Back to all papers

CT-derived topological intratumoral heterogeneity predicts major pathological response to neoadjuvant immunochemotherapy in resectable non-small-cell lung cancer: a two-center study.

July 27, 2026pubmed logopapers

Authors

Zhou Y,Hong T,Qi H,Zhao J,Wu Z,Du J,Liu L,Zou F

Affiliations (4)

  • Department of Radiology, Jiangxi Cancer Hospital & Institute, Jiangxi Clinical Research Center for Cancer, The Second Affiliated Hospital of Nanchang Medical College, Nanchang, China.
  • Liaoning University of Traditional Chinese Medicine, Shenyang, China.
  • Department of Clinical Engineering, Southern Medical University, Nanfang Hospital, Guangzhou, China.
  • Department of Radiology, The Second Affiliated Hospital, Jiangxi Medical College, Nanchang University, Nanchang, China.

Abstract

To develop and validate a pretreatment computed tomography (CT)-derived model for predicting major pathological response (MPR) to neoadjuvant immunochemotherapy in resectable non-small-cell lung cancer (NSCLC). This multicenter retrospective study included 287 patients with stage IB-III NSCLC treated with neoadjuvant immunotherapy plus chemotherapy followed by surgery. Patients from Center A formed the training cohort (n = 224), and patients from Center B formed the external validation cohort (n = 63). Two-dimensional (2D) and three-dimensional (3D) intratumoral heterogeneity scores (ITHscores) were derived from venous-phase pretreatment CT using local radiomics, K-means clustering and connected-component topology. Candidate predictors included clinicopathological variables, ITHscores and blood-derived inflammatory indicators. Eight machine-learning algorithms were compared, and the final model was selected using cross-validation and feature ablation. Shapley additive explanations (SHAP) analysis was used to interpret feature contributions. Non-MPR tumors showed significantly higher 2D and 3D ITHscores than MPR tumors in both cohorts (all P < 0.001). CatBoost was selected as the optimal algorithm. SHAP analysis identified 2D ITHscore and 3D ITHscore as the two most influential predictors, with mean absolute SHAP values of 0.403 and 0.326, respectively. The final parsimonious model retained 2D ITHscore, 3D ITHscore and systemic immune-inflammation index. It achieved AUCs of 0.824 in the training cohort and 0.792 in the external validation cohort, outperforming the clinical and ITH-only models. Calibration and decision curve analyses showed acceptable agreement and potential clinical utility. CT-derived topological intratumoral heterogeneity, combined with systemic inflammation, may provide a non-invasive pretreatment biomarker for predicting MPR in resectable NSCLC.

Topics

Carcinoma, Non-Small-Cell LungLung NeoplasmsTomography, X-Ray ComputedImmunotherapyJournal ArticleMulticenter Study

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.