无线电技术
体素
医学
肿瘤分级
质心
Boosting(机器学习)
人工智能
免疫组织化学
肿瘤科
正电子发射断层摄影术
成像生物标志物
核医学
内科学
模式识别(心理学)
放射科
肿瘤异质性
错误发现率
变异系数
表达式(计算机科学)
梯度升压
病理
蛋白质表达
作者
Ruxi Chang,Ruiyan WANG,Xiang LIU,Jungang Gao,Weishan ZHANG,Xiaoyi DUAN
出处
期刊:Quarterly Journal of Nuclear Medicine and Molecular Imaging
[Edizioni Minerva Medica]
日期:2026-11-01
标识
DOI:10.23736/s1824-4785.26.03675-7
摘要
BACKGROUND: This study aims to explore the role of radiomic features and PET image heterogeneity in predicting PD-L1 expression in NSCLC. METHODS: F-FDG PET/CT and PD-L1 immunohistochemical assessment, PD-L1 positivity was defined as a tumor proportion score (TPS) ≥1%. Metabolic heterogeneity was assessed using the coefficient of variation (COV, SUVsd / SUVmean × 100%), the distance from the hottest voxel to the tumor centroid (NHOC), and the distance from the hottest voxel to the nearest tumor boundary (NHOP). Tumors were semi-automatically segmented using a 40% SUVmax threshold. To capture the most comprehensive peritumoral information, the ROI is expanded by 15 mm. Radiomic features were extracted from intratumoral and 15-mm peritumoral regions on PET and CT images and a model is built with tumor and surrounding features using Gradient Boosting Decision Tree. The model effectiveness is assessed using AUC, sensitivity, specificity, accuracy, and F1 score. RESULTS: The study included 96 patients (median age: 64 [59, 70] years), of whom 58 were male. 63 patients were PD-L1 expression positive, while 33 were negative. COV and NHOP was significantly higher in the PD-L1 positive group (Z=-2.688, P=0.007; Z=-3.704, P<0.001), while no significant differences noted in other parameters between the groups (P>0.05 for all). Both COV and NHOP are independent predictors of PD-L1 positivity(P<0.05). The combined radiomics model outperforms the individual radiomics models, with the integrated intratumoral CT, PET, and peritumoral CT radiomics model achieving the highest predictive performance (AUC=0.816). The addition of metabolic heterogeneity parameters (COV and NHOP) further improved discrimination, resulting in all evaluation metrics exceeding 0.8. CONCLUSIONS: F-FDG PET/CT metabolic parameters, especially COV and NHOP, predict PD-L1 expression and improve targeted treatment decision-making.
科研通智能强力驱动
Strongly Powered by AbleSci AI