Abstract: It is of great significance for practical medical diagnosis to establish interpretable and evolvable models for intelligent diagnosis and treatment of diseases by using the massive medical data and medical knowledge accumulated in the process of clinical diagnosis and treatment. This paper studies intelligent diagnosis and treatment method based on the combination of knowledge guidance and data drive. By introducing medical knowledge and human- computer intelligent interaction means, we combined medical knowledge, data prior and doctor experience. Taking thyroid nodular diseases as the research object, this paper built a human- machine collaborative medical image intelligent diagnosis system by combining technologies such as intelligent diagnosis of medical images, automatic generation of medical reports, and the "human in the loop" collaborative mechanism. The accuracy of medical diagnosis and report generation was demonstrated through quantitative experiments. At the same time, it demonstrated the clinical usability of the system through its functional effects, effectively carry out intelligent medical imaging assisted diagnosis of "human in the loop", and improved diagnostic efficiency and level.