Radiomics Models of Dynamic Contrast-Enhanced MRI for Evaluation of the Expression Levels of CD3+, CD4+ & CD8+ Tumor-Infiltrating Lymphocytes in Advanced Gastric Carcinoma

Abstract

Abstract Objective To explore the effectiveness of machine learning classifiers based on dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) in predicting the expression levels of CD3+, CD4+ and CD8+ tumor-infiltrating lymphocytes (TILs) in patients with advanced gastric cancer (AGC). Materials and Methods This study investigated 103 patients with confirmed AGC through DCE-MRI and immunohistochemical staining. Utilizing Omni Kinetics software, radiomics features (Ktrans, Kep, and Ve) were extracted and underwent selection via variance threshold, SelectKBest, and LASSO methods. A logistic regression model was constructed, and 10-fold cross-validation assessed its performance. Immunohistochemical staining was used to evaluate CD3+, CD4+, and CD8+ T-cell expression. A receiver operating characteristic curve was used to calculate the model performance. Results Eight radiomics characteristics were used in the CD3 model to produce AUCs of 0.857 (training) and 0.863 (test). The CD4 model used seven characteristics to produce AUCs of 0.867 (training) and 0.817 (test). The CD8 model used six characteristics to attain AUCs of 0.876 (training) and 0.820 (test). Clinical usefulness was verified by a careful decision curve study. Conclusions Machine learning classifiers based on DCE-MRI have the potential to accurately predict CD3+, CD4+, and CD8+ tumor-infiltrating lymphocyte expression levels in patients with AGC.