Deep Skin Diseases Diagnostic System With Dual-Channel Image and Extracted Text

The document discusses the development and testing of an Artificial Intelligence (AI) model, DIET-AI, designed to diagnose 31 common skin diseases. This is in response to the high number of skin disease patients (~200 million per year) in China and the low doctor-patient ratio. The AI model, which uses Computer Vision (CV), Natural Language Processing (NLP), and Speech Recognition (SR), was trained on a dataset of over 200,000 images and 220,000 medical records from Asian patients. The model was tested on 6043 cases collected from 15 hospitals in 7 provinces in China between September 1 and December 1, 2021. The performance of DIET-AI was compared with that of 6 doctors of varying seniority, and it was found that the AI's average performance was at least as good as that of the doctors. The model's effectiveness was confirmed by comparing the area under curve (AUC), sensitivity, and specificity. The study concludes that the DIET-AI model is effective in clinical scenarios and can be further validated in more complex clinical scenarios.