|
内容記述 |
Objective.The forceps-type positron emission counter (PEC) was developed for intraoperative quantification of 18F-fluorodeoxyglucose uptake in individual lymph nodes.It employs coincidence detection between two opposing 2 x 2 bismuth germanate crystal arrays, yielding only 16 lines of response (LORs).This study investigates the feasibility of image reconstruction from these 16 LORs to enable repositioning guidance and sensitivity-corrected quantification.Approach.Seven reconstruction algorithms---spanning different data-fidelity models (least-squares and Poisson likelihood), regularization priors (none, L_2 Tikhonov, and total variation), and optimization frameworks---were compared using Geant4-based Monte Carlo simulations.Phantom experiments assessed reconstruction quality and count-rate dependence, localization accuracy across the field of view, hot-background robustness, and dual-source separation.Main results.Among the tested design axes, regularization has the largest impact on reconstruction quality, and the optimal algorithm depends on the task: MAP-EM-TV (Poisson-likelihood with TV regularization) achieves the highest structural similarity while PLS-TV (least-squares with TV) achieves the best quantitative accuracy (FoV-sum CV reduced by 21.5 percentage points) and a mean localization error of 1.27~mm.Near-optimal image quality is reached at $\sim$3{,}000 counts, with spatial information retained down to ~100 counts; image quality degrades substantially below source-to-background ratio 12:1.Significance.The results demonstrate that the 16-LOR PEC can provide spatial information beyond count-rate or center-of-gravity estimation, enabling repositioning guidance and sensitivity-corrected quantification for intraoperative use.As the study relies on a shared forward model, validation with a physical prototype is the key next step. |
