Tencent Jarvis Lab is dedicated to developing advanced medical artificial intelligence algorithms and deploying them in daily clinical practice. However, real clinical data is often imperfect with various issues, e.g., large variations of images from different hospitals, scarcity of annotated labels, and discrepancy of labels provided by different physicians. To cope with these challenges, we propose various methods to learn from such imperfect clinical data to build a robust medical image analysis system. In this talk, I will present some of our recent work on image style transferring to reduce the variations of images from different hospitals, small-sample learning to cope with label scarcity, and noisy-label learning to directly learn from potentially inconsistent annotations from multiple experts to leverage the intrinsic information from the raw labels.
Yefeng Zheng received B.E. and M.E. degrees from Tsinghua University, Beijing, China in 1998 and 2001, respectively, and a Ph.D. degree from University of Maryland, College Park, USA in 2005. After graduation, he worked at Siemens Corporate Research in Princeton, New Jersey, USA on medical image analysis before joining Tencent in Shenzhen, China in 2018. He is now a Distinguished Scientist and Director of Tencent Jarvis Lab, leading the company’s initiative on medical AI. He has published 200+ papers on top journals/conferences and invented 70+ US patents. His work has been cited more than 10,000 times with an h-index of 55. He is a Fellow of Institute of Electrical and Electronics Engineers (IEEE), a Fellow of American Institute for Medical and Biological Engineering (AIMBE), an Associate Editor of IEEE Transactions on Medical Imaging, and was a Program Co-Chair of International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) 2021.