MIT Researchers Combine Deep Learning and Physics to Fix MRI Scans

In the realm of medical imaging, MRI stands out for its exceptional visualization of soft tissues, surpassing the capabilities of X-rays and CT scans. However, the Achilles’ heel of MRI lies in its susceptibility to motion artifacts – even the slightest movement during a scan can compromise image quality, costing considerable resources and time.

This can also lead to potentially misleading diagnoses and suboptimal treatment outcomes. But now, MIT researchers are armed with a new deep learning model that is designed to rectify motion-related distortions in brain MRI.

The leading author of the paper, Nalini Singh, who is also a researcher affiliated with the Abdul Latif Jameel Clinic for Machine Learning in Health at MIT, explains to MIT News, “Motion is a common problem in MRI,…It’s a pretty slow imaging modality.”

To address this challenge, Singh and her team have devised a pioneering solution named “Data Consistent Deep Rigid MRI Motion Correction.” This method constructs motion-free images from distorted data, all without altering the scanning process.

Central to this hybrid approach is the preservation of coherence between the output image and the actual measurements it represents. Without this coherence, the model risks generating “hallucinatory” images – deceptively realistic but clinically inaccurate representations that can severely compromise diagnostic reliability.

Patients grappling with neurologically induced involuntary movement, as seen in Alzheimer’s and Parkinson’s diseases, would benefit from artifact-free MRI scans. Studies from the University of Washington Department of Radiology indicate that motion interferes with approximately 15% of brain MRI scans.

This recurring issue contributes to an annual expenditure of approximately $115,000 per scanner on repeat scans. Daniel Moyer, an assistant professor at Vanderbilt University said of the model, “This line of work from Singh and company is the next step in MRI motion correction.”

He went on to say, “Not only is it excellent research work, but I believe these methods will be used in all kinds of clinical cases: children and older folks who can’t sit still in the scanner, pathologies which induce motion, studies of moving tissue, even healthy patients will move in the magnet.”

Moyer, Concluded by saying “In the future, I think that it likely will be standard practice to process images with something directly descended from this research.”