CT-Based Topological Analysis Predicts Pancreatic Cancer Survival and Chemotherapy Benefit

CT-based topological data analysis classifier (ra-CAPR) independently predicted overall survival (HR 1.42, P=0.005) and adjuvant chemotherapy benefit in pancreatic ductal adenocarcinoma, outperforming conventional radiomics (AUC 0.77 vs 0.72).

• Retrospective study of 1452 patients across transcriptomic and imaging cohorts; TDA classifier developed in 122 institutional patients, externally validated in 50 (TCIA), and tested in 657 surgical patients.
• ra-CAPR achieved cross-cohort AUCs of 0.774 and 0.744 versus 0.715 and 0.621 for conventional radiomics, with markedly lower feature distributional shift (26% vs 75%).
• Limitations: retrospective design, single-institutional development, small external validation set, no prospective evaluation.

RadPigeon summaries are original and for information only. They are not clinical advice.