AI-driven discovery of a simple and generalizable computer tomography imaging biomarker for improved clinical T staging in early-stage lung adenocarcinoma
The weights of the lung nodule automatic segmentation model based on nnUNet: https://zenodo.org/records/16608410
quantitative features calculation.py/: functions of quantitative features calculation based on different thresholds (tggo and ts).
Determination of the optimal component thresholds combination and the optimal quantitative feature to obtain the modified clinical T stage (cTm):
0_import_lib_data.R/: load required libs and define the thresholds ranges of extracting GGO and solid components.
1_DFS_prediction-performance.R/: LASSO regression for DFS prediction based on 5 fold cross-validation.
1_OS_prediction-performance.R/: LASSO regression for OS prediction based on 5 fold cross-validation.
1_Pathology_prediction-performance.R/: LASSO regression for pathology prediction based on 5 fold cross-validation.
2_Optimal_threshold_determination.R/: Determination of optimal thresholds based on the highest mean C-index of pathology, DFS, and OS prediction.
3_Optimal_feature_determination.R: Determination of optimal quantitative feature based on the highest mean C-index of pathology, DFS, and OS prediction.
4_cTm.R/: The modification of current clinical T stage according to the value of SM%.
cTm_application.py/: The code for building an executable program for cTm demonstration.