Building Deep Learning Models for Medical Imaging with Limited Training Data
Abstract:
Advances in medical imaging and artificial intelligence have created new opportunities to detect and characterize disease, assist clinical interpretation, and extract physiological information that was previously difficult to obtain. Despite these advances, developing reliable deep learning models remains challenging when training data are limited.

In this talk, I will provide an overview of selected projects in my lab, with an emphasis on strategies for addressing the scarcity of training data. Examples will include hemodynamic assessment from 4D flow MRI and image enhancement of 4D flow MR images, as well as disease assessment from thoracic CT, among other ongoing imaging projects. These will illustrate different strategies for model development and highlight the opportunities and practical challenges of applying deep learning to medical imaging when there is limited training data.
Bio:
Amir A. Amini received his Ph.D. in Electrical Engineering and Computer Science from the University of Michigan. He is currently Endowed Chair in Bioimaging and Professor of Electrical and Computer Engineering at the University of Louisville. His previous faculty appointments were at Yale University and Washington University in St. Louis. He has held numerous leadership roles in medical imaging and image-analysis conferences, serving as a scientific program committee member, program chair, and conference chair. He was Symposium Co-Chair of SPIE Medical Imaging in 2007 and the IEEE International Symposium on Biomedical Imaging in 2018. He has served as an Associate Editor for IEEE Transactions on Medical Imaging, IEEE Transactions on Biomedical Engineering, IEEE Reviews in Biomedical Engineering, IEEE Open Journal of Engineering in Medicine and Biology, and Computerized Medical Imaging and Graphics. He served as Vice President for Publications of the IEEE Engineering in Medicine and Biology Society from 2020 to 2021. He became Editor-in-Chief of IEEE Transactions on Biomedical Engineering in 2025.

With support from the NIH, NSF, private foundations, and industry, Dr. Amini’s laboratory has conducted research on the development and application of MRI methods for motion and flow measurement. His laboratory also develops deep learning-based biomedical image-analysis methods across multiple imaging modalities, with applications that include computer-aided diagnosis and radiation therapy for lung cancer and heart failure diagnosis from echocardiography. He received the University of Massachusetts Amherst College of Engineering Distinguished Alumni Award in 2020. He was elected a Fellow of IEEE in 2007, the American Institute for Medical and Biological Engineering in 2017, SPIE in 2018, the Asia-Pacific Artificial Intelligence Association in 2021, and the International Academy of Medical and Biological Engineering in 2024.