|
Abstract
Deep learning methodologies have demonstrated noteworthy success in the realm of medical image analysis, often delivering accuracy comparable to that of human experts. As efforts progress to integrate these models into clinical environments, however, significant challenges arise concerning model robustness, interpretability, and generalizability. Our studies reveal that deep learning models often fall short in terms of robustness and generalizability when applied to images with slight differences compared to the training sets. This problem also impacts model interpretability, with trustworthiness of explanations often being neglected. This presentation delves into our recent research addressing these critical issues via innovative machine learning techniques. Additionally, we will explore the implications of large language models for medical image analysis. The conversation will highlight how these models can potentially transform the way we comprehend and interact with medical imaging data.
|
|
Bio
Dr. Pingkun Yan is the P.K. Lashmet Career Development Chair Associate Professor at the Department of Biomedical Engineering at Rensselaer Polytechnic Institute (RPI). Before joining RPI, he was a Senior Scientist of Philips Research working at the clinical site at the National Institutes of Health (NIH). His research focuses on translational medical imaging informatics and image-guided intervention using artificial intelligence and machine learning techniques through close collaboration with clinicians.
Dr. Yan has published over 80 peer-reviewed articles in well-recognized journals including Nature Communications, PNAS, Medical Image Analysis, IEEE T-MI, IEEE T-CSVT, IEEE T-BME, IEEE T-ITB, Medical Physics, and top international conferences including MICCAI, ICCV, CVPR, and ISBI. His publications have been cited more than 7,700 times. His research work has also resulted in 10+ patent filings and issued patents.
Dr. Yan is a recipient of the NSF CAREER award and NIBIB Trailblazer award. He received the MICCAI Best Paper Award in 2005 for his work on segmenting blood vessels from magnetic resonance angiography by modeling the capillary action. In 2008, he was recognized as one of the four finalists of the Innovation in Industry Award by the New York Academy of Sciences, which stood out from 120 nominees across the state of New York, for his contribution to prostate cancer diagnosis.
He is currently serving as an associate editor of multiple international journals, including Machine Vision and Applications (Springer) and Neurocomputing (Elsevier). He co-organized 6 international workshops and 4 international journal special issues and has served as a program committee member for more than 50 international conferences. He is also a regular reviewer of a large number of international journals and conferences. Dr. Yan is a senior member of the National Academy of Inventors (NAI) and the Institute of Electrical and Electronics Engineers (IEEE).
|