Related Experiment Video
Updated: Aug 5, 2026
Segmentation and Linear Measurement for Body Composition Analysis using Slice-O-Matic and Horos
Published on: March 21, 2021
Computed tomography body composition-based prediction of postoperative quality of life in non-small cell lung cancer
Chuanlin Shan1, Xiaomin Sun1, Degui Wang2
1Department of Thoracic Surgery, Third Ward, Anhui Provincial Chest Hospital, Hefei, China.
Objectives:
This paper aims to design an explainable machine learning model capable of predicting postoperative quality of life in elderly NSCLC patients by integrating CT derived body composition metrics with standard clinical indicators.
Methods:
A total of 200 elderly NSCLC patients undergoing lobectomy were retrospectively analyzed and divided into High QoL (n=124) and Low QoL (n=76) groups based on 6-month FACT-L scores. The cohort was split into training (n=140) and validation (n=60) sets. CT derived features (PMA, SMI, muscle attenuation, LAMA), clinical variables, and functional indicators (including 6 min walk distance [6MWD] decline) were collected. Multivariate regression identified independent predictors. Machine learning models (Logistic regression, Random Forest, and XGBoost) were developed and evaluated using ROC curves. SHAP analysis assessed model interpretability.
Results:
Low QoL patients showed lower PMA, SMI, and muscle attenuation, but higher LAMA and greater 6MWD decline (all p<0.001). Multivariate analysis identified 6MWD decline, PMA, and muscle attenuation as independent predictors (all p<0.001). XGBoost achieved the best performance (AUC: 0.892 training, 0.854 validation). SHAP analysis highlighted 6MWD decline as the most influential factor.
Conclusions:
Multimodal integration of CT body composition and functional indicators accurately predicts postoperative QoL, supporting personalized perioperative care.
