Vpliv rekonstrukcijskih algoritmov pri slikanju s pozitronsko emisijsko tomografijo na izraženost značilnega presnovnega možganskega vzorca pri parkinsonovi bolezni

Tomše, Petra (author), Trošt, Maja (mentor)

Namen: Značilen presnovni možganski vzorec pri parkinsonovi bolezni (angleško: Parkinson’s Disease Related Pattern - PDRP) je vzorec možganske aktivnosti, ki je specifičen za parkinsonovo bolezen (PB). Določimo ga s statistično analizo slik možganov bolnikov s PB, posnetih s 18F-fluorodeoksiglukozo (FDG) in pozitronsko emisijsko tomografijo (PET). PDRP je slikovni biološki označevalec PB. Izraženost PDRP lahko izračunamo iz slik FDG PET za vsakega preiskovanca posebej. V raziskavi nas je zanimalo, ali vrsta rekonstrukcijskega algoritma slik FDG PET vpliva na izraženost PDRP. Zasnova raziskave, opis metod in preiskovancev: Pri slovenski skupini 40 bolnikov s PB, 40 zdravih preiskovancev (ZP), ter 25 bolnikov z atipičnim parkinsonizmom (AP) smo s kamero Biograph mCT na Kliniki za nuklearno medicino Univerzitetnega kliničnega centra Ljubljana (UKCL) posneli FDG PET možganov. Ameriška skupina 20 bolnikov s PB in 7 ZP je slikanje FDG PET možganov s kamero GE Advance opravila na The Feinstein Institute for Medical Research (FIMR). Slike 20 slovenskih bolnikov s PB in 20 ZP smo uporabili za določitev slovenskega PDRP(SLOV) z multivariatno statistično analizo SSM/PCA (angleško: Scaled Subprofile Model/Principal Component Analysis). Ostale slike smo uporabili za validacijo vzorca, tako da smo preverili, ali izraženost PDRP(SLOV) zanesljivo razlikuje med skupinami preiskovancev, ter ali izraženost PDRP(SLOV) in izvirnega PDRP(FIMR) dobro korelirata. Nato smo raziskali, kakšen vpliv imajo različni rekonstrukcijski algoritmi slik PET na izraženost PDRP(SLOV) in PDRP(FIMR). Proučili smo naslednje rekonstrukcijske algoritme: analitične FBP (angleško: Filtered Backprojection), FBP+TOF (TOF – angleško: Time Of Flight), 3DRP (angleško: 3D Reprojection) in FORE-FBP (FORE – angleško: Fourier rebinning), ter iterativne OSEM (angleško: Ordered Subset Expectation Maximization), OSEM+TOF, OSEM+PSF (PSF – angleško: Point Spread Function), OSEM+PSF+TOF in FORE-Iterative. Rezultati: Določili smo presnovni možganski vzorec PDRP(SLOV), ki se kaže z relativno povišano presnovo v palidumu, putamnu, talamusu, možganskem deblu, malih možganih in senzomotorični skorji, povezano z relativno znižano presnovo v posteriorno parietalni, okcipitalni in frontalni skorji. Z analizo izraženosti PDRP(SLOV) lahko razlikujemo med bolniki s PB in ZP oziroma bolniki z AP (p<0,001). S korelacijo izraženosti PDRP(SLOV) in PDRP(FIMR) (r=0,977, p<0,001) smo pokazali zelo dobro primerljivost med obema mrežnima vzorcema. Potrdili smo, da izraženost PDRP dobro korelira med referenčnim in ostalimi rekonstrukcijskimi algoritmi (r&#8805;0,993, p<0,001). Zaključki: Z raziskavo smo določili PDRP(SLOV) in pokazali, da njegova izraženost dobro razlikuje med bolniki s PB in ZP ter bolniki z AP in da je vzorec dobro primerljiv z izvirnim PDRP(FIMR). Potrdili smo, da različne vrste rekonstrukcijskih algoritmov slik PET nimajo pomembnega vpliva na izraženost PDRP. Rezultati raziskave bodo pripomogli k širši klinični in raziskovalni uporabi tega slikovnega biološkega označevalca PB v svetovnem merilu.

• Language: Slovenian
• Keywords: parkinsonova bolezen, FDG PET, rekonstrukcijski algoritmi, značilni presnovni možganski vzorci
• Work type: Doctoral dissertation
• Organization: MF - Faculty of Medicine
• Year: 2019
• PID: 20.500.12556/RUL-106394
• COBISS.SI-ID: 299087104
• Publication date in RUL: 21.02.2019

Secondary language

• Language: English
• Title: The effect of image reconstruction algorithms in positron emission tomography on the expression of characteristic metabolic brain network for Parkinson’s disease

Abstract

Aim: Parkinson’s disease related pattern (PDRP) is a metabolic brain network characteristic for Parkinson's disease (PD). It is identified with statistical analysis of PD patients’ brain images acquired with 18F-fluorodeoxyglucose (FDG) and positron emission tomography (PET). PDRP is an imaging biomarker of disease process in PD. PDRP expression can be determined prospectively for each individual patient. The aim of our study was to explore whether a type of FDG PET image reconstruction algorithm affects the PDRP expression. Study design, methods and participants: Brain scans of a Slovenian cohort of 40 PD patients, 40 healthy control (HC) participants and 25 patients with atypical parkinsonism (AP) were acquired with FDG PET at Department of nuclear medicine of University Medical Centre Ljubljana (UMCL), using Biograph mCT scanner. An American cohort of 20 PD patients and 7 HC subjects was scanned at The Feinstein Institute for Medical Research (FIMR) using GE Advance camera. Images of 20 Slovenian PD patients and 20 HC were used to identify a Slovenian PDRP(SLOV) with Scaled Subprofile Model/Principal Component Analysis (SSM/PCA). Other images were used for PDRP(SLOV) validation, performed by testing discrimination between subject groups based on the expression of PDRP(SLOV) and correlation between expressions of PDRP(SLOV) and PDRP(FIMR). Afterwards we explored the effect of various PET image reconstruction algorithms on the expression of PDRP(SLOV) and PDRP(FIMR). We studied the following reconstruction algorithms: analytical FBP (Filtered Backprojection), FBP+TOF (TOF = Time of flight), 3DRP (3D Reprojection), FORE-FBP (FORE = Fourier rebinning) and iterative OSEM (Ordered Subset Expectation Maximization), OSEM+TOF, OSEM+PSF (PSF = Point Spread Function), OSEM+PSF+TOF, FORE-Iterative. Results: We determined metabolic brain network PDRP(SLOV), characterized by relative hypermetabolism in pallidum, putamen, thalamus, brain stem, cerebellum and sensory-motor cortex, associated with relative hypometabolism in posterior parietal, occipital and frontal cortex. PDRP(SLOV) discriminates PD, AP and HC subjects (p<0.001). PDRP(SLOV) and PDRP(FIMR) expression correlation (r=0.977, p<0.001) implies very good similarity between the brain networks. We confirmed significant correlation of PDRP expression between the reference and other reconstruction algorithms (r&#8805;0.993, p<0.0001). Conclusions: We identified PDRP(SLOV) and showed that its expression reliably discriminated PD patients from HC and AP subjects. PDRP(SLOV) has good similarity to the original PDRP(FIMR). We confirmed that different types of PET reconstruction algorithms have no significant impact on the expression of PDRP. Our work will contribute to implementation of this imaging biomarker in clinical and research applications worldwide.

• Keywords: Parkinson's disease, FDG PET, reconstruction algorithms, characteristic metabolic brain networks