$125,000 Award

Project Overview

MRI is a useful imaging study to diagnose disease throughout the body. Unfortunately, MRIs are time-consuming, expensive, and uncomfortable for patients. This team proposes to use the latest developments in artificial intelligence (AI) using a set of techniques called deep learning (DL) to make MRI faster. Specifically, they propose to use DL to ‘synthesize’ or ‘switch’ one MRI sequence into another, which could speed up the MRI acquisition process by multiple fold. They will build on prior knee MRI work and expand to other body parts, as well as perform external validation of their AI FastMRI tools. The interdisciplinary team is assembled of experts across diagnostic radiology, imaging informatics and IT, and computer science. The ultimate goal is to take this promising technology from bench to bedside to improve hospital efficiency and the patient experience.

Project Team

PIs:

  • Paul Yi, MD
  • Vishwa Parekh, PhD

Research Assistants: To be named

Imaging Informatics:

  • Michael Toland (UMMS Imaging IT)
  • Peter Kamel, MD
  • Eliot Siegel, MD
  • Garzia Zapata, MD

Midpoint Progress Updates – October 2023

The team has installed hardware purchased with the Innovation Challenge award funding which enabled them to successfully train an MRI knee model. This model effectively shows high similarity scores for synthetic T2 fat saturated images from T1/PD images. Future plans include fine turning the MRI knee models and then moving on to other body parts, most likely starting with the brain.