Fig. 1: Features selection, SHAP analysis and models assessment.

A Features selection with LASSO regression in sub-region1 radiomics model (left panel), multichannel 2D DL model (middle panel), and 3D DL model (right panel). B, C Feature importance based on XGBoost algorithm with SHAP analysis in sub-region1 model (left panel), multichannel 2D DL model (middle panel), and 3D DL model (right panel). D The ROC curves of DLRad1, DLRad2, 2D DL, 3D DL, and sub-region1 radiomics model in training set (left panel), internal validation set (middle panel), and external validation set (right panel). Note: 2D two-dimensional, 3D three-dimensional, DL deep learning, LASSO least absolute shrinkage and selection operator, ROC receiver operating characteristic, SHAP Shapley Additive Explanations, XGBoost eXtreme Gradient Boosting.