CASCADE Conformal Prediction: Uncertainty-Adaptive Prediction Intervals for Two-Stage Clinical Decision Support
CASCADE introduces a novel conformal prediction framework for Parkinson's Disease management that adapts prediction intervals based on uncertainty from an initial screening classifier. By using epistemic uncertainty from Stage 1 to scale Stage 2 predictions, it creates dynamic safety margins. This method balances accurate risk assessment with efficient dose forecasting by ensuring intervals are tight for confident patients and wide for ambiguous ones.