摘要：Medical artificial intelligence (AI) is a hot point in AI utilities now, of which clinical diagnosis is one important AI administration. Diagnostic AI (D-AI) is the simulation of doctor’s clinical thinking for clinical analysis and decision. These abilities come from many-years medical education and clinical experience. Therefore, D-AI should also be based on medical knowledge from medical university and clinical experience. However, at present most D-AI is based on big data analysis, it is now right. The first big question of big data based D-AI is that it is not evidence-based analysis which is absolutely required by evidence-based medicine in clinical analysis and decision. The second is that its algorithm can not be interpreted by medical knowledge, thus could induce algorithm black box, which is now commonly suggested to avoid. A medical knowledge based D-AI algorithm is needed to set up right D-AI, simulating clinician’s clinical thinking.
