system. The projection pi isgiven by [24]:pi =N∑j=1Hijfj (1.7)After acquiring the projection measurements, a statistical model is used to de-scribe how the projection measurements vary around the expected mean. A costfunction is often used to define the ’best’ image, and the Maximum Likelihood ap-proach is most commonly used for PET since it offers unbiased, minimum varianceestimates as the number of measurement increases.1.2 Kinetic ModelingTo obtain quantitative measurements from physiological tracer distribution, we oftenneed a kinetic model to relate PET data to tissue functions. In these models,radiotracer moves between different tissue compartments, which represent differentbiochemical states of the radiotracer and its metabolites. By assumption, there isan uniform radiotracer distribution inside each compartment.In a compartment model, there a fixed number of states with specific interactionsamong them and arrows represent pathways where radiotracers flow between each121.2. Kinetic Modelingcompartment. The change in concentration in each tissue compartment is describedby a linear, first-order ordinary differential equations (ODE) of the concentrationsin all other compartments. From these ODEs, tracer kinetics are the convolution ofthe input function and response function in other compartments.Compartment models can be used to fit the tissue concentrations as a functionof time from the measured PET data. Kinetic models used in PET quantificationare discussed in the following sections.1.2.1 Time Activity CurvesThe time activity curve (TAC) gives the radioactivity value in each region-of-interest(ROI) or pixel across a sequence of PET images (i.e. scanning time) as shownin Figure1.5. Radioactivity of the tracer reaches maximum then stabilizes as theisotope decays. To quantify this curve, we need to fit compartment models to obtainquantitative information.Standard Uptake ValuesStandard uptake values (SUV) can be estimated from TACs. SUV is defined as theratio of 1) the mean tissue radioactivity concentration c of ROI (Mbq/kg), and 2)the injected activity (Mbq) divided by the body weight (kg). SUV(t)=c(t)/(injectedactivity(t) / body weight). SUV represents the ratio of (1) the image derived ra-dioactivity concentration found in certain ROI, and (2) as reference the radioactivityconcentration in the hypothetical case of an even distribution of the injected radioac-tivity across the whole body. Average SUV in each ROI was also obtained over thelast 30mins (50-80mins) time frame. SUV can be significantly affected by imagenoise, low image resolution and/or user biased ROI selection.1.2.2 Reference Tissue ModelThe commonly used reference tissue model (RTM) in PET neuroscience studies isillustrated in Figure1.6 and parameters are explained in Table1.2.2. The referenceregion is modeled as a ’single-tissue’ compartment (CR), while the target region ismodeled as a ’two-tissue’ compartment (CNDandCS).CP is the arterial input function, which represents the cumulative availability ofthe radiotracer in arterial plasma. Tissue concentration normalized to the cumula-tive arterial concentration is often used as the gold standard for quantitative PET131.2. Kinetic ModelingFigure 1.5: Time activity curve for [11C]-DASB tracer with SRTM model fit in theleft caudate region. The estimated parameters from SRTM are shown. Error barsare estimated from the scanner count-rates [4]studies. However, since getting arterial blood samples during the scan can be achallenge sometimes, arterial input function can be sometimes replaced by referencetissue if one exists.The rate of which tracer moves from one compartment to another is proportionalto the tracer concentration in the first compartment [3]. For the non-displaceablecompartment CND, we have the following differential equation describing the changein radioactivity concentration:dCNDdt= K1CP − k2CND − k3CND + k4CS (1.8)In RTM, the two target tissue compartments do not represent different physicalspaces. The specifically bound compartment CS is where tracers bind to theirspecific target. The non-specifically bound compartment CND is where tracers may141.2. Kinetic ModelingFigure 1.6: Systematic diagram of the reference tissue model [8]. Using a referenceregion, parameter Cp can be eliminated.either bind to non-target molecules or unbound (free).The reference tissue compartment CR, which is assumed to have no specificbinding but similar non-specific binding as the target compartment, is used to esti-mate the input function without measuring tracer concentration in plasma CP . Thereference tissue compartment only contains non-specifically bound and free tracer[3][8].For RTM, the operational equation is given by:CT (t) = R1[CR(t) + a · CR(t)⊗e−ct + b · CR(t)⊗e−dt] (1.9)where t is the time after tracer administration and⊗is convolution.R1 is defined as the ratio of tracer deliver rates between the target and referenceregions:R1 =K1K ′1(1.10)151.2. Kinetic ModelingSymbol Definition UnitsCP Tracer concentration in plasma kBq.mL−1CND Tracer concentration in non-specifically boundtarget tissuekBq.cm−3CS Tracer concentration in specifically bound targettissuekBq.cm−3CR Tracer concentration in reference tissue kBq.cm−3K1 Rate constant for transporting tracer from arte-rial plasma to reference tissuemL.cm−3.min−1K ′1 Rate constant for transporting tracer from arte-rial plasma to reference tissuemL.cm−3.min−1k2 Rate constant for transporting tracer from ref-erence tissue to venous plasmamin−1k′2 Rate constant for transporting tracer from ref-erence tissue to venous plasmamin−1k3 Rate constant for transporting tracer from non-displaceable to specifically bound compartmentmin−1k4 Rate constant for transporting tracer fromspecifically bound to non-displaceable compart-mentmin−1Table 1.3: Symbols used in the reference tissue model, as shown in Figure1.6. [3][4]The parameters a,b,c,d can be estimated from different combinations of rateconstants. Because we assume that the volumes of distribution for non-displaceabletracer are equal in the reference and target region, so thatK1k2=K1′k2′. Thisassumption allows us to reduce the number of independent parameters from five(a,b,c,c and R1) to four (R1,k2,k3,and BPND), which are then estimated usingnonlinear regression analysis [8]. The two assumptions of RTM are: 1) there is nospecific binding in the reference tissue compartment and 2) K1/k2 is the same inreference and target tissue compartments.Binding PotentialThe goal of PET study is to estimate all the rate constants in the compartmentmodel using the fit parameters. Instead of estimating these rate constants directly,which is prone to statistical noise in model fitting, more robust parameter is usedto combine these rate constants. Because the rate constants are highly covariated,the overall error for the combined parameter is less than the error for each rate161.2. Kinetic Modelingconstant.In a ligand receptor binding system, ligand-receptor kinetics is described by theMichaelis-Menten equation.L+R↔ LRwhere L = ligand, R = receptor, LR = ligand-receptor complex.Binding potential (BP) is most commonly used to estimate the rate constants.InPET, BP values are combined measures of availability and affinity of neuroreceptors,and is the ratio of Bmax oto KD:BP = Bmax/KD = receptordensityxaffinity (1.11)where Bmax is the total concentration of receptors in the tissue, and KD is theradioligand equilibrium dissociation constant.There are different definitions of BP, but for our interest, non-displaceable bind-ing potential (BPND) is used, which defined as [3]:BPND =k3k4(1.12)1.2.3 Simplified Reference Tissue ModelRTM has four parameters to be estimated which is often too complex for the noisyPET data. In most cases, it can be replaced by the simplified reference tissue model(SRTM).Compared to the original RTM, SRTM reduces the number of tissue compart-ments to one instead of two (i.e. combined the specifically bound compartment CSand non-specifically bound compartment CND). This reduces the number of param-eters from four to three (eliminate k3) and reduces the variability in the parameterestimates. SRTM is used to quantify the receptor kinetics from PET measurementsusing input function derived from a reference region without acquiring an arterialinput function [8] [25]. The three parameters used in SRTM are R1 (relative deliv-ery in tissue compartment compared to the reference region), k2’ (the clearance rateconstant from the reference region) and BPND (k3/k4) estimated using nonlinearfitting or more complex models.SRTM uses the following three assumptions to estimate specific binding in tissueregions of interest as a function of the reference region [25][26]:171.2. Kinetic Modeling1. reference tissue compartment has no specific binding2. the volumes of distribution for non-displaceable tracer are equal in the refer-ence and target region, so thatK1k2=K1′k2′.3. there is no difference between the specific and the non-specific compartment,so TAC can be fitted by an one-tissue compartment model.SRTM was used to generate BPND values for covariance pattern analysis in theserotonergic system, which was the main part of this thesis work.1.2.4 Two-Step Simplified Reference Tissue ModelSRTM calculates one BPND value for each ROI using regional TAC, but sometimesparametric images of BPND values are of particular interest. In parametric images,each voxel is related to some physiological parameter. This is done by applyingtraditional model to each individual voxel separately.Although there is only one true value of k2’, SRTM estimates k2’ value for eachpixel of the image. A two-step method (SRTM2) was developed [25][26]:1. R1, k2 and k2’ values are calculated using SRTM for all brain pixels. A globalk2’ value is calculated from all pixels outside the reference region.2. Fix k2’ value to the averaged global value and calculate functional images ofBP and R1 using a two-parameter fit.SRTM2 was used to generate parametric BPND images for DASB tracer andwas compared with regional BPND values from SRTM.1.2.5 Logan Plot MethodCompared to RTM or SRTM, data fitting using linear regression methods is inparticular interest due to faster computational time . Logan plot is a graphicalmethod which reduces the number of parameters by transforming the model equation(1.9) to a linear equation evaluated at several time points and interpret the slopesand intercepts of the linear equations [27]. This method is independent of thespecific model structure of the reference tissue and uses a global clearance rate k2’as SRTM2. Logan plot method was used to generate regional BPND values for RACand DTBZ tracers.181.3. Network Analysis1.3 Network AnalysisFunctional imaging techniques allow us to quantify brain activity to study patho-physiology of neurodegenerative disorders, but absolute activity may not providethe complete picture. In addition, activity variability between subjects and brainregions increases the difficulty to quantify PET signal.Moeller and colleagues [28] proposed a data-driven, statistical regional covariancemodel based on multivariate principal component analysis (PCA), namely the ScaledSubprofile Model (SSM). SSM models the sources of subject and region variationas spatially distributed networks in functional images. SSM is able to identify agroup-dependent, region-specific, disease specific spatial covariance patterns in thebrain that can be used study the heterogeneous regional interactions in differentpatient groups and to discriminate patients from healthy controls [29][30]. Networkanalysis has been proven to be more robust than local binding analysis and moresensitive to small changes. For the serotonergic system, which is the main focus ofthis thesis work, most studies on serotonergic pathways have been focused on localbinding, disease-specific alteration of the functional network across the entire brainis still unknown.1.3.1 OverviewDisease-specific metabolic network abnormalities have been used to accurately dis-criminate between PD patients and controls using SSM. The so-called PD-relatedpattern (PDRP) derived from 18F-fluorodeoxyglucose (FDG) PET was characterizedby increased pallido-thalamic and pontine activity associated with relative reducedactivity in the cortical motor regions [31] [32] and was found to correlate consis-tently with Unified Parkinson’s Disease Rating Scale (UPDRS) motor scores [30]and clinical response to therapy [33].Similar network analysis has also been applied to identify the PD-related cogni-tive pattern (PDCP) using FDG as a potential biomarker of cognitive functioning inPD. PDCP pattern was characterized by relative increased activity in the cerebellarvermis and dentate nuclei with associated reduced activity in frontal and parietalassociation areas [34]. It was also shown that brain network patterns associatedwith motor and cognitive functions are orthogonal of each other [32]. By applyingnetwork analysis to the serotonergic system, we can further study the functioningrole of serotonergic pathways in PD.191.3. Network Analysis1.3.2 Principal Component AnalysisSSM is based on PCA to decompose sources of variation (deviation) in the data intoa set of linearly uncorrelated/orthogonal vectors called Principal Components (PCs).PCs are ranked based on the variance accounted by each PC in the subject by regiondata matrix, so that the first PC accounts for the largest variance. PCA is doneby Singular Value Decomposition (SVD) of the data matrix (detailed mathematicalsteps are listed in the next subsection) [35].1.3.3 SSM/PCA AnalysisAfter minimizing substantive variability in subjects and brain regions, we can iden-tify the significant spatial covariance patterns in the combined control and patientgroups. Detailed computational steps of region-based SSM are listed below [30][28]:1. subject PET images are smoothed and normalized onto a common template(e.g. the normalized Talairach-like space (MNI)) using Statistic ParametricMapping (SPM) software. This step ensures all functional activity measure-ments at different locations in the brain are mapped onto the same coordinatesystem in a one-to-one correspondence fashion. (This step can be eliminatedwhen using regional BPND values).2. regional quantitative measurements (e.g. BP values or SUV) are obtainedusing pre-defined brain region mask. Value in each region r (1,...,N) of eachsubject s (1,...,M) are combined together to form the subject by region datamatrix Psr.3. (optional) depending on the characteristics of the quantitative measurements,logarithmically transformation of Psr can be applied. Logarithmically trans-formation makes highly skewed distributions less skewed, and changes multi-plicative scaling effect into additive components which can be then removedby double centering step (step 4) . For BP data, we did not apply any trans-formation before applying statistical analysis.Psr −→ LogPsr4. LogPsr is centered with respect to subject means in each region LGMRrand region means in each subject GMPr obtain the Subject Residual Profile201.3. Network Analysis(SRPsr). LGMRr is the mean across subjects of brain data of region r,and GMPr is the mean across regions of subject s. Through this doublecentering process, the resulting matrix SRPsr is the deviation of the meansubject and mean regional values which represents a coordinate system thatrelates differences from the mean values. Double centering ensures the resultwas invariant to subject and regional scaling effects.SRPsr = Psr − LGMRr −GMPrwhere LGMRr = meanregion(LogPsr)GMPr = meansubject(LogPsr)−meansubject(LGMRs)(1.13)5. Singlular Value Decomposition (SVD) of SRPsr matrix to derive the regionalbrain patterns and the associated subject scores. We first determine the MxMsubject by subject covariance matrix Ssub in matrix format as:Ssub = SRP ∗ SRP T (1.14)Eigenvalue decomposition of the matrix Ssub results in eigenvalues (λk, k=1,...,M)and eigenvectors (ek, k=1,...,M):Ssubek = λkek (1.15)After left multiplying both sides by SRP T :SRP TSsubek = λkSRPT ek(SRP TSRP )SRP T ek = λkSRPT ek(1.16)where (SRP TSRP ) is the NxN region by region covariance matrix Sreg. Wecan also get the Group Invariant Subprofile vectors (GISk) as eigenvectors ofmatrix Sreg using the same eigenvalues as before λkSregGISk = λkGISkwhere Sreg = SRPTSRP and GISk = SRPT ek(1.17)ek vectors weighted by the square root of their corresponding λk eigenvaluesgives the subject score vectors (Scorek) whose elements represent the pattern211.3. Network Analysisvector GISkScorek =√λek (1.18)So as a result of SUV, SRPk for each subject is expressed as sum of the GISkmultiplied by corresponding subject score (Scorek) along each PC:SRPk =∑kScorekGISk (1.19)The eigenvalues represent the Variance Accounted For (V AFk) for each vector:V AFk = λk/(λ1 + λ2 + ...+ λk) (1.20)Identify Disease-Related patternTherefore, for each PC, we have the subject scores for each subject and GIS valuesfor each region. To identify significant topographic covariance profiles which bestdistinguish between subject groups, PCs related to disease are judged by discrim-inative accuracy between groups. A single or linearly combination of GIS vectorshas to separate the subject scores of two subject groups at a pre-specified statisticalthreshold. Only the first few PCs accounting for a relatively significant amount ofvariation in the original data matrix, which correspond to major sources of spatialvariance, should be considered. The general steps for identification of significantdisease-related spatial covariance pattern are listed below:1. choose PCs accounting for a significant amount of total variation in the datamatrix for further analysis. This eliminate PCs accounting for low percentageof total variance due to noise.2. choose PCs satisfying a pre-specified statistical threshold. This is done byapplying two-sample T-test on the subject scores along each PC, and selectonly PCs with p-value lower than a pre-specified threshold (e.g. p < 0.001) toeliminate noise. This threshold is based the characteristics of the data matrix.3. subject scores of the selected PCs are entered into logistic regression modelswith groups (binary number) as dependent variables and subject scores as221.3. Network AnalysisFigure 1.7: Overflow of the Scaled Subprofile Model.231.4. Parkinson’s Diseaseindependent variables. The combination of PCs with the lowest Akaike In-formation Criterion (AIC) [36] is selected as the one which best distinguishesbetween groups.4. p-value of likelihood-ratio test for this combination of Z-transformed subjectscores (based on equation1.21 ) is used to examine the level of discriminationbetween groups.5. PCs are then combined using the coefficients from the regression model toyield a single disease-related covariance patternZscore = (score−mean(score(1 : NC)))/std(score(:)) (1.21)To obtain the most robust results that is most suitable for our dataset, wecombined SSM analysis with bootstrap resampling techniques to identify the disease-related covariance pattern. Details about the modification can be found in themethod section.Topographic Profile RatingAfter obtaining the significant disease-related covariance pattern, we can apply theforward application to calculate the subject scores for a specific pattern on individualbasis using equation1.22. This process is called the topographic profile rating (TPR)[28].SRP Tk GISk = Scorek (1.22)1.4 Parkinson’s Disease1.4.1 BackgroundParkinson’s disease (PD) is the second most common progressive neurodegenerativedisorder of the central nerves system (CNS), and has a prevalence of approximately0.3% of the entire population. PD affects around 100,000 Canadians with a cost ofabout $2.5-5 billion annually. The prevalence of PD increases significantly with age,affecting about 4.4% of people over 50 years of age and 11.9% over 80 years of age[37].241.4. Parkinson’s Disease1.4.2 SymptomsMotor Deficits The most common symptoms of PD are abnormalities in mo-tor system, including resting tremor, rigidity and difficulty initiating and sustain-ing voluntary movements. These motor abnormalities are known to caused by thepresence of Lewy bodies in the brain, which results in the degradation of nigrostri-atal dopamine (DA) neurons and therefore alters the activity of the motor cortico-striatopallido-thalami cortical (CSPTC) pathways. It is known that the motorsymptoms of PD start to occur when about 50% of the nigral dopaminergic neuronshave died, resulting in 80% reduction in the striatal DA content. [38]. Pathologicprocess of PD begins in the dorsal motor nucleus, proceeding to midbrain and fore-brain in different stages of the disease as predicted by the Braak hypothesis [39].This implies the existence of a relatively long preclinical period during which severaldisease-induced neurochemical changes take place.Cognitive Deficits Apart from common motor abnormalities, PD patients oftenexperience cognitive deficits which may occur before or along with motor deficit.Some of these deficits, such as depression and dementia, occur in more than onethird of PD patients and take years to develop before initial motor symptom onsetwhich make them potential preclinical markers of PD. These motor and cognitivedeficits can have a great impact on the quality of life. There is currently no curefor PD, medications and surgeries aim to reduce symptoms and improve quality oflife through coping mechanism which help patients adapting to motor and cognitivelimitations, mainly by using levodopa or DA agonists. [40] [41] However, treatment-induced cognitive complications are common side effects. [42].Exact causes of these cognitive symptoms are still unclear. Some studies havesuggested that the degradation of nigrostriatal DA content may also contributeto the cognitive deficits in PD due to the direct connections between the ventraltegmental area (VTA) and the prefrontal cortex [43]. However, changes in thedopaminergic pathway alone cannot fully explain the cognitive deficits of the disease.Several studies have suggested that the serotonergic pathway may contribute toseveral non-motor disturbance.Treatment There is currently no cure for PD, medications and surgeries aim toreduce symptoms and improve quality of life. The most commonly used medica-tion is the pharmacological replacement of DA, which is mostly accomplished by251.4. Parkinson’s Diseaselevodopa (L-DOPA). L-DOPA is a precursor of DA and is converted to DA in thedopaminergic neurons. Since motor symptoms occur as a result of DA degenera-tion, administration of L-DOPA temporarily diminishes the motor symptoms. DAagonists can also be used. Agonist binds to DA receptor and activates the receptorto produce a biological response.Motor function responds well to the DA therapy. The most common complica-tion as a result of DA therapy is dyskinesia, which is involuntary muscle movementand can range from slight tremor of the hands to uncontrollable movement of thebody. Treatment of cognitive symptoms in PD mainly aims to reduce symptoms orimprove quality of life through coping mechanism which help patients adapting tocognitive limitations [40].In order to better understand the disease, we will look closer at two neurologicalpathways inside the brain: dopaminergic pathway which is the most commonlyknown pathway to be affected by PD and serotonergic pathway which is linkedmore closely to cognitive abnormalities of the disease.1.4.3 Dopaminergic SystemMotor abnormalities are caused by the presence of Lewy bodies, which results inthe degradation of nigrostriatal DA neurons. Pathologic process of PD begins inthe dorsal motor nucleus, proceeding to midbrain and forebrain [39].Figure 1.8: Chemical structure of dopamine moleculeDopamine (3,4-dihydroxyphenylethylamine) is the neurotransmitter that con-trols the dopaminergic pathway, which has a characteristic anatomical pattern inthe brain. DA is involved in the regulation of locomotor activity, emotion and neu-roendocrine secretion. The chemical structure of dopamine is shown in Figure1.8.261.4. Parkinson’s DiseaseFigure 1.9: Neuronal projections of four DA systems in human brain [9]There are four central DA pathways which are defined neuroanatomically asshown in Figure1.9 [10][9].1. The nigrostriatal system projects from the substantia nigra to the caudatenucleus and putamen, where nearly 80% of DA is located here. Destructionof this pathway results in sever motor dysfunction.2. The mesolimbic system originates in the ventral tegmental area of the midbrainand projects to nucleus accumbens, olfactory tubercle, hippocampus, septalnuclei, and amygdala. This system is heavily involved in emotions, memoryand the reward system.3. The mesocortical system originates in the ventral tegmental area and projectsto the anterior cingulate cortex, septum, neocortex and prefrontal cortex.These cortical regions are important for motivation, cognition and emotionalcontrol.4. The tuberoinfundibular system projects from hypothalamus to the pituitarygland, which regulates the neuroendocrine functions.271.4. Parkinson’s DiseaseDTBZIt is known that the motor symptoms of PD start to occur when about 50% ofthe nigral dopaminergic neurons have died, resulting in 80% reduction in striatalDA content. We used [11C]-dihydrotetrabenazine (DTBZ) to estimate dopaminergicdenervation. DTBZ is a presynaptic vesicular monoamine type 2 (VMAT2) marker,which competes with DA and binds to these VMAT2 DA transporters within DA-producing neurons. VMAT2 binding site is a specific protein located in the mem-branes of presynaptic vesicles, so DTBZ is used to access membrane DA transporterbinding. Even though DTBZ is not specific to DA, it is found that over 95% ofVMAT2 binds to DA terminal in the striatum (largely in the storage vesicles in thepredominant dopaminergic terminals) and does not undergo any major disease ortreatment induced regulatory changes. Studies showed that there is striatal DTBZuptake reduction is correlated with motor disability and disease progression of PD[44]. The advantage of DTBZ is not up regulated by DA, so there is a consistentrate of decrease in the course of disease progression [45].1.4.4 DA releaseThe DA transmission process involves the production, release, reuptake, breakdownand diffusion of DA as shown in Figure1.10. DA is produced from tyrosine by DOPAdecarboxylase (DDC) in the presynaptic neuron. DA is then taken up by storagevesicles which carry and release DA into the synapse. Once in the synapse, DA canreact with 5 different post-synaptic DA receptors which trigger various cascade ofintracellular signaling in different parts of the brain. For example, D2 receptors aremainly located in the striatum, limbic areas, hypothalamus and pituitary gland [10].Abnormal DA release in the synapse was shown to relate to treatment-inducedmotor complications in PD patients [46] due to dramatic changes in receptor oc-cupancy. DA release pattern was analyzed in details using double RAC scans asshown in later sections.RACRadiotracer [11C]-raclopride (RAC) is used to study the change in DA release pat-tern due to various stimuli. RAC is a dopamine D2 receptor antagonist and can beused to evaluate the amount of DA changes and the synaptic DA loss. The antag-onist binds to the receptor and competes with DA released due to any stimuli for281.4. Parkinson’s DiseaseFigure 1.10: Systematic diagram of DA synapse showing the production, release,reuptake, break down and diffusion of DA. [10]binding to these receptors [47]. Therefore, if DA is released because of any stimuli,the ability of RAC to bind to the receptors will be reduced, resulting in a lowerRAC BPND value for scans taken during stimuli. There is evidence that differentdisease-induced changes in the synaptic DA levels (referred to as DA release pattern)may be related to different manifestation of the disease and responses to therapy[46][47].Clinically, RAC shows an asymmetric binding in the left and right hemispheres.There is evidence that in advanced PD subjects, improvement in bradykinesia andrigidity scores following DA medication administration significantly correlated withreduction in RAC binding, suggesting an increased DA release into the synapse.291.4. Parkinson’s DiseaseAs discussed before, L-DOPA treatment, although remains as the most effectivetreatment of PD symptoms, can lead to motor fluctuations and complications suchas L-DOPA-induced dyskinesia (LIDs). Although the origin of such complicationsis not clear, it is thought that both pre- and post-synaptic DA systems in com-bination with non-DA systems play an important role in the development of thecomplications.1.4.5 Serotonergic SystemSerotonin neurons (5-HT) are originated in the dorsal raphe nuclei which are thenprojected to the basal ganglia (particularly in the striatum) and to the frontal cor-tex and limbic system as shown in Figure1.12 [48]. Serotonin (5-hydroxytryptamine,5-HT) is a monoamine brain neurotransmitter. The chemical formula for serotoninis N2OC10H12 and its chemical structure is shown in Figure 1.11. Serotonin is pro-duced from the hydroxylation and decarboxylation of the amino acid tryptophan.In the central nervous system, serotonin is only synthesized by the neurons of Raphenuclei which are distributed along the length of the brainstem in nine pairs. Sero-tonin is widely distributed in the brain and serves an important role in the regulationof mood, sleep, appetite, memory, learning, etc. Low level of serotonin is associatedwith depression, bipolar disorders, fear and anxiety, since serotonin is required forthe metabolism of stress hormones [11].The serotonergic system is involved in different types of psychopathlogical con-ditions associated with PD, especially depression, weight and appetite problems[49][50]. The degeneration of serotonergic terminals occur earlier in the diseasecompared to dopaminergic system. PD patients exhibit progressive, nonlinear lossof serotonergic function, which starts in the caudate, thalamus, hypothalamus andanterior cingulate cortex and expands to the basal ganglia and limbic system asdisease progresses [51].It was shown that 5-HT neurons share the same monoamine biosynthetic compo-nents with DA neurons, which contributes to DA processing in denervated striatumin PD subjects [52] and play a role in levodopa-induced dyskinesia (LID) by releas-ing DA as a false neurotransmitter. An autoradiographic study has shown a higherSERT level in the putamen of dyskinetic than non-dyskinetic levodopa-treated sub-jects [42].301.5. LRRK2 MutationFigure 1.11: Chemical structure of serotoninDASBTo study the brain serotonergic system, we used second generation [11C]-DASB PETimaging tracer to measure the level of serotonin transporter (SERT) binding andto estimate serotonin neuronal integrity. DASB has high specificity and sensitivityfor SERT, and low affinity for DA transporter (DAT). Studies have suggested adecrease in striatal 5-HT [48] and some observed a striatal hyperinnervation in PDpostmortem study and animal models.1.5 LRRK2 MutationThe human leucine-rich repeat kinase 2 (LRRK2) gene was discovered in 2004 andis the greatest known genetic contributor to PD. Majority of PD is sporadic PD,meaning the cause of the disease is unknown. About 10 % of the disease is relatedto genetic mutation in the LRRK2 gene. In the US, LRRK2 mutation accounts forapproximately 0.5% of simplex PD (e.g. single occurrences in a family) and 2%-6%of familial PD. [53]Clinical characteristics of sporadic PD and LRRK2 mutation associated PDpatients are quite similar. The motor symptoms are comparable between the twoPD groups. Cognitive impairment does not appear to be more common in LRRK2associated PD than in typical sporadic disease [54].Studies have shown that there is an increased DA turnover and increased SERTbinding in asymptomatic LRRK2 mutation carriers compared to healthy controls311.6. Research ObjectivesFigure 1.12: Serotonin projections in the brain. Serotonin is produced in nucleusraphe and projects onto other brain regions [11][55][56]. Studying LRRK2 associated PD in comparison with healthy and sporadicPD populations can help us to further understand the contribution of genetic mu-tations to the disease, and understand the presymptomatic stages of the disease.1.6 Research ObjectivesPET imaging has been an effective tool to study altered neurological pathways inpatients with Parkinson’s disease. Even though motor symptoms of this disease aremainly due to the loss of neurons in the dopaminergic pathway, interactions betweenthe dopaminergic and serotonergic pathways may contribute to treatment-inducedcomplications (motor and cognitive) in later stages of the disease.The overall objective of this thesis work is to investigate the interactions betweenthe dopaminergic and serotonergic pathways in presymptomatic and early stages ofthe disease, and their contributions to treatment-induced complications and non-motor symptoms in PD subjects.In the first part of the project (Chapter 4), we applied a regional covariance321.6. Research Objectivespattern analysis to [11C]-DASB data in PD subjects (in early and moderate stages).Regions with relatively preserved binding in the disease-specific networks may actas a compensatory mechanism for the dopaminergic system. We also applied thesame analysis to asymptomatic LRRK2 mutation carriers to investigate if there isa distinct mutation-specific network, which may explain the increased risk of PD inthis group of subjects and maybe used as an earlier marker before motor symptomonset.In the second part of the project (Chapter 5), we analyzed the medication-induced DA release pattern on early PD subjects (less than 5 years of disease du-ration) using double [11C]-RAC scans. We want to examine if altered DA releasepattern in response to treatment would contribute to treatment-induced motor com-plications.The combination of abnormal medication-induced DA release pattern and up-regulation of the serotonergic system may be able to explain the occurrence oftreatment-induced complications and non-motor symptoms in PD patients, and actas a potential early marker for the disease.33Chapter 2Methods2.1 Subjects and Clinical InformationSubjects We studied 18 non-demented patients with PD, 9 asymptomatic LRRK2mutation carriers (LRRK2-AMC), and 9 healthy control volunteers (HC). The 18 PDsubjects were further divided into 12 sporadic PD (sPD) and 6 LRRK2 mutation-associated PD (LRRK2-PD) subjects. All subjects in this study had no clinicalhistory of depression or had received anti-depressant therapy, and they had no othermedication with known action on the serotonergic system. All healthy controls hadno history of neurological or psychiatric disorders and were not on any medication.All groups were matched for age and sex (Table2.1).Control sPD gPD UnaffectedNo of Subjects 9 13 8 9Age (years) 56±14 59±8 66±15 48±10Disease Duration (years) 3±3 8±7UPDRS total off 16±7 22±10UPDRS total on 10±6Gender 8M/5F 3M/5F 6M/6FTracers DASB DASB/DTBZ/MP/RAC DASB/DTBZ/PBR/PMP DASB/DTBZ/PBR/PMPTable 2.1: Participants clinical information: values reported as mean±standarddeviation; Disease duration has been accounted from the time of PD motor symp-toms initiation (not from time of clinical diagnosis); M=Males; F=Females; UP-DRS=Unified Parkinson’s Disease Rating Scale; UPDRS off=UPDRS without drug(LDOPA) intervention; UPDRS on=UPDRS when on drug medication.All subjects were used in the DASB SSM pattern analysis study. Only sPDsubjects were involved in DA release study because only they had double RACscans to evaluate DA release patterns.Clinical Evaluation PD subjects were clinically evaluated with the Unified Parkin-son’s Disease Rating Scale (UPDRS). UPDRS was used to measure severity of mo-tor symptoms and therefore monitor disease progression. UPDRS off was measuredwhen subjects were taken off LDOPA medication; UPDRS on was measured whensubjects were still on medication. By calculating the change in UPDRS with and342.2. Scanning Protocol and Image Processingwithout medication, we can relate the amount of DA released with the effectivenessof the medication.Age of disease onset for PD patients was defined as reported by the patients,not at diagnosis. Montreal Cognitive Assessment (MoCA) scores were also recordedto access mild cognitive dysfunction. The assessment tests on attention, executivefunctions, memory, language, visuoconstructional skills, conceptual thinking, cal-culations, and orientation. Hoehn and Yahr scale was used to access the diseasesymptoms progression. Beck depression inventory was used to access the severity ofdepression.2.2 Scanning Protocol and Image Processing2.2.1 Scanning ProtocolTo perform [11C]-DASB PET scans, a mean dose of 555Mbq of DASB radiotracerwas administered by intravenous injection over 60 seconds, and acquisition timeof 80 minutes was used. For DTBZ and RAC scans, 60mins scan time was used.PET images were obtained on a high resolution research tomography with an in-plane resolution of 2.3mm. Patients stopped medication for at least 18 hours beforescanning. MRI scans were performed as resting-state MRI at UBC 3 Tesla MRIcenter.2.2.2 Image ProcessingReconstructed images were summed to create the entire dynamic set using Matlabbased Statistical Parametric Mapping (SPM12) software. Pre-defined high-contrastregion-of-interest (ROI) templates were developed in Montreal Neurological Institute(MNI) space using MRI and PET data from healthy controls.Subject PET images were coregistered to the corresponding MRI images usingSPM12 software, and then warped onto the MNI space to obtain the correspondingtransformation matrix. Inverse transformation was applied to the MNI space ROIsto place ROIs onto the original PET images.ROI SelectionFor DASB, ROIs were manually defined on both hemispheres for 21 non-overlappingROIs which are known to be related to the serotonergic system: anterior and pos-352.2. Scanning Protocol and Image Processingterior cingulate (ACC and PCC), amygdala, caudate, cerebellum, dorsolateral pre-frontal cortex (DLPFC), hypothalamus, insula, medulla, midbrain, orbital frontalcortex (OFC), pons, pedunculopontine nucleus (PPN), putamen, substantia nigra(SN), thalamus, ventral striatum (VS), hippocampus, ventral tegmental area (VTA),dentate nucleus (DN), and globus pallidus (GP).ROIs for DTBZ and RAC scans were placed on the ventral and dorsal striatumregions (1 on caudate and 3 on putamen) for both hemispheres. Four consecutivesaggital slices (17mm) were selected for data extraction for the dorsal striatumregions, 3 consecutive saggital slices (7.5mm) were used for the ventral striatum dataextraction. Occipital Cortex was used as the reference region which was defined by 3ROIs placed on the same slices as the striatum regions for DTBZ. The Cerebellumwas used as the reference region which was defined by a single ROI placed on 3consecutive saggital slices (12.75mm)Quantitative MeasurementsFor DTBZ and RAC, Logan plot method was used to obtain BPND values in eachof the 5 ROIs.For DASB, we obtained 3 sets of quantitative data from these PET/MRI im-ages: regional non-displaceable binding potential (BPND) values, parametric BPNDvalues, and standard uptake values (SUV) for all subjects:• regional BPND values in each ROI were obtained using Simplified ReferenceTissue Model (SRTM) with cerebellum as reference region• parametric BPND values were obtained using two-step Reference Tissue Model(SRTM2), which used k2’ values generated from RTM (k2’=0.05 min−1 forDASB) in the first step and fixed k2’ value in the second step• regional SUV obtained over the last 30mins (50-80mins) time frameTo reduce noise in the network analysis, BPND values instead of SUVs werechosen as the quantitative measurement of choice. In the next chapter, we comparedthese quantitative measurements for DASB tracer to 1) validate DASB parametricBPND algorithm and 2) to choose the best quantitative measurement for networkanalysis.36Chapter 3DASB Parametric ValidationIn this chapter, we validated the SRTM2 method for DASB tracer by comparingglobal k2’ values in the first step of SRTM2 with reported k2’ values in literatureand comparing parametric BPND values with regional BPND values obtained usingSRTM.3.1 MethodsTo calculate parametric BPND values for DASB data using SRTM2, we need to fixa global k2’ value for the second step of the kinetic model as discussed previously.There are two ways to find the global k2’ value:• k2’ value for each pixel of the brain image was estimated using SRTM firstand then averaged outside the reference region to obtain the global k2’ valuefor each subject [25]• set k2’ value based on literature, which was reported to be 0.056min−1 forDASB tracer from one-tissue-compartment kinetic modeling [57] [26] [58]To validate the parametric BPND values, we first compared the k2’ values ob-tained from the first step of SRTM for 7 healthy control subjects to check thevariation and agreement of the calculated k2’ values with reported values. In thesecond step of SRTM2, after fixing k2’, BPND values were estimated using two-parameter fit. We then compare the regional BPND values obtained from SRTMwith the averaged parametric BPND values in each ROI.3.2 Results3.2.1 k2’ ValuesThe calculated k2’ values for 7 control subjects had an average of 0.052±0.017min−1as shown in Figure3.1. The calculated k2’ values agree with the reported true373.3. Discussionk2’ value 0.056min−1 from one-tissue-compartment kinetic parameter values usingcerebellum as the reference region [57].Figure 3.1: estimated k2’ values for each subject compared to reported k2’ value asshown with red line3.2.2 BPND ValuesTo validate SRTM2 parametric BPND values, we compared the regional values ob-tained from SRTM and averaged values from SRTM2 in each ROI as shown inFigure3.2, and averaged regional and parametric values for each subject as shownin Figure3.3.When correlating regional BPND from SRTM and SRTM2, we found a significantcorrelation between average SRTM BPND values and averaged parametric BPNDvalues in each brain region (R2=0.967) as shown in Figure3.4.3.3 Discussionk2’ values from SRTM2 showed excellence agreement with reported k2’ value andaveraged parametric BPND values also showed excellent agreement with regionalBPND values from SRTM. Parametric BPND values can be used for network analysis383.3. DiscussionFigure 3.2: Averaged BPND values from SRTM and SRTM2 in each ROIat voxel level and to explore tracer gradient inside a ROI. However, to reduce noise,network analysis in this project focused on 20 pre-defined ROIs instead of individualvoxel. The analysis can be extended to voxel level in the future.393.3. DiscussionFigure 3.3: Averaged BPND values from SRTM and SRTM2 in each subjectFigure 3.4: Correlation between SRTM and SRTM2 BPND values in each ROI.Equation of the linear regression model is shown in the Figure40Chapter 4SSM Pattern Analysis inSerotonergic System4.1 MethodsTo identify network abnormalities, we used a region-based network modeling ap-proach, the Scaled Subprofile Model (SSM), defined previously [30]. This approachis based on multivariate principal component analysis (PCA) and models the sourcesof subject and region variation as spatially distributed networks in functional im-ages [28]. SSM is able to identify a group-dependent, region-specific, disease specificspatial covariance patterns in the brain that can be used study the heterogeneousregional interactions in different patient groups and to discriminate patients fromhealthy controls.[28] [29]In this study, we applied SSM on BPND values in 20 brain regions to identify sig-nificant regional covariance networks which best distinguish between subject groupsbased on individual expression of the principal components (PC).4.1.1 Pattern IdentificationTo limit the analysis to a subset of PCs related to the disease, individual PCs mustaccount for at least 5% of the total subject by region variability in the data. Toimprove the stability of the selected PCs and the estimated weight for each PC inthe logistic regression model, bootstrap resampling was performed 1000 times toobtain the frequency histogram of PCs entering the regression model.In the case of identifying disease-related pattern using HC and PD subjects, only7 out of 28 PCs made the 5% cutoff and were used to generate the PC histogram(Fig4.1). The scree test was then used to determine the frequency cutoff to choosethe optimal number of PCs to include in the model.After selecting the subset of PCs, frequency histograms of the estimated model414.1. MethodsFigure 4.1: Frequency Histogram of the included PCs. Only 7 PCs account forgreater than 5% of total variance in the data. The first 3 PCs with the highestfrequency were entered into the logistic regression model to obtain the combineddisease-related pattern.parameters of the selected PCs were used to determine the weights of each PC.Subject scores for each selected PC were then entered singly or in linear combinationinto a forward stepwise logistic regression models (Matlab scripts, Mathworks). Thecombination of PCs with the lowest Akaike Information Criterion (AIC) score [36]was selected to yield a single disease/mutation-related covariance pattern. Thismodel was used to estimate the corresponding weights (coefficients) on each pattern,that in linear combination, best discriminated between subject groups. In this case,histograms of model coefficients of the selected PCs gave the weights of 0.174, 0.267and 0.232 for PC2, PC4 and PC5 respectively.The final disease/mutation-related spatial covariance pattern was a linear com-bination of the regional weights of the selected PCs. The same coefficients werethen applied to subject scores for the three PCs to compute the combined SPDRPsubject score for each individual subject. Regional weights for this specific combi-nation of PCs were Z-thresholded at 1 to select significant regions contributing tothe corresponding covariance pattern. Network expression for new subjects or testset in validation was computed using the projection of the subject data onto thecorresponding spatial maps. This process, the topographic profile rating (TPR),was defined previously.424.2. ResultsFigure 4.2: Frequency Histogram of the model parameters of the included PCs inthe logistic regression model. The mean of each frequency histogram was used asthe coefficient for each corresponding PC in the logistic regression model.4.1.2 ValidationTo validate the disease/mutation-related spatial covariance pattern, we preformed5-fold cross-validation with 1000 iterations. All subjects in the analysis were dividedinto 5 groups; 4 groups were used as training set to obtain the pattern; 1 group wasused as test set to examine the robustness of the pattern by calculating individualsubject expression of this specific pattern. The sensitivity and specificity of eachdiscrimination were determined using Receiver Operating Curve algorithm (ROC).4.2 Results4.2.1 Absolute BPND ValuesBefore applying PCA to the data, we performed group analysis on the mean BPNDvalues in all 20 regions in all 4 subject groups to validate the choice of the ROIs andget a sense of possible regions might appearing in the covariance brain pattern.Looking at BPND values, there was a significant decrease in BPND values incaudate, amygdala and putamen in PD compared to HC subjects. PD subjectsshowed a lower BPND in all 20 regions compared to HC subjects, but there is arelatively smaller decrease in hypothalamus compared to other regions. There wasa significant higher BPND values in ACC, amygdala, hypothalamus and medulla inthe asymptomatic LRRK2 mutation carrier (LRRK2-AMC) subjects compared toHC subjects. There was no significant difference between LRRK2-associated PD(LRRK2-PD) and sporadic PD (sPD) in any brain region.There was a significant age correlation in left hypothalamus, left amygdala andright PPN regions in HC BPND values. We did not observe any significant cor-434.2. ResultsFigure 4.3: Serotonergic Parkinson’s disease-related pattern (SPDRP) identifiedby spatial covariance analysis of DASB PET scans from 17 PD patients. Thispattern was characterized by a relative decreased SERT binding in caudate andputamen, and a covarying increased SERT binding in hypothalamus, hippocam-pus, anterior cingulate (ACC), amygdala, and medulla. Only regions that signifi-cantly contributed to the network at Z>1. Regions with positive weights (increasedbinding) are colour-coded red; those with negative weights (decreased binding) arecolour-coded blue.relation between BPND values and age or age of onset in any brain region in PDsubjects. With age as a covariate, there was a significant group differences in amyg-dala (p=0.008), caudate (p=0.011), putamen (p=0.043), hypothalamus (p¡0.001)and medulla (p=0.005) between LRRK2-AMC and HC subjects.4.2.2 PD vs HCDisease-specific spatial covariance pattern was derived using BPND values in 20brain regions from 9 HC and 18 PD subjects. Subject scores significantly separatedHCs from PD patients (p<0.0001). The pattern was obtained from PC2, PC4 andPC5, which accounted for 27% of the total variance in the subject by region BPNDdata set. The serotonergic Parkinson’s disease-related pattern (SPDRP) was char-acterized by a relative decreased SERT binding in caudate and putamen, and acovarying increased SERT binding in hypothalamus, hippocampus, anterior cingu-444.2. ResultsFigure 4.4: SPDRP expression in HC, PD and LRRK2-AMC subjects. There was asignificant separation between HC (blue) and PD (red) groups (p¡0.0001). LRRK2-AMC subjects did not show an elevation of this disease-specific pattern.late (ACC), amygdala, and medulla. Only regions that significantly contributed tothe network at Z>1.Correlation with Clinical MeasurementsThere was an almost significant negative correlation between SPDRP expression(subject scores) and age of onset (p=0.0579) as shown in Fig.4.5. There was alsoan almost significant positive correlation between SPDRP expression and UPDRSmotor scores (p=0.0558) as shown in Fig.4.6. No correlation was observed for diseaseduration or age.SPDRP Expression in Asymptomatic LRRK2 Mutation CarriersLRRK2-AMC subjects did not show an elevated expression of SPDRP compared toHC subjects (p=0.14).4.2.3 Asymptomatic LRRK2 Mutation Carrier vs HCAsymptomatic LRRK2 mutation-related spatial covariance pattern (LRRK2-AMRP)was derived using BPND values in the same 20 brain regions from 9 HC and 9 asymp-454.2. ResultsFigure 4.5: SPDRP expression vs age of disease onset in PD subjects. SPDRPexpression in PD subjects showed an almost significant correlation with age of onset(p=0.0579).Figure 4.6: SPDRP expression vs UPDRS motor score in PD subjects. SPDRPexpression in PD subjects showed an almost significant correlation with UPDRS(p=0.0558).tomatic LRRK2 mutation carriers (LRRK2-AMC). Subject scores were significantlyhigher in LRRK2-AMC subjects compared to HC (p<0.0001). The pattern was ob-tained from PC3, PC1, PC2 and PC5. The resulting LRRK2-AMRP was comprisedof a relatively decreased binding in putamen, and relatively preserved binding in hy-pothalamus, amygdala, PPN, midbrain, substantia nigra (SN) and medulla.464.2. ResultsFigure 4.7: Serotonergic asymptomatic LRRK2 mutation-related spatial covariancepattern (LRRK2-AMRP) identified by spatial covariance analysis of DASB PETscans from 9 LRRK2-AMC subjects. This pattern was characterized by a relativedecreased SERT binding in putamen, and a covarying increased SERT binding inhypothalamus, amygdala, midbrain, PPN, SN and medulla. Only regions that signif-icantly contributed to the network at Z>1. Regions with positive weights (increasedbinding) are colour-coded red; those with negative weights (decreased binding) arecolour-coded blue.Figure 4.8: LRRK2-AMRP expression in HC, PD and LRRK2-AMC subjects.There was a significant separation between HC (blue) and PD (red) groups(p¡0.0001). LRRK2-AMC subjects did not show an elevation of this disease-specificpattern.474.2. ResultsLRRK2-AMRP Expression in PD subjectsPD subjects did not show an elevated expression of LRRK2-AMRP compared toHC subjects (p=0.65).ValidationWe preformed 5-fold cross validation with 1000 iterations to confirm the obtainedpattern. In each iteration, subjects were divided randomly into five subsets: foursubsets were used to train the model to obtain the covariance pattern, and theremaining one group with each subset containing approximately equal number ofmembers from different groups was used to test the accuracy of the classification.Receiver Operator Curve (ROC) was used to examine the accuracy and speci-ficity of the pattern. Area under curve (AUC) for training set is 0.98 and 0.62 fortesting set for the PD vs HC pattern (SPDRP).Figure 4.9: ROC for training set for 1000 iterations.Figure 4.10: ROC for testing set for 1000 iterations484.3. Discussion4.3 DiscussionIn this study, we applied network analysis to DASB PET data from sporadic PD(sPD), LRRK2-associated PD (LRRK2-PD) and asymptomatic LRRK2 mutationcarriers (LRRK2-AMC) to identify a novel disease or mutation-related spatial co-variance pattern in the serotonergic pathways.4.3.1 Use of PCA to Identify Disease/Mutation-RelatedTopographiesFunctional imaging techniques allow us to quantify brain activity to study patho-physiology of neurodegenerative disorders, but absolute tracer binding or metabolicactivity may not provide the complete picture. In addition, variability in subjectsand brain regions increases the difficulty in quantitative analysis. Comparing totraditional group analysis using BPND amplitudes, pattern analysis is proven to bemore robust and sensitive to small change in brain physiology.Studies have suggested that cognitive processes depend on interactions amongdistributed brain regions, which are characterized by brain connectivity [59]. Thescaled subprofile model (SSM) employed in this analysis was able to examine thesubject by region interactions in SERT binding, while eliminating global and region-specific effects in the data. Before applying PCA, BPND data was double-centeredto obtain the residual regional BPND values which contain relevant biological infor-mation independent of the global mean.A more detailed review of the mathematical principles and basic assumptionsunderlying this method was discussed previously [60].4.3.2 ROI-based Network AnalysisAll the 20 ROIs in this analysis were chosen based on prior knowledge about theserotonergic system. Choice of ROIs were also confirmed by analyzing the absoluteBPND values between groups. In this project, network analysis was done on ROI-level only to 1) reduce noise compared to BPND values and 2) compare to absoluteregional BPND values. Future analysis can be extended to voxel level when no priorknowledge or hypothesis is present.494.3. Discussion4.3.3 PC1For FDG PDRP, the PC1 was found to best separate between subject groups andno other PCs were entered into the logistic regression model. In our case, PC1was mainly contributed by the noise in the ROIs. Regions contributed the most toPC1 included midbrain, PPN, SN and pons, which were also shown to have highernoise in raw data compared to other regions. TACs in these regions have the largestvariation, which results in large variance in BPND in these regions.4.3.4 Pattern Identification and ValidationROIs contributing to SPDRP or LRRK2-AMRP were selected based on a pre-definedthreshold (Z-transformed regional weights are greater than 1). We note that con-tributions from two hemispheres were asymmetrical in some regions, with relativelygreater involvement of one hemisphere or the other. This is likely attributed to thenoise in the BPND values before the PCA analysis.ValidationIn 5-fold cross validation, accuracy for test set is not optimal (AUC=0.62). Thisrelatively low AUC was mainly due to high false positive rate in the classification ofHC subjects. However there is good accuracy in the classification of PD subjects.Due to the imbalance in the number of HC and PD subjects and the low number ofsubjects in the analysis, the logistic model can suffer the overfitting problem. Waysto improve AUC for the test set include 1) reducing the number of PCs includedin the regression model, 2) balancing the number of subjects in two groups, and 3)including more subjects into the analysis.4.3.5 Comparison with Other Network AnalysisFor PD subjects, they showed a relatively decreased SERT binding in caudate (mostsignificant) and putamen, and relatively preserved SERT binding in hypothalamus(most significant), hippocampus, medulla, dentate nucleus, anterior cingulate andamygdala compared to HC. Here, we compared SPDRP obtained from PD vs HCsubjects to common network analysis to examine if there is any similarities betweenthese networks.504.3. DiscussionFDG/PET PDRP and PDCPPCA-based covariance analysis has been previously applied to FDG/PET data asdiscussed before in the Introduction Chapter. The Parkinson’s Disease RelatedPattern (PDRP) and Parkinson’s Disease Cognitive Pattern (PDCP) were validatedacross different patient populations.PDRP as previously defined using FDG PET was characterized by hyperme-tabolism in the thalamus, globus pallidus (GP), pons, and primary motor cortex,with associated relative metabolic reduction in the lateral premotor and posteriorparietal areas. PDCP defined also with FDG PET was characterized by relativelyincreased activity in the cerebellar vermis and dentate nuclei, with associated de-creased activity in frontal and parietal association areas [30].No direct correlation was found between SPDRP and FDG PDRP or FDGPDCP. We do not expect to see a close connection between these distinct patterns,because DASB specifically targets the serotonergic pathway while FDG targets brainmetabolic activities.Resting-State fMRI ConnectivitySPDRP also did not resemble any of the known resting-state fMRI connectivitynetworks. This is also expected since resting-state fMRI looks at brain activationinstead of any specific neurological pathways in the brain.However, by applying network analysis (such as Independent Component Anal-ysis (ICA))) to resting-state fMRI images for the same subjects, we can incorporatethe serotonergic network with functional activation network. Advance statisticalanalysis (such as joint ICA) can be applied to both DASB and resting-state fMRIdata together to examine the intrinsic network underlying both modalities.4.3.6 Possible Functional Basis for SPDRP and LRRK2-AMRPTopographyIn this section, we look into the regions involved in SPDRP and LRRK2-AMRPand try to link them with their functional roles in the brain related to the disease.514.3. DiscussionRegions with Decreased BindingCompared to the localized SERT binding reduction (as measured by absolute BPNDvalues in each ROI), PD subjects showed a significantly lower SERT binding com-pared to HC subjects in caudate and putamen, which were also shown to have arelative decreased binding in SPDRP.Previous studies showed that PD subjects had a significant decreased absoluteSERT binding in caudate (30%) and putamen (26%) compared to HC subjects[61] [62]). Unlike in the dopaminergic system where posterior putamen is moreaffected by the disease, there is a preferential loss of SERT in caudate [62] whichis consistent with the greater relative decreased binding in caudate compared toputamen in SPDRP.Regions with Increased BindingRegions with significantly higher absolute BPND values (ACC, amygdala, hypotha-lamus and medulla) in LRRK2-AMC compared to HC subjects all showed a relativepreserved binding in SPDRP. This may indicate that upregulated regions in SP-DRP are affected before motor symptoms onset and may act as a compensatorymechanism in the serotonergic system.According to Braak hypothesis of PD brain pathological staging, Lewy body andneurite deposition occur in raphe nuclei (in brainstem) at stage 2; substantia nigra(SN) and midbrain are affected at stage 3 where clinical motor symptoms start tooccur [39].Both SN and midbrain showed upregulated binding in LRRK2-AMRP, but notin SPDRP. One study suggested SERT loss in striatum regions of PD precedes theloss in midbrain region [63], which may indicate the upregulation in midbrain and SNtries to compensate the loss of SERT in the striatum regions in the presymptomaticstage of the disease.Regions with relative preserved binding were also shown to be involved in cogni-tive impairments of PD. For PD patients with depression, hypothalamus and amyg-dala showed a higher SERT binding than PD patients without depression [? ][49].These two regions both showed upregulated binding in SPDRP and LRRK2-AMRP.PD patients with abnormal BMI changes showed significantly higher SERT bindingin hypothalamus compared to cases with no significant BMI changes [50]. Thesefindings may suggest upregulation of SERT function starting from presymptomatic524.3. Discussionstages of the disease may be related to depression and BMI changes in PD patients.Comparing SPDRP with LRRK2-AMRP, putamen was the only region withrelative decreased binding in LRRK2-AMRP. This may indicate that even thoughcaudate is more affected after motor symptom onset, putamen is still affected earlierby the disease in the presymptomatic stage.In the presymptomatic LRRK2-AMRP, there are more regions with relativelyincreased binding than in SPDRP. This may indicate a stronger upregulation in thepresymptomatic stage compared to symptomatic stage of the disease. This strongerupregulation may act as a compensatory mechanism in the serotonergic system, butthis over activation of serotonergic system may lead to various non-motor symptomsafter disease onset.4.3.7 Correlation with Clinical MeasurementsWe observed an almost significant correlation between SPDRP expression and UP-DRS motor scores, which indicate more sever patients (in terms of motor disability)may have a higher expression of SPDRP.We also observed an almost significant negative correlation between SPDRPexpression and age of disease onset, meaning that younger onset patients have ahigher expression of SPDRP. This finding can be combined with the DA releasefindings discussed in the next chapter to examine the possible relationship betweenserotonergic network.53Chapter 5Levodopa-Induced DopamineRelease in PDIn the second part of the project, we analyzed the drug-induced DA release patternon early PD subjects (less than 5 years of disease duration) using double RAC data.The combination of abnormal drug-induced DA release pattern and upregulation ofthe serotonergic system may be able to explain the occurrence of treatment-inducedcomplications in PD patients.5.1 ObjectivesIn this study, we investigated the levodopa (LDOPA)-induced DA release pattern for10 early (less than 5 years disease duration) sporadic PD subjects using double [11C]-RAC scans to examine the relationships between DA release pattern, DA denervationand SERT binding.Since we already observed some regions with upregulated SERT binding in PDsubjects, we want to investigate if the combination of abnormal drug-induced DArelease pattern and upregulation of the serotonergic system may be able to explainthe occurrence of treatment-induced complications in PD patients.5.2 Methods5.2.1 Subjects and Quantitative MeasurementsWe examined the correlations between LDOPA-induced DA release (as measuredby double RAC scans), DTBZ BPND values and DASB BPND values for 10 earlysporadic PD subjects.DA release was calculated as RAC BPND values without drug minus RAC BPNDvalues with drug. Unified Parkinson’s Disease Rating Scale (UPDRS) was measured545.2. MethodsNo.Subjects Age Disease Duration UPDRS on drug UPDRS off drug10 57.8± 8.7 years 2.1± 1.4 years 10.8± 5.4 13.7± 4.7Table 5.1: Clinical information for 10 early sporadic PD subjectswith and without drug to evaluate the severity motor symptoms in each subject.The change in UPDRS was defined as the difference between UPDRS off and ondrug, which should correspond to the effectiveness of the drug on individual subject.BPND values were extracted in 3 striatum regions (caudate, ventral striatumand putamen) using either Logon or RTM method with cerebellum used as thereference region. The 3 striatum regions were further divided into the most andleast affected sides for each subject, which was defined using the averaged DTBZBPND in the putamen regions (i.e. the side with lower DTBZ putamen BPNDhad severer degradation of dopamine-producing neurons, so was defined as the mostaffected side).5.2.2 Regression ModelsWe used linear and multiple regression analysis to examine the relationship betweenDA release, DTBZ BPND values, DASB BPND values and clinical measurementsin each striatum region. Detailed regression method is as follows:1. DA release values were regressed on each of the possible explanatory variables(predictors) separately (linear regression)• DTBZ BPND• DASB BPND• RAC baseline BPND• Age of onset• Change in UPDRS (UPDRS off drug − UPDRS on drug)• disease duration• gender2. Multiple regression was applied on all predictors which survived a cut-off cri-terion from the linear regression• Cut-off criterion: correlation p-value < 0.3555.3. ResultsCorrelations between DA release and all the above variables were examined ineach of the striatum regions for all subjects.5.3 Results5.3.1 Correlations in Striatum RegionsMost Affected PutamenIn the most affected putamen region, DA release showed a significant correlation withage of onset and change in UPDRS (p=0.013 and p=0.049) in linear regression asshown in Figure5.1 and Figure5.2. No variable was significant in multiple regressionafter correcting for other variables.Figure 5.1: Correlation between DA release and age of onset for sPD subjects inmost affected putamen region.Least Affected PutamenIn the least affected putamen region, DA release correlated significantly with age ofonset and RAC baseline BPND values (p=0.030 and p=0.012). Age of onset andRAC baseline BPND values increased significance level when entered in multipleregression (p=0.00060 and p=0.00029) after correcting for each other as shown inFigure5.3 and Figure5.4.565.3. ResultsFigure 5.2: Correlation between DA release and change in UPDRS for sPD subjectsin most affected putamen region.Figure 5.3: Correlation between DA release and RAC baseline BPND values aftercorrecting for age of onset for sPD subjects in least affected putamen region.CaudateNo correlation was found between DA release and any possible explanatory variables.However, there was a significant correlation between RAC baseline BPND valuesand age of onset for both the most and the least affected Caudate (p=0.0077 andp=0.015) as shown in Figure5.5.575.3. ResultsFigure 5.4: Correlation between DA release and age of onset after correcting forRAC baseline BPND values for sPD subjects in least affected putamen region.Figure 5.5: Correlation between RAC baseline BPND values and age of onset forsPD subjects in caudate region.5.3.2 Disease SeverityDisease severity can be defined by either UPDRS off medication or DTBZ BPNDvalues. DA release did not correlate with either disease severity measures.In the 10 early PD patients, UPDRS off drug did not depend on age (p=0.067)or age of onset (p=0.12). DA depletion (defined by DTBZ BPND values) did notdepend on age (p=0.66), age of onset (p=0.8), UPDRS off (p=0.52), change inUPDRS or disease duration.585.4. Discussion5.3.3 Correlation with Other TracersNo significant correlation was found between DA release and DTBZ or DASB BPNDvalues in any of the striatum regions.5.4 Discussion5.4.1 Correlations with Disease SeverityStudies have shown that age of onset did not influence the absolute severity ofnigrostriatal damage as measured by DTBZ BPND values [46][65], which agreeswith the fact that we did not see any significant correlation between DTBZ BPNDvalues and age or age of onset in early PD subjects.5.4.2 DA Release Correlation with Age of Onset and Change inUPDRSOur results suggest that, for early PD patients, younger onset patients have an in-creased DA release in response to LDOPA stimuli and a trend towards better motorresponse to the medication. This was indicated by a strong negative correlationbetween DA release and age of onset in the putamen region after adjusting for RACBPND at baseline , and a trend of positive correlation between DA release andchange in UPDRS motor scores.Younger onset patients have a higher DA release while motor symptoms severity(as measured by UPDRS motor scores) and DA depletion (as measured by DTBZBPND values) remain relatively the same compared to older onset patients. Thisfinding implies that DA release in younger onset patients undergoes larger alter-ation, which results in larger swing in synaptic DA levels. This large swing orimbalance may contribute to greater risk of motor fluctuations, which may explainage-dependent occurrence of complications. We are following these PD subjectsclinically (follow-up 3 years after first scan) to see if they would develop treatment-induced motor complications.These results will be validated by including more subjects into the study.595.4. Discussion5.4.3 Relationship with Serotonergic SystemAs shown before in Chapter 4, there was an almost significant negative correlationbetween SPDRP expression and age of onset for PD subjects.5-HT terminals also participate in DA re-uptake (Berger 1978) and can me-tabolize L-DOPA into DA. DA released in the striatum by 5-HT terminals acts asfalse neurotransmitter which contributes to L-DOPA-induced dyskinesia (LID). Thedecline in SERT levels i the striatum of PD precedes the decline in midbrain andother regions, so we see a relative preserved binding in some brain regions. Theupregulation in these serotonergic projection regions may act as a compensatorymechanism in the serotonergic pathway for the loss of dopamine-producing neuronsin early stages of the disease.60Chapter 6Conclusions and Future WorkIn this project, we used a PCA-based network analysis to investigate whether a dis-ease or a LRRK2 mutation specific spatial covariance pattern exists in the serotoner-gic system. A disease-related SPDRP and a mutation-related LRRK2-AMRP werefound. Brain regions with a relatively preserved binding may act as presymptomaticmarker and/or compensatory mechanism for the disease, which also may link to non-motor symptoms of the disease. We also investigated the altered medication-inducedDA release pattern in early PD subjects. The combination of abnormal medication-induced DA release pattern and upregulation of the serotonergic system may beable to explain the occurrence of treatment-induced complications and non-motorsymptoms in PD patients, and act as a potential early marker for the disease.As discussed before, more subjects will be included in the analysis to confirmthe results and improve statistical power. Patients involved in DA release study willhave follow-ups to check if medication-induced motor complications occur. Networkanalysis can also be extended to voxel level after reducing noise in the parametricBPND images. Other network analysis can also be applied to combine differentimaging modalities (PET and fMRI) or different PET tracers together to gain betterunderstanding of the disease.61Bibliography[1] T. F. Massoud and S. S. Gambhir, “Molecular imaging in living subjects: seeingfundamental biological processes in a new light,” Genes & development, vol. 17,no. 5, pp. 545–580, 2003.[2] C. S. Levin and E. J. Hoffman, “Calculation of positron range and its effect onthe fundamental limit of positron emission tomography system spatial resolu-tion,” Physics in medicine and biology, vol. 44, no. 3, p. 781, 1999.[3] R. B. Innis, V. J. Cunningham, J. Delforge, M. Fujita, A. Gjedde, R. N. Gunn,J. Holden, S. Houle, S.-C. Huang, M. Ichise, et al., “Consensus nomenclature forin vivo imaging of reversibly binding radioligands,” Journal of Cerebral BloodFlow & Metabolism, vol. 27, no. 9, pp. 1533–1539, 2007.[4] M. D. Walker, Quantitative dynamic 3D pet scanning of the body and brainusing LSO tomographs. PhD thesis, University of Manchester, 2009.[5] M. Phelps and E. Hoffman, “N. mull ani, c. higgins, m. ter-pogossian,” designconsiderations for a positron emission transaxial tomograph (pet iii),”,” IEEETrans. Biomed. Eng. NS-23, pp. 516–522, 1976.[6] J. A. Fessler and J. M. Ollinger, “Signal processing pitfalls in positron emissiontomography,” IEEE Signal Processing Magazine, vol. 14, no. 1, pp. 43–55, 1997.[7] N. E. Bolus, R. George, B. R. Newcomer, et al., “Pet/mri: the blended-modalitychoice of the future?,” Journal of nuclear medicine technology, vol. 37, no. 2,pp. 63–71, 2009.[8] A. A. Lammertsma and S. P. Hume, “Simplified reference tissue model for petreceptor studies,” Neuroimage, vol. 4, no. 3, pp. 153–158, 1996.[9] B. Capuano, I. Crosby, and E. Lloyd, “Schizophrenia: genesis, receptorologyand current therapeutics,” Current medicinal chemistry, vol. 9, no. 5, pp. 521–548, 2002.62Bibliography[10] K. Leuner and W. Mu¨ller, “The complexity of the dopaminergic synapses andtheir modulation by antipsychotics.,” Pharmacopsychiatry, vol. 39, pp. S15–20,2006.[11] M. S. Sodhi and E. Sanders-Bush, “Serotonin and brain development,” Inter-national review of neurobiology, vol. 59, pp. 111–174, 2004.[12] D. W. Townsend, Z. Cheng, D. Georg, W. Drexler, and E. Moser, “Grandchallenges in biomedical physics,” Frontiers in Physics, vol. 1, no. 1, 2013.[13] D. Delbeke, “Oncological applications of fdg pet imaging: brain tumors, col-orectal cancer lymphoma and melanoma,” The Journal of Nuclear Medicine,vol. 40, no. 4, p. 591, 1999.[14] S. M. Ametamey, M. Honer, and P. A. Schubiger, “Molecular imaging withpet,” Chemical reviews, vol. 108, no. 5, pp. 1501–1516, 2008.[15] R. D. Evans and A. Noyau, The atomic nucleus, vol. 582. McGraw-Hill NewYork, 1955.[16] H. E. Johns, Physics of radiology. Charles River Media, 1983.[17] D. L. Bailey, Quantification in 3D positron emission tomography. PhD thesis,University of Surrey, 1996.[18] C. Thompson, A. Dagher, D. Lunney, S. Strother, and A. Evans, “A techniqueto reject scattered radiation in pet transmission scans,” in Physics and Engi-neering of Computerized Multidimensional Imaging and Processing, pp. 244–253, International Society for Optics and Photonics, 1986.[19] M. Xu, P. Cutler, and W. Luk, “Adaptive, segmented attenuation correctionfor whole-body pet imaging,” IEEE Transactions on Nuclear Science, vol. 43,no. 1, pp. 331–336, 1996.[20] J. M. Ollinger and J. A. Fessler, “Positron-emission tomography,” IEEE SignalProcessing Magazine, vol. 14, no. 1, pp. 43–55, 1997.[21] Y. Vardi, L. Shepp, and L. Kaufman, “A statistical model for positron emissiontomography,” Journal of the American Statistical Association, vol. 80, no. 389,pp. 8–20, 1985.63Bibliography[22] J. Qi and R. H. Huesman, “Theoretical study of lesion detectability of mapreconstruction using computer observers,” IEEE transactions on medical imag-ing, vol. 20, no. 8, pp. 815–822, 2001.[23] A. Alessio and P. Kinahan, “Pet image reconstruction,” Nuclear medicine,vol. 2, 2006.[24] D. Borys, K. Szczucka-Borys, and K. Gorczewski, “System matrix computationfor iterative reconstruction algorithms in spect based on direct measurements,”International Journal of Applied Mathematics and Computer Science, vol. 21,no. 1, pp. 193–202, 2011.[25] R. N. Gunn, A. A. Lammertsma, S. P. Hume, and V. J. Cunningham, “Para-metric imaging of ligand-receptor binding in pet using a simplified referenceregion model,” Neuroimage, vol. 6, no. 4, pp. 279–287, 1997.[26] Y. Wu and R. E. Carson, “Noise reduction in the simplified reference tissuemodel for neuroreceptor functional imaging,” Journal of Cerebral Blood Flow& Metabolism, vol. 22, no. 12, pp. 1440–1452, 2002.[27] J. Logan, J. S. Fowler, N. D. Volkow, A. P. Wolf, S. L. Dewey, D. J. Schlyer,R. R. MacGregor, R. Hitzemann, B. Bendriem, S. J. Gatley, et al., “Graphicalanalysis of reversible radioligand binding from timeactivity measurements ap-plied to [n-11c-methyl]-(-)-cocaine pet studies in human subjects,” Journal ofCerebral Blood Flow & Metabolism, vol. 10, no. 5, pp. 740–747, 1990.[28] J. Moeller and S. Strother, “A regional covariance approach to the analysis offunctional patterns in positron emission tomographic data,” Journal of CerebralBlood Flow & Metabolism, vol. 11, pp. A121–A135, 1991.[29] G. E. Alexander and J. R. Moeller, “Application of the scaled subprofile modelto functional imaging in neuropsychiatric disorders: a principal componentapproach to modeling brain function in disease,” Human Brain Mapping, vol. 2,no. 1-2, pp. 79–94, 1994.[30] D. Eidelberg, “Metabolic brain networks in neurodegenerative disorders: afunctional imaging approach,” Trends in neurosciences, vol. 32, no. 10, pp. 548–557, 2009.64Bibliography[31] D. Eidelberg, J. Moeller, V. Dhawan, P. Spetsieris, S. Takikawa, T. Ishikawa,T. Chaly, W. Robeson, D. Margouleff, S. Przedborski, et al., “The metabolictopography of parkinsonism,” Journal of Cerebral Blood Flow & Metabolism,vol. 14, no. 5, pp. 783–801, 1994.[32] C. Lozza, J.-C. Baron, D. Eidelberg, M. J. Mentis, M. Carbon, and R.-M. Marie´,“Executive processes in parkinson’s disease: Fdg-pet and network analysis,”Human brain mapping, vol. 22, no. 3, pp. 236–245, 2004.[33] T. Eckert and D. Eidelberg, “Neuroimaging and therapeutics in movementdisorders,” NeuroRx, vol. 2, no. 2, pp. 361–371, 2005.[34] C. Huang, P. Mattis, C. Tang, K. Perrine, M. Carbon, and D. Eidelberg,“Metabolic brain networks associated with cognitive function in parkinson’sdisease,” Neuroimage, vol. 34, no. 2, pp. 714–723, 2007.[35] S. Wold, K. Esbensen, and P. Geladi, “Principal component analysis,” Chemo-metrics and intelligent laboratory systems, vol. 2, no. 1, pp. 37–52, 1987.[36] H. Akaike, “An information criterion (aic),” Math Sci, vol. 14, no. 153, pp. 5–9,1976.[37] L. M. de Lau and M. M. Breteler, “Epidemiology of parkinson’s disease,” TheLancet Neurology, vol. 5, no. 6, pp. 525–535, 2006.[38] T. Wichmann and M. R. DeLONG, “Pathophysiology of parkinson’s disease:the mptp primate model of the human disorder,” Annals of the New YorkAcademy of Sciences, vol. 991, no. 1, pp. 199–213, 2003.[39] H. Braak, K. Del Tredici, U. Ru¨b, R. A. de Vos, E. N. J. Steur, and E. Braak,“Staging of brain pathology related to sporadic parkinsons disease,” Neurobi-ology of aging, vol. 24, no. 2, pp. 197–211, 2003.[40] B. Dubois and B. Pillon, “Cognitive deficits in parkinsons disease,” Journal ofneurology, vol. 244, no. 1, pp. 2–8, 1996.[41] A. Lees and E. Smith, “Cognitive deficits in the early stages of parkinson’sdisease,” Brain, vol. 106, no. 2, pp. 257–270, 1983.65Bibliography[42] M. Poletti and U. Bonuccelli, “Acute and chronic cognitive effects of levodopaand dopamine agonists on patients with parkinsons disease: a review,” Thera-peutic advances in psychopharmacology, vol. 3, no. 2, pp. 101–113, 2013.[43] R. Cools, “Dopaminergic modulation of cognitive function-implications for l-dopa treatment in parkinson’s disease,” Neuroscience & Biobehavioral Reviews,vol. 30, no. 1, pp. 1–23, 2006.[44] A. J. Stoessl, “Positron emission tomography in premotor parkinson’s disease,”Parkinsonism & related disorders, vol. 13, pp. S421–S424, 2007.[45] C. S. Lee, A. Samii, V. Sossi, T. J. Ruth, M. Schulzer, J. E. Holden, J. Wudel,P. K. Pal, D. La Fuente-Fernandez, D. B. Calne, et al., “In vivo positron emis-sion tomographic evidence for compensatory changes in presynaptic dopamin-ergic nerve terminals in parkinson’s disease,” Annals of neurology, vol. 47, no. 4,pp. 493–503, 2000.[46] R. de la Fuente-Ferna´ndez, V. Sossi, Z. Huang, S. Furtado, J.-Q. Lu, D. B.Calne, T. J. Ruth, and A. J. Stoessl, “Levodopa-induced changes in synapticdopamine levels increase with progression of parkinson’s disease: implicationsfor dyskinesias,” Brain, vol. 127, no. 12, pp. 2747–2754, 2004.[47] V. Sossi, K. Dinelle, G. J. Topping, J. E. Holden, D. Doudet, M. Schulzer, T. J.Ruth, A. J. Stoessl, D. La Fuente-Fernandez, et al., “Dopamine transporterrelation to levodopa-derived synaptic dopamine in a rat model of parkinsons:an in vivo imaging study,” Journal of neurochemistry, vol. 109, no. 1, pp. 85–92,2009.[48] M. Politis and C. Loane, “Serotonergic dysfunction in parkinson’s disease andits relevance to disability,” The Scientific World Journal, vol. 11, pp. 1726–1734, 2011.[49] M. Politis, K. Wu, C. Loane, F. Turkheimer, S. Molloy, D. Brooks, and P. Pic-cini, “Depressive symptoms in pd correlate with higher 5-htt binding in rapheand limbic structures,” Neurology, vol. 75, no. 21, pp. 1920–1927, 2010.[50] M. Politis, C. Loane, K. Wu, D. J. Brooks, and P. Piccini, “Serotonergic medi-ated body mass index changes in parkinson’s disease,” Neurobiology of disease,vol. 43, no. 3, pp. 609–615, 2011.66Bibliography[51] M. Politis, K. Wu, C. Loane, L. Kiferle, S. Molloy, D. J. Brooks, and P. Piccini,“Staging of serotonergic dysfunction in parkinson’s disease: an in vivo 11 c-dasbpet study,” Neurobiology of disease, vol. 40, no. 1, pp. 216–221, 2010.[52] M. Politis, K. Wu, C. Loane, D. J. Brooks, L. Kiferle, F. E. Turkheimer, P. Bain,S. Molloy, and P. Piccini, “Serotonergic mechanisms responsible for levodopa-induced dyskinesias in parkinsons disease patients,” The Journal of clinicalinvestigation, vol. 124, no. 3, pp. 1340–1349, 2014.[53] M. Farrer, J. Stone, I. Mata, S. Lincoln, J. Kachergus, M. Hulihan, K. Strain,and D. Maraganore, “Lrrk2 mutations in parkinson disease,” Neurology, vol. 65,no. 5, pp. 738–740, 2005.[54] M. Farrer and O. A. Ross, “Lrrk2-related parkinson disease,” 2012.[55] D. Wile, K. Dinelle, J. McKenzie, N. Heffernan, M. Adam, Q. Miao, C. Za-betian, Z. Wszolek, J. Aasly, M. Farrer, et al., “Striatal and cortical elevationsin serotonin transporter binding precede motor onset in asymptomatic patientswith lrrk2 gene mutations,” in MOVEMENT DISORDERS, vol. 30, pp. S26–S26, WILEY-BLACKWELL 111 RIVER ST, HOBOKEN 07030-5774, NJ USA,2015.[56] V. Sossi, R. Nandhagopal, D. Wile, M. Schulzer, J. McKenzie, K. Dinelle,M. Farrer, Z. Wszolek, J. Aasly, and A. Stoessl, “Increased dopamine turnover:A possible contributor to the increased risk of parkinson’s disease in lrrk2 mu-tation carriers,” in MOVEMENT DISORDERS, vol. 30, pp. S24–S25, WILEY-BLACKWELL 111 RIVER ST, HOBOKEN 07030-5774, NJ USA, 2015.[57] M. Ichise, J.-S. Liow, J.-Q. Lu, A. Takano, K. Model, H. Toyama, T. Suhara,K. Suzuki, R. B. Innis, and R. E. Carson, “Linearized reference tissue para-metric imaging methods&colon; application to &lsqb; 11c&rsqb; dasb positronemission tomography studies of the serotonin transporter in human brain,”Journal of Cerebral Blood Flow & Metabolism, vol. 23, no. 9, pp. 1096–1112,2003.[58] W. G. Frankle, M. Slifstein, R. N. Gunn, Y. Huang, D.-R. Hwang, E. A. Darr,R. Narendran, A. Abi-Dargham, and M. Laruelle, “Estimation of serotonintransporter parameters with 11c-dasb in healthy humans: reproducibility and67Bibliographycomparison of methods,” Journal of nuclear medicine, vol. 47, no. 5, pp. 815–826, 2006.[59] O. Sporns, “Network attributes for segregation and integration in the humanbrain,” Current opinion in neurobiology, vol. 23, no. 2, pp. 162–171, 2013.[60] J. R. Moeller and C. G. Habeck, “Reciprocal benefits of mass-univariate andmultivariate modeling in brain mapping: applications to event-related func-tional mri, h215o-, and fdg-pet,” International journal of biomedical imaging,vol. 2006, 2006.[61] M. Guttman, I. Boileau, J. Warsh, J. Saint-Cyr, N. Ginovart, T. McCluskey,S. Houle, A. Wilson, E. Mundo, P. Rusjan, et al., “Brain serotonin transporterbinding in non-depressed patients with parkinson’s disease,” European journalof neurology, vol. 14, no. 5, pp. 523–528, 2007.[62] S. J. Kish, J. Tong, O. Hornykiewicz, A. Rajput, L.-J. Chang, M. Guttman,and Y. Furukawa, “Preferential loss of serotonin markers in caudate versusputamen in parkinson’s disease,” Brain, vol. 131, no. 1, pp. 120–131, 2008.[63] S. E. Kim, J. Y. Choi, Y. S. Choe, Y. Choi, and W. Y. Lee, “Serotonin trans-porters in the midbrain of parkinsons disease patients: a study with 123i-β-citspect,” Journal of Nuclear Medicine, vol. 44, no. 6, pp. 870–876, 2003.[64] K. A. Frey, R. A. Koeppe, M. R. Kilbourn, T. M. Vander Borght, R. L. Albin,S. Gilman, and D. E. Kuhl, “Presynaptic monoaminergic vesicles in parkinson’sdisease and normal aging,” Annals of neurology, vol. 40, no. 6, pp. 873–884,1996.[65] V. Sossi, R. de la Fuente-Ferna´ndez, M. Schulzer, J. Adams, and J. Stoessl,“Age-related differences in levodopa dynamics in parkinson’s: implications formotor complications,” Brain, vol. 129, no. 4, pp. 1050–1058, 2006.68


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items