Life is a complex and cross-scale information system. Research and explorations based on information technology are expected to promote new medical paradigms and accelerate medical discovery. Meanwhile, bioscience could also inspire the innovation of information technology. Artificial intelligence has shown great potential in using medical images and multimodal data to diagnose common diseases. However, current diagnostic models face challenges in complex clinical scenarios. The main reason is that it is difficult for machines to understand the high-level semantic information of medical images and thus cannot provide an evidence-based explanation of diagnoses.
To resolve such critical AI-related biomedical issues, Peking University professor Guangyu Wang is committed to leveraging emerging information technologies and promoting new digital healthcare models. Her unique cross-field training and professional experience in medicine, computer science, and bioinformatics provide an excellent foundation for interdisciplinary research.
Guangyu proposed a “data-driven, evidence-fusion” theory for intelligent medical computing. During the pandemic, she developed an AI system for lung-lesion semantic map segmentation, automatic diagnosis, and dynamic evaluation, and the work was awarded as the Best Paper of Cell journal 2020. In addition, she developed an intelligent triage system for the diagnosis and discrimination of viral and non-viral pneumonia using chest X-rays based on semantic-based medical image recognition and reasoning.
To address the challenges of diverse time-series data and complex complications of follow-up cohort, Guangyu built an intelligent diagnosis and treatment method for major chronic diseases which could identify patients with major chronic diseases (such as diabetes and chronic kidney disease) five years in advance and dynamically quantify the disease risk.
Guangyu also established a systematic evaluation approach for the verifications and applications of intelligent medical models. She proposed a comprehensive adversarial attack framework that combined software-hardware tests and mutual validation between simulation and experiment. This framework can help the generalization of medical artificial intelligent systems.
In the future, Guangyu plans to carry out research on cognitive science-inspired approaches for intelligent semantic computing. Also, she is dedicated to promoting privacy protection, data-sharing, and computation using multi-source and multi-mode health and medical big data based on secure federated learning. She hopes that her research will create new digital healthcare paradigms and transform medical technology development into disease treatment and prevention, and health management.
