IntroductionThe Food Drug Administration authorized the first autonomous artificial intelligence (AI) diagnostic system in 2018, which is for detecting diabetic retinopathy1. Since then, AI machine learning (ML) based predictive technologies are rapidly made available for incorporation into clinical workflows2, e.g., for early sepsis detection3 and predicting surgery time4. However, recent studies revealed problems of prediction models under various medical scenarios, e.g., missed detection in m...