Clinical Day

Massachusetts Institute of Technology

Wednesday | April 8, 2026 | 8:30 – 9:15

Abstract: Computational physiological models are increasingly proposed as decision-support tools in medicine, yet their adoption is limited by unresolved questions of trust, transparency, and predictive credibility. This talk uses a physics-informed cardiovascular modeling framework—Windkessel Physics-Informed Neural Networks (WPINNs)—as a concrete case study to examine how Verification, Validation, and Uncertainty Quantification (VVUQ) can be operationalized for biological systems. We show how embedding mechanistic cardiovascular models directly into learning enables parameter identifiability, physiological interpretability, and robust prediction under sparse and biased data, while exposing clear domains of validity. Through perturbation-based validation using real human data and synthetic ground truth, we demonstrate feature-level agreement—directionality, gain, time scales, and causal relationships—rather than reliance on pointwise accuracy alone. Finally, we illustrate how physics-residual-based uncertainty serves as a transparent indicator of model reliability, linking prediction confidence to violated assumptions. These results argue that trustworthy computational models for medicine require VVUQ frameworks that are mechanism-aware, perturbation-driven, and decision-relative, rather than purely data- or accuracy-centric.

Biography:

Dr. Roozbeh Jafari is a principal investigator at the MIT. He is also an adjunct professor in Electrical and Computer Engineering at Texas A&M University. He joined MIT from Texas A&M where he was the Tim and Amy Leach Professor in Electrical and Computer Engineering and in the School of Engineering Medicine. He was formerly a Principal Staff at MIT Lincoln Laboratory.

Jafari received his PhD in computer science from the University of California, Los Angeles, and completed a postdoctoral fellowship at the University of California, Berkeley. His research interests lie in the areas of wearable computer design, sensors, systems, and AI for digital health paradigms and, most recently, digital twin for precision health. His laboratory developed groundbreaking technologies including smart rings, smart watches and e-tattoos for continuous cardiovascular monitoring, cuffless blood pressure sensing systems, and precision medicine digital twins. He has raised more than $89 million for research with $25 million directed toward his lab. His research has been funded by the NSF, NIH, DoD (TATRC), DTRA, DIU, AFRL, AFOSR, DARPA, SRC, and industry (Texas Instruments, Tektronix, Samsung & Telecom Italia). He has published more than 250 papers in refereed journals and conferences. He has served as the general chair and technical program committee chair for several flagship conferences in the area of wearable computers. Jafari is the recipient of the National Science Foundation CAREER award (2012), the IEEE Real-Time & Embedded Technology & Applications Symposium best paper award (2011), the Andrew P. Sage best transactions paper award (2014), the ACM Transactions on Embedded Computing Systems best paper award (2019), the William O. and Montine P. Head Memorial research award for outstanding engineering contribution award from the College of Engineering at Texas A&M University (2019), the dean of engineering excellence award at Texas of A&M University (2021), and the TEES research impact award at Texas A&M University (2021). He has also been named the Texas A&M Presidential Fellow (2019).

Jafari serves on the editorial board for Nature Digital Medicine, the IEEE Transactions on Biomedical Circuits and Systems, the IEEE Sensors Journal, IEEE Internet of Things Journal, IEEE Journal of Biomedical and Health Informatics, IEEE Open Journal of Engineering in Medicine and Biology, and the ACM Transactions on Computing for Healthcare. He is the past chair of the IEEE Wearable Biomedical Sensors and Systems Technical Committee as well the IEEE Applied Signal Processing Technical Committee (elected). He serves on scientific panels for funding agencies frequently, has served as a standing member of the NIH Biomedical Computing and Health Informatics (BCHI) study section (2017-2021), and was the inaugural chair of the NIH Clinical Informatics and Digital Health (CIDH) study section (2020-2022). He is a fellow of the American Institute for Medical and Biological Engineering (AIMBE) and the Institute of Electrical and Electronics Engineers (IEEE).

University of Tel-Aviv

Wednesday | April 8, 2026 | 9:15 – 10:00

Abstract: The rapid evolution of artificial intelligence is fundamentally reshaping the landscape of medical imaging and diagnostic workflows. In this talk, I will discuss the integration of advanced AI to transform healthcare support and clinical decision-making. I will focus on three key themes: First, I will explore how Foundation models pre-trained on general-purpose imagery can be adapted to enhance medical image detection and segmentation. Second, I will demonstrate how Generative modeling facilitates earlier disease detection, enabling more proactive screening and personalized patient care. These methods are applied across X-ray, CT, and MRI modalities, with a specific focus on liver tumor and Pulmonary Embolism detection. I will conclude by presenting our latest Multimodal fusion and Prediction models, which integrate imaging with clinical data. This holistic approach provides a comprehensive view of the patient, ultimately driving the next generation of clinical decision support tools.

Biography: Prof. Greenspan heads the Medical Image Processing Lab in the School of Biomedical Engineering at Tel-Aviv University. In 2021, she joined the Icahn School of Medicine at Mount Sinai, NY, as Co-Director of the AI and Emerging Technologies (AIET) PhD concentration. In 2010, Dr. Greenspan Co-founded RADLogics Inc., a startup providing AI tools to Radiologists and Oncologists. Professor Greenspan has authored over 250 publications in leading international journals and conferences, with more than 31,000 Google Scholar citations (H-index: 71). She is a member of SPIE, IEEE_ISBI and MICCAI. She served as Lead Co-editor for the 2016 IEEE TMI Special Issue on Deep Learning and co-edited the first book on Deep Learning for Medical Image Analysis, in 2017. She served as Program Chair for IEEE ISBI 2020 and MICCAI 2023 and will serve as the Keynote Chair for MICCAI 2027. Dr. Greenspan is consistently ranked in the top 1-2% of cited researchers in the fields of medical imaging and biomedical engineering (Stanford university ranking). In 2019 she was selected to be one of “Top-30 Women AI Leaders in Drug Discovery and Advanced Healthcare.” In recognition of her contributions to medical computing, she was named a MICCAI Fellow in 2024.

Aival

Wednesday | April 8, 2026 | 10:15 – 11:00

Abstract: Recent years have seen a massive increase in the number of clinical AI products for image interpretation available on-the-market. The usefulness of these products in a clinical setting depends on criteria that are far wider than the benchmarks often used in academia, and that require knowledge of how the products will be used in the clinical workflow. With the growing capabilities of AI and the availability of data, there is an ever clearer path for algorithms developed in academic research to impact real-world clinical practice. In this talk I will highlight some of the important considerations when evaluating AI products that may help to inform algorithmic development for practical use.

Biography: Kanwal is the Founder of Aival, which is building the data infrastructure layer for healthcare AI. She has been working in medical imaging AI since finishing her PhD in 2007: first in developing novel algorithms then in commercialising these through industry and startups. Aival unifies a hospital’s multimodal datasets to be able to rapidly train, evaluate, and monitor advanced AI models while retaining data sovereignty.

School of Computing and Data Science, The University of Hong Kong

Wednesday | April 8, 2026 | 11:00 – 11:45

Abstract: Computational pathology is transforming diagnostic practice by leveraging artificial intelligence to extract clinically relevant insights from Whole Slide Images (WSIs). The integration of multimodal AI offers new opportunities for building interpretable, accurate, and scalable diagnostic tools. In this talk, I will present our recent advances that demonstrate how incorporating domain knowledge and biological context can significantly enhance histopathological analysis. We will first introduce a knowledge-guided framework that integrates expert-derived knowledge into AI models, enabling more generalizable and clinically meaningful predictions across diverse cancer tasks. We then showcase a strategy for inferring cellular-level phenotypes directly from histology images, providing a cost-effective alternative to spatial transcriptomics for characterizing the tumor microenvironment and predicting patient outcomes. Together, these works reflect a shift toward more human-aligned, knowledge grounded AI systems for computational pathology.

Biography: Dr. Lequan Yu is an Assistant Professor at School of Computing and Data Science, The University of Hong Kong, and a former postdoctoral fellow at Stanford University. He received his Ph.D. and B.Eng. from The Chinese University of Hong Kong and Zhejiang University, respectively. His research focuses on medical AI, multimodal learning, and precision oncology. He has published over 100 papers, including Nature Communications, npj Digital Medicine, Cell Gemonics, and TPAMI, with 22,000+ Google Scholar citations. Dr. Yu was ranked by Clarivate Analytics in the top 1% of the citation list in 2023-2025 and won the MICCAI 2023&2024 Young Scientist Publication Impact Award Runner-Up. He serves as the Associated Editor of npj Digital Medicine, the Guest Associated Editor of TMI, and the Area Chair of NeurIPS, ICLR, CVPR, AAAI, and MICCAI.

Case Western Reserve University (USA)

Wednesday | April 8, 2026 | 14:00 – 14:55

Abstract: Chemical contrast agents have long been integral to clinical diagnostic imaging. However, growing concerns about their safety, cost, and environmental impact have prompted the need for alternative solutions. In this talk, Dr. Shuo Li will present his pioneering work on AI contrast-enhanced imaging. This transformative approach leverages cutting-edge machine learning techniques to synthesize contrast-enhanced images without chemical agents. This innovative technology reduces patient risk and healthcare costs and opens new frontiers for precision imaging. Dr. Li will showcase recent breakthroughs from his lab, highlight clinical applications across cardiology, oncology, and neurology, and discuss the future potential of AI-driven imaging in reshaping medical diagnostics.

Biography: Dr. Li is a global leader in conducting multi-disciplinary research to enable artificial intelligence (AI) in healthcare. He is a Leonard Case, Jr. endowed professor at Case Western Reserve University (USA). Before that, he was an associate professor at Western University (Canada) and a scientist at the Lawson Health Research Institute. He was a scientist at GE Healthcare (2006-2015). He is a committee member in multiple highly influential conferences and societies. He is most notable for serving on the prestigious board of directors in the MICCAI society (2015-2024), where he is also the general chair for the MICCAI 2022 conference, which is the most influential AI-in-imaging conference. He has over 300 publications, acted as the editor for six Springer books, and serves as an associate editor for several prestigious journals in the field. Throughout his career, he has received several awards from GE, various institutes, and international organizations. He is a Fellow of SPIE, AAIA, IET, AIMBE, and IAMBE, and a member of the National Academy of Artificial Intelligence (NAAI).

USD AI Research, The University of South Dakota

Wednesday | April 8, 2026 | 15:25 – 16:20

Abstract: AI and clinical medicine are often portrayed as natural partners, yet their collaboration remains challenging. Drawing on NIH experience and extensive AI research, this talk explores why technically sophisticated models frequently fail in real clinical settings. It highlights mismatches in data assumptions, explainability, accountability, and workflow integration that fuel distrust between clinicians and AI scientists. While much effort focuses on building robust models, clinical relevance is often overlooked. Models trained on narrow or homogeneous datasets struggle to generalize, emphasizing the need for cross-population training and testing to ensure applicability across diverse patient populations. Using examples from medical imaging and precision medicine, the talk shows how algorithmic success alone does not guarantee clinical impact. True progress requires reframing AI as a collaborative tool that augments human expertise — human-in-the-loop machine learning (active learning). The session concludes with pathways toward human-centered, interpretable, and population-aware AI systems. Take-home message: learning, not just limited to training, drives meaningful clinical impact.

Biography: Prof. KC Santosh is the Chair of the Department of Computer Science (since 2020) and founding director of the USD AI Research (since 2015, https://ai-reseach-lab.org)) at the University of South Dakota (USD). He served as Graduate Program Director for seven years (2017–24) and was previously a research fellow at NIH and a PostDoc at INRIA (France). With over $8.7 million in funding (DOD, NSF, and ED, to name a few), he has authored 12 books and 290+ peer-reviewed research articles (such as IEEE TPAMI, IEEE TAI, and IEEE TMI) as well as delivered 90+ keynote talks, including TEDx talk. He serves as an associate editor for multiple prestigious journals such as IEEE Trans on AI, IEEE Trans of Medical Imaging and Int J of Machine Learning &Cybernetics, chaired multiple premier conference events such as IEEE Conference on AI, IEEE Conference on Cognitive Machine Intelligence and IEEE CBMS, and leads multiple review panels such as NSF and Mitacs (Canada). He is a trained leader, having completed leadership programs such as Deans/Chairs 1.0 (CCAS, Spring 2021), Deans/Chairs 2.0 (CCAS, Summer 2024), and the President Executive Leadership Institute (USD, 2021/22). He leads AI+x initiative that primarily includes curriculum innovation, outreach and visibility (USD AI Symposium, IEEE), and interdisciplinary research collaboration (e.g., South Dakota Biomedical Computation Collaborative – supported by a $6.5M award from the U.S. Department of Education). He brings extensive experience in curriculum innovation—particularly in interdisciplinary initiatives and shared governance (e.g., physics, business analytics, psychology, biology, and biomedical engineering, etc.)—as well as in program assessment and evaluation (as PEV), including ABET accreditation during the 2016/17 and 2022/23 cycles. His leadership has driven a 3,000% growth in AI enrollment at USD. His contributions have established USD as a pioneer in AI programs within the state of South Dakota. In fundraising, he secured a $2.0 million endowment promise to further strengthen the reputation of the department. He is a member of the NIST’s AI Standards and Innovation and a U.S. Speaker for AI education. For more info. https://kc-santosh.org/.