項目的第三部分檢查了質子治療劑量沉積的精確度，重點是使用中研院PET的模塊化設計以及SiPMs和STiC asic讀出，進行範圍驗證。該研究展示了130 MeV質子輻照PMMA時正子發射體的深度分佈，並將長庚紀念醫院緊湊的32通道設置的測量結果與Geant4模擬進行比較。通過計時能力和不同同位素的多指數擬合分析的結果驗證了解剖變化估計在連續治療期間的實施。此外，我們介紹了AS-PET的模塊化設計及其模擬成像性能，突出其在質子治療範圍驗證中的潛力。這項論文研究可以幫助改善平板PET系統的PET圖像質量，可作為粒子治療中治療計劃和質量保證的有價值工具。;The first part of the thesis explores the frontier of Positron Emission Tomography (PET) imaging through the development of a novel rotating dual-head PET system, aimed at overcoming the inherent spatial resolution limitations within compact PET detectors. Central to this method is the introduction of a three-angle reconstruction technique utilizing the basic Maximum Likelihood Estimation Method (MLEM). We use GATE/Geant4 10.4 simulation toolkit to perform detector simulation and have developed our own image reconstruction framework.Image resolution and detector sensitivity are estimated. The resolution along limited-angle axis is improved from 3.6 mm to 1.7 mm using three-angle reconstruction approach.
The second part of the thesis delves into enhancing dose distribution accuracy in particle therapy through deep learning. Focusing on the transition from detector data to intrinsic dose distributions, we utilize Monte Carlo simulations with GATE/Geant4 on a human CT phantom exposed to high-energy protons. Employing a conditional generative adversarial network, we develop a neural network model to infer dose maps from PET coincidence distributions. Our model, evaluated by mean relative error and deviations in Bragg peak position, demonstrates deviations within 1% for dose and 2% for range in mono-energetic irradiations, with performance sustained under realistic spread-out Bragg peak conditions. This work underscores the feasibility of deep learning for mapping low count data to dose distributions, promising advancements in particle therapy imaging.
The third part of project examines the precision of proton therapy dose deposition, focusing on range verification using the Academia Sinica PET′s modular design with SiPMs and STiC asic readout. This study presents the positron emitter depth distribution in PMMA irradiated by 130 MeV protons, comparing measurements from a compact 32-channel setup at Chang Gung Memorial Hospital with Geant4 simulations. The results, validated by timing capabilities and multi-exponential fit analysis of different isotopes, confirms the implementation on anatomical change estimation during successive treatment sessions. Additionally, we introduce the AS-PET′s modular design and its simulated imaging performance, highlighting its potential in range verification for proton therapy. This thesis research can help to improve the PET image quality for flat-panel PET systems and can be a valuable tool for treatment planning and quality assurance in particle therapy.
