Schizophrenia is a serious mental illness, and its clinical manifestations are complex and varied. At present, the clinical diagnosis of schizophrenia mainly relies on history inquiry, interview, and scale evaluation, etc., which is prone to missed diagnosis and misdiagnosis. Therefore, many researchers have attempted to use deep learning method for the adjuvant diagnosis of schizophrenia. In this paper, the data types and deep learning model used in the automatic diagnosis of schizophrenia are summarized, and presence of challenges and future development trends in this field are analyzed and prospected. 

无