Date: Wednesday, Dec 3rd, 2025
Time: 10:00 AM Central Time
Location: Zoom
Title: “AI-Driven Advances for Fast, Affordable, and Personalized MRI”
Abstract: Machine learning (ML) is opening new possibilities for the design and operation of medical imaging systems. This talk will present recent advances from our lab toward intelligent, ML-based magnetic resonance imaging (MRI). First, I will introduce a novel healthcare paradigm that enables rapid, affordable longitudinal imaging by combining standard high-field MRI exams with scans using new-generation low-field systems. Those are small, compact, inexpensive, and portable, but they suffer from reduced image quality and low SNR. To address this, we developed ViT-Fuser, a vision-transformer that learns personalized features from the subject’s past high-quality scans and leverages them to boost image quality in point-of-care, low-cost scans. Next, the talk will focus on data acquisition. Although recent ML methods have enabled automated optimization of MRI acquisition protocols, those studies rely on supervised learning from targets. Therefore, they limit the possibility of discovering novel protocols and image contrasts. To overcome this, we developed a framework for self-supervised data-acquisition optimization that maximizes tissue contrast without any targets; experiments with in vivo brain MRI data demonstrated its effectiveness. Together, our methods leverage ML to advance affordable and accessible healthcare.
Bio: Prof. Efrat Shimron heads the Medical Artificial Intelligence (AI) and Magnetic Resonance Imaging (MRI) Lab at the Technion. Her research develops AI-driven methods for fast, robust, and affordable MRI, combining signal processing, machine learning, physics, and medical science. She also develops techniques for next-generation, compact low-field MRI systems. Prof. Shimron is a recipient of the European Research Council (ERC) Starting Grant, the MIT Rising Stars award, and other awards.