影像科学与光化学2024，Vol.42Issue(2)：148-154,7.DOI:10.7517/issn.1674-0475.231107
基于对比增强CT的影像组学模型可预测乳腺癌患者的肿瘤浸润淋巴细胞水平
Enhanced CT-Based Radiomcis Model can Predict Tumour-infiltrating Lymphocyte Levels in Breast Cancer Patients
摘要
Abstract
Objective:To develop a novel radiomics model based on contrast-enhanced CT for predicting tumor-infiltrating lymphocytes(TIL)levels in breast cancer(BC)patients.Methods:210 patients with non-specific invasive BC were retrospectively enrolled(training group:147,validation group:63),and high-dimensional radiomcis features in the contrast-enhanced CT images of the patients were extracted using the Pyradiomics software package,followed by the use of the Mann-Whitney U-test,Spearman's correlation coefficient,and the least absolute shrinkage and selection operator(LASSO)algorithm for stepwise feature screening.The filtered optimal features were developed in the training set and validated in the validation set by non-linear support vector machine(NLSVM)method and the discriminative ability of the model was verified in the validation set.Afterwards,the global interpretation of the model with feature importance ranking is performed using the Shapley additive and exPlanatory(SHAP)algorithm.Results:The area under curve(AUC)of training and validation groups were 0.824(95％CI:0.762-0.886)and 0.766(95％CI:0.624-0.909),respectively,and there were three positively influencing features features four negatively influencing features for the radiomcis model,of which log-sigma-5-0-mm-3D_firstorder_Maximum has the greatest influence on the model,and the larger its value,the smaller the model output SHAP value.Conclusion:Contrast-enhanced CT-based radiomcis model can help clinicians accurately predict the level of tumor infiltrating lymphocyte in BC patients before treatment and facilitate personalised treatment for BC patients.关键词
影像组学/肿瘤浸润淋巴细胞/乳腺癌/预测模型Key words
radiomcis/TIL/breast cancer/predictive model引用本文复制引用
李永,崔书君,杨飞,张凡,殷晓霞..基于对比增强CT的影像组学模型可预测乳腺癌患者的肿瘤浸润淋巴细胞水平[J].影像科学与光化学,2024,42(2):148-154,7.基金项目
河北省医学重点课题(20210091) （20210091）
