質子放射治療在精準醫療中具備極大潛力，但射程不確定性是其臨床 應用的主要挑戰。雙面板在線式正子造影系統(Dual-Panel On-line PET System)能透過原位即時監測質子誘發之短半衰期同位素(如 15O)，提供可靠 的射程驗證。然而，受限於系統非連續型偵測器的幾何排列，其投影數據嚴 重欠缺，導致三維影像重建易產生嚴重的有限角度假影與離焦模糊，降低了 定性與定量的準確性。 為改善此幾何限制造成的影像劣化，本研究利用 GATE 蒙地卡羅模擬 軟體，建構與臨床相符的系統模型。研究中首先進行取樣完整性與系統靈敏 度的量化分析， 結果顯示，取樣完整性與靈敏度會隨偏離視野中心及加大探 頭間距而急遽下降。 接著，透過三維空間的格點掃描實驗，獲取不同空間位置的點射源投影 數據，藉此建構具備空間變異性的影像系統矩陣。 最後，本研究結合列表模式期望值最大化演算法(List-Mode Maximum Likelihood Expectation Maximization， LM-MLEM)與移變理查森-露西反卷積 演算法(Shift-Variant Richardson-Lucy Deconvolution Algorithm， SV-RL)進行 影像重建與比較，以評估解析度恢復的效果;Proton radiotherapy holds great potential in precision medicine， but range uncertainty remains a major challenge in its clinical application. The Dual-Panel On-line PET System can provide reliable range verification through in-situ， realtime monitoring of proton-induced short-half-life isotopes (such as 15O). However， restricted by the geometric arrangement of the system′s non-continuous detectors， its projection data is severely deficient. This deficiency makes three-dimensional image reconstruction highly susceptible to severe finite-angle artifacts and defocus blur， thereby degrading qualitative and quantitative accuracy. To address the image degradation caused by these geometric limitations， this study utilized the GATE Monte Carlo simulation software to construct a system model consistent with clinical settings. Initially， a quantitative analysis of sampling completeness and system sensitivity was performed. The results indicated that both sampling completeness and sensitivity decline sharply as the position deviates from the center of the field of view (FOV) and as the distance between the detector panels increases. Subsequently， a three-dimensional grid scan experiment was conducted to acquire projection data from point sources at various spatial locations， thereby constructing an imaging system matrix with spatial variation. Finally， this study combined the List-Mode Maximum Likelihood Expectation Maximization algorithm and the Shift-Variant Richardson-Lucy Deconvolution Algorithm to perform image reconstruction and comparison， aiming to evaluate the effectiveness of resolution recovery.
