An Integrated AI Medical Diagnosis Application Based on Deep Learning
Haoyang Li, Xiaoyang Li, Bolin Shen, Zhi Zheng · 2021 2nd International Seminar on Artificial Intelligence, Networking and Information Technology (AINIT) · 2021
AI medical diagnosis has become an easy and effective approach method by which patients could get their health reports faster. This paper introduced an integrated AI medical diagnosis application with an image recognition system, chat system, and user interface. With the powerful help of the Convolutional Neural Network (CNN), an accurate image diagnosis system for breast cancer is ready to use. The chat system provides a bond between AI and patients through communication, using the Q&A sub-system based on Knowledge Graph (KG) and the casual chat sub-system based on Natural Language Process (NLP). Health condition query and common converse are satisfied respectively by Q&A sub-system and casual chat sub-system. Depending on the user’s intentional probability of medical problems, the casual chat system will interfere if the user’s intention is not matched in Knowledge Graph. Besides the deep learning part to process patient’s data, an easy-to-use user interface is also designed to guide people getting intended information, using Flask framework to offer API and jQuery to send a request. Our accuracy achieves 0.88 in breast cancer detection, and the result of GradCAM indicates the CNN network has a high performance on such a task. The practical experiment shows that the patient’s intention can proceed well in the chat system and that application can provide people advantageous feedback according to their health condition.