摘要目的:探究有序子集最大期望值迭代（OSEM）法和贝叶斯惩罚似然（BPL）法重建 18F-氟代脱氧葡萄糖（ 18F-FDG）PET-MRI的不对称指数（AI）对颞叶癫痫灶的诊断价值。 方法:横断面研究。选取2020年1月至2023年6月于华中科技大学同济医学院附属协和医院PET中心行 18F-FDG一体化PET-MRI脑成像的颞叶癫痫患者。将每例患者的PET源数据分别采用OSEM法、BPL法（β=150）及BPL法（β=350）3种方式重建。对小脑皮质、海马、颞叶海马旁回、颞上回、颞中回、颞下回、颞叶内嗅皮层、颞横回、颞极共9对双侧脑区进行定量分析，提取每个脑区标准化摄取值均值（SUV mean）并计算AI。以小脑皮质为参考区，将各脑区按照AI值分为癫痫致痫灶区（AI≥20%）和癫痫泛化区（5%<AI<20%）。比较不同重建方法的SUV mean和AI差异。 结果:共纳入颞叶癫痫患者18例，年龄（29.5±14.7）岁，男6例、女12例。BPL法（β=150）和BPL法（β=350）重建的双侧小脑皮质区SUV mean均高于OSEM法（左侧：6.88±1.86比6.80±1.83比6.57±1.76，右侧：6.84±1.81比6.76±1.79比6.49±1.69；均 P<0.001）。在小脑皮质区和癫痫泛化区，3种重建法的AI差异均无统计学意义（均 P>0.05）；在癫痫致痫灶区，BPL法（β=350）的AI低于OSEM法（24.68±4.13比25.42±3.64， P=0.048），BPL法（β=150）与OSEM法的AI差异无统计学意义（ P>0.05）。 结论:BPL法重建 18F-FDG PET-MRI的AI对颞叶癫痫灶诊断与OSEM法无差异。

abstractsObjective:To evaluate the diagnostic value for temporal lobe epilepsy focus localization based on the asymmetry index (AI) derived from images reconstructed using the ordered subsets expectation maximization (OSEM) and the Bayesian penalized likelihood (BPL) methods on 18F-fluorodeoxyglucose ( 18F-FDG) PET-MRI. Methods:This is a cross-sectional study. We collected and analyzed 18F-FDG PET-MRI brain images from 18 patients diagnosed with temporal lobe epilepsy at the PET Center of Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, between January 2020 and June 2023. The acquired PET list-mode data for each patient were reconstructed using three methods respectively: OSEM, BPL (β=150), and BPL (β=350). Quantitative analysis was performed on 9 pairs of bilateral brain regions: cerebellar cortex, hippocampus, parahippocampal gyrus, superior temporal gyrus, middle temporal gyrus, inferior temporal gyrus, entorhinal cortex, transverse temporal gyrus, and temporal pole. The mean standardized uptake value (SUV mean) for each region was extracted, and the AI for each region was calculated. The cerebellar cortex was considered the reference region, and other brain regions were classified into the epileptogenic zone (AI≥20%) and epileptic propagation zone (5%<AI<20%). Differences in SUV mean and AI among the three reconstruction methods were analyzed using one-way ANOVA or the Friedman test. Results:A total of 18 patients with temporal lobe epilepsy were included [6 males, 12 females, mean age (29.5±14.7) years]. In both the right and left cerebellar cortex, the SUV mean values in the BPL (β=150) and BPL (β=350) were higher than those in the OSEM (left: 6.88±1.86 vs 6.80±1.83 vs 6.57±1.76; right: 6.84±1.81 vs 6.76±1.79 vs 6.49±1.69; all P<0.001). There were no significant differences in AI among the three reconstruction methods for the cerebellar cortex and the epileptic propagation zone (all P>0.05). In the epileptogenic zone, the AI in the BPL (β=350) was lower than that in the OSEM (24.68±4.13 vs 25.42±3.64, P=0.048), while there was no significant difference in AI between the BPL (β=150) and OSEM ( P>0.05). Conclusion:Compared with the OSEM method, AI in images reconstructed with the BPL method demonstrates no significant difference for localizing epileptogenic foci in temporal lobe epilepsy on 18F-FDG PET-MRI.