Deep-learning computer-assisted decision-making (DL-CAD) medical device software has high technical content, high clinical value and high social attention. Center for Medical Device Evaluation (CMDE) has introduced several policies to encourage technological innovation and promote industry development. As an innovative clinical application technology, neither manufacturer nor reviewer has enough experience with registration and evaluation of DL-CAD devices. In consequence, many misunderstanding exist in its clinical evaluation requirement. This article aims to provide reference for investigators, manufacturers and reviewers, as well as summarize elements of clinical trial design, and bias control methods, by investigating and surveying global supervision requirements & experience, academic research achievements.