Chest Computed Tomography Habitat and Body Composition for Survival Prediction in Stage IV Lung Cancer

Objectives To integrate clinical characteristics, CT habitat radiomic features, and quantitative body composition parameters in patients with stage IV non-small-cell lung cancer (NSCLC) and develop a multidimensional model for predicting overall survival. Materials and methods Overall, 196 patients with stage IV NSCLC were retrospectively enrolled and divided into a training set (n = 137) and an internal test set (n = 59). An additional 50 patients formed an external test set. Radiomic features of the entire tumor, tumor habitat features based on K-means clustering, and body composition at the L3 vertebral level were extracted using 3D Slicer. Random forest and LASSO(LASSO) regression were utilized for feature selection, followed by the construction of unimodal, bimodal, and multi-dimensional integrated prognostic models. Model performance was evaluated via Harrell’s concordance index, time-dependent receiver operating characteristic (ROC) curves, and Kaplan–Meier survival analysis. The SHapley Additive exPlanations(SHAP) method was used for model interpretability analysis. Results The multidimensional fusion model achieved the best performance, with area under the curve values of 0.842, 0.811, and 0.817 in the training, internal, and external test sets, respectively. It significantly outperformed single-modality models and effectively stratified patients into high- and low-risk groups (log-rank p = 0.00067). C-indices were 0.850, 0.790, and 0.760, respectively. SHAP analysis identified body composition features as key contributors to survival prediction. Conclusion The proposed model demonstrated robust prognostic performance and may support individualized management of advanced NSCLC.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-18
DOI
https://doi.org/10.21203/rs.3.rs-11442726/v1
Primary Topic
Radiomics and Machine Learning in Medical Imaging
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Chest Computed Tomography Habitat and Body Composition for Survival Prediction in Stage IV Lung Cancer

Bosen Xie, Lei Yue, Duiming Yang, Hao Hu et al.
6 links
Research Square
Radiomics and Machine Learning in Medical Imaging
preprint

Chest Computed Tomography Habitat and Body Composition for Survival Prediction in Stage IV Lung Cancer

Bosen Xie, Lei Yue, Duiming Yang, Hao Hu, Jiageng Li, Yubo Wang, Yadan Yin, Mengxue Kong, Chunyuan Wei, Yang Fu, Bin Yang, peiqi wang
preprint en

Abstract

Objectives To integrate clinical characteristics, CT habitat radiomic features, and quantitative body composition parameters in patients with stage IV non-small-cell lung cancer (NSCLC) and develop a multidimensional model for predicting overall survival. Materials and methods Overall, 196 patients with stage IV NSCLC were retrospectively enrolled and divided into a training set (n = 137) and an internal test set (n = 59). An additional 50 patients formed an external test set. Radiomic features of the entire tumor, tumor habitat features based on K-means clustering, and body composition at the L3 vertebral level were extracted using 3D Slicer. Random forest and LASSO(LASSO) regression were utilized for feature selection, followed by the construction of unimodal, bimodal, and multi-dimensional integrated prognostic models. Model performance was evaluated via Harrell’s concordance index, time-dependent receiver operating characteristic (ROC) curves, and Kaplan–Meier survival analysis. The SHapley Additive exPlanations(SHAP) method was used for model interpretability analysis. Results The multidimensional fusion model achieved the best performance, with area under the curve values of 0.842, 0.811, and 0.817 in the training, internal, and external test sets, respectively. It significantly outperformed single-modality models and effectively stratified patients into high- and low-risk groups (log-rank p = 0.00067). C-indices were 0.850, 0.790, and 0.760, respectively. SHAP analysis identified body composition features as key contributors to survival prediction. Conclusion The proposed model demonstrated robust prognostic performance and may support individualized management of advanced NSCLC.

Research Square
Kunming Medical University (CN), Calmette Hospital (KH), Panzhihua Central Hospital (CN), Baoshan University (CN)
National Natural Science Foundation of China
Good health and well-being
Radiomics and Machine Learning in Medical Imaging
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon