Causally Constrained and Uncertainty-Aware Cardiopulmonary Risk Prediction
Researchers introduce the Causal Envelope Agent (CE-Agent), a two-stage framework that compiles epidemiological evidence into constraints on risk-factor weights and estimates local baseline risk plus newly added CT phenotypes from cohort data. It uses closed-form uncertainty propagation to handle missing clinical or CT inputs in milliseconds and return calibrated risk intervals. Evaluated across 53,452 lung-screening participants and 6,814 multi-ethnic adults, CE-Agent outperformed cardiovascular calculators in lung screening, corrected Pooled Cohort Equations risk overestimation, generalized across hospitals and scanners, and matched lung cancer screening models.
The method offers a practical, uncertainty-aware approach to joint cardiopulmonary risk prediction that could improve screening triage and reduce overfitting in clinical AI.
Comments & Questions (0)
No comments yet
Be the first to comment!