IBIIS-AIMI Seminar: Computationally Modeling the Neuroradiologist - Andreas Rauschecker, MD, PhD

Abstract: Most attempts to "solve" radiology with AI map images directly to an output: in the past to a diagnosis, and in the modern age to a radiology report. We have taken a different approach, modeling how a neuroradiologist actually works by extracting clinically-relevant features from a brain MRI and then reasoning over them. The resulting diagnostic engine, built in 2018, approached neuroradiologist-level diagnostic accuracy across fifty neurologic diseases, and yet, it did not change clinical practice. Understanding its limitations has shaped everything my lab has done since. In this talk I will trace the path from that diagnostic engine to disease-agnostic quantification of clinical brain MRI, and then to brain tumors, where segmentation, uncertainty quantification, longitudinal modeling, and reading-room deployment converge on measurements clinicians can act on. I will argue that quantitative descriptions of abnormality, rather than diagnosis, is what makes personalized care from imaging possible across the long tail of diseases radiologists see and questions they are asked to answer. I show how this approach leads not only to improved radiology reporting but ultimately to a more quantitative understanding of neurological disease.

About: Andreas Rauschecker, MD, PhD, is Associate Professor In-Residence in the Department of Radiology & Biomedical Imaging at UCSF, where he is Co-Executive Director and Clinical Director of the Center for Intelligent Imaging (ci2) and Program Director of the Neuroradiology Fellowship. He is a practicing neuroradiologist. As Principal Investigator of the Brain Research and Artificial Intelligence in Neuroimaging (BRAIN) lab, his research focuses on developing and applying AI algorithms to analyze and quantify clinical brain MRI across a range of pathologies, with the aim of improving our understanding of neurological disease and translating these methods into clinical practice. He completed his MD/PhD in Neuroscience at Stanford (under the mentorship of Brian Wandell), his radiology residency and informatics fellowship at the University of Pennsylvania, and his neuroradiology fellowship at UCSF.

Attendance is open to the Stanford and AIMI affiliate community. Please contact aimicenter@stanford.edu for the Zoom link if you would like to attend virtually.