分子影像学杂志2026，Vol.49Issue(5)：594-603,10.DOI:10.12122/j.issn.1674-4500.2026.05.05
瘤体联合瘤周影像组学模型可预测非小细胞肺癌免疫治疗疗效
Combined intratumoral and peritumoral radiomics model predicts immunotherapy efficacy in non-small cell lung cancer
摘要
Abstract
Objective To investigate the predictive value of radiomics models based on non-contrast and arterial phase CT images of intratumoral,peritumoral(5,10,15 mm),and combined intratumoral-peritumoral regions for immunotherapy efficacy in patients with non-small cell lung cancer(NSCLC).Methods This retrospective study enrolled 127 patients with NSCLC who underwent immunotherapy and pre-treatment contrast-enhanced CT from January 2020 to January 2024 at the First Affiliated Hospital of Bengbu Medical University.A multi-input(non-contrast,arterial phase,and their combination),multi-region of interest(intratumoral,peritumoral 5,10,15 mm),and multi-model strategy was used for image analysis.The primary analytical model was pre-specified as the model constructed from intratumoral and 15 mm peritumoral features extracted from combined non-contrast and arterial phase images.Tumor regions of interest(ROIs)were delineated on the largest cross-sectional area using uAI Research Portal software.Peritumoral ROIs of 5,10,15 mm were generated via platform expansion.After absolute maximum normalization,optimal radiomic features were selected using LASSO regression and maximum relevance minimum redundancy algorithms to construct the radiomics model.Following preprocessing,models were developed using identical algorithms and ROC curves were plotted.Four metrics,including area under the curve(AUC),sensitivity,specificity,and accuracy,were analyzed for both the training and test sets.Results A total of 21 radiomics models(intratumoral,peritumoral,combined)were constructed based on non-contrast,arterial phase,and combined images under different peritumoral expansion conditions.All models were built following feature selection from an initial pool of 2264 features.The combined model using non-contrast plus arterial phase images with a 15 mm intratumoral-peritumoral region performed best,achieving an AUC of 0.8871(95%CI:0.8380-0.9362)in the training set and 0.7095(95%CI:0.5903-0.8286)in the test set.Conclusion Preliminary findings suggest that radiomics features combining intratumoral and peritumoral(15 mm)information from both non-contrast and arterial phase CT images may hold value for predicting immunotherapy response in NSCLC,potentially aiding clinical decision-making.关键词
非小细胞肺癌/影像组学/免疫治疗/瘤周区域Key words
non-small cell lung cancer/radiomics/immunotherapy/peritumoral region引用本文复制引用
文欣园,胡尹迪,李轶涵,游昕楠,谢波,马宜传..瘤体联合瘤周影像组学模型可预测非小细胞肺癌免疫治疗疗效[J].分子影像学杂志,2026,49(5):594-603,10.基金项目
安徽省临床医学研究转化专项立项项目(202304295107020072) （202304295107020072）
