Clinical-Anchored Incremental Survival Learning: A Stability-Constrained Framework for Molecularly Enhanced Survival Prediction in Lung Adenocarcinoma

CAISL improves internal survival-risk prediction in lung adenocarcinoma (LUAD) by raising mean C-index to 0.7246 versus 0.6970 for clinical Cox.

• Why it matters: High-dimensional molecular inputs in LUAD can overfit and destabilize feature selection, limiting reliable prognostic discrimination beyond clinical models.
• What they did: The CAISL framework combines a sparse molecular risk correction with a clinical baseline, evaluated in TCGA LUAD (470 patients, 119 deaths) using five seeded, event-stratified five-fold cross-validation.
• The result: CAISL also shows higher time-dependent AUC at three and five years and a lower integrated Brier score (0.164 vs 0.181), with stability metrics indicating consistent feature selection.

AI summary of the abstract · may contain errors · not medical advice · read the paper.