中国临床医学影像杂志2025，Vol.36Issue(3)：194-199,6.DOI:10.12117/jccmi.2025.03.009
基于MRI影像组学评估局部进展期直肠癌新辅助放化疗反应:一项两中心多设备研究
MRI radiomics for assessing treatment response of neoadjuvant chemoradiotherapy in locally advanced rectal cancer:a two-center,multi-vendor study
摘要
Abstract
Objective:To explore the value of MRI radiomics in evaluating pathological complete response(pCR)after neoadjuvant chemoradiotherapy(nCRT)in patients with locally advanced rectal cancer(LARC).Methods:Between October 2021 and January 2024,a total of 203 patients with LARC were retrospectively gathered from two medical centers:Nanfang Hospital of Southern Medical University(Center 1,n=142)and Guangdong Provincial Hospital of Traditional Chinese Medicine(Center 2,n=61).High-resolution MRI examinations following nCRT were conducted using 3.0T MR equipment from four different models across the two centers.Subsequently,two radiologists delineated the volume of interest(VOI)encompassing the entire tumor lesions on T2WI and diffusion weighted imaging(DWI)images obtained after nCRT.Subsequently,radiomic features were extracted from these delineated VOIs for further analysis.Patients from Center 1 were stratified into a training set(n=113)and an internal validation set(n=29)at a ratio of 4∶1,while the patients from Center 2 were utilized as an independent external test set.Pearson correlation coefficients was used for dimensionality reduction.Radiomics models were constructed with five classifiers,including support vector machine,linear discriminant analysis,random forest,Logistic regression,and gaussian process,and the classifier with the best model performance was selected.Evaluating the model's assessment performance was based on the area under the curve(AUC)of the receiver operating characteristic(ROC)curve.Results:A total of 1 888 radiomics features were extracted from T2WI and DWI images after nCRT.After dimensionality reduction and feature selection,twelve most valuable radiomic features were obtained,comprising 5 features from the T2WI sequence and 7 features from the DWI sequence.These features encompass 1 morphological feature,1 first-order feature,1 laplacian of gaussian feature and 9 wavelet features.The radiomics model was constructed using gaussian process.The model achieved AUC values of 0.999,0.812,and 0.803 for the training set,internal validation set,and external test set,respectively.Conclusion:A predictive MRI-based radiomic model demonstrates promising accuracy in assessing treatment response to nCRT in patients with LARC.This model has been validated using an independent external test set,suggesting its potential utility for guiding clinical decision-making.关键词
直肠肿瘤/磁共振成像Key words
Rectal Neoplasms/Magnetic Resonance Imaging分类
医药卫生引用本文复制引用
贾子琪,林耿斌,袁文静,张汉良,崔亮,林韵颖,刘岘,吴元魁,陈维翠..基于MRI影像组学评估局部进展期直肠癌新辅助放化疗反应:一项两中心多设备研究[J].中国临床医学影像杂志,2025,36(3):194-199,6.基金项目
国家自然科学基金(82202259) （82202259）
广州市科技局市员联合资助项目(2023A03J0245) （2023A03J0245）
