Dual-Regulator Conformal Prediction for Medical Digital Twin Uncertainty

SUMMARYA digital twin is a virtual model that interacts bidirectionally with physical systems through closed-loop optimization. It has wide applications in aerospace, construction, and healthcare. Verification, validation, and uncertainty quantification of predictive models help ensure robust decisions and reliable outputs. Based on the previously developed conformal prediction method, mapped-dual adaptive conformal prediction (M-DACP) is proposed as extension, targeting application scenarios where multidimensional outputs jointly participate in decision-making. It enhances the ability to handle complex conditions such as non-exchangeable data and distribution shifts, especially when input information is limited. Using Apnea-ECG database, the reliability numerical analysis results and the diversity of result format, referable information provided by optimal transport (OT) are demonstrated, offering directions for future research.