Artificial Intelligence Can Help Improve MRI Scanning Speed

Published in Radiology, a study from the NYU Gross School of Medicine and Meta AI Research found that artificial intelligence (AI) can be used to reconstruct magnetic resonance imaging (MRI) scans faster than standard methods, leading to improved access and shorter wait times.

Despite the various benefits that accompany traditional MRI scans, there are several flaws in the process, such as the long wait times that patients may experience.

In 2018, NYU Langone Health and Meta AI Research began an effort known as fastMRI that aimed to use AI to enhance the speed of MRI scanning. Previously, in a proof-of-concept study, the NYU Langone team simulated accelerated scans by removing 75 percent of raw data acquired by conventional MRI scans. The fastMRI scans constructed images comparable to those of the standard scans.

In the new study, researchers only used 25 percent of the data to perform accelerated scans and had the AI model fill in the missing information. The study involved 170 participants with a diagnostic knee MRI using both a conventional MRI and an accelerated AI protocol.

Six musculoskeletal radiologists reviewed the MRI examinations to find meniscal or ligament tears, as well as bone marrow or cartilage abnormalities. To lower the bias risk, they were unaware of which images were reconstructed with AI.

Following evaluation, they concluded that the AI-reconstructed images were similar to the standard images for identifying tears or abnormalities. They also found that the image quality of accelerated scans exceeded that of the standard images.

“Our new study translates the results from the earlier laboratory-based study and applies it to actual patients,” says Michael P. Recht, MD, the Louis Marx professor of radiology and chair of the Department of Radiology at NYU Grossman School of Medicine, in a press release. “FastMRI has the potential to dramatically change how we do MRI and increase accessibility of MRI to more patients.”

Aside from the efficiency of fastMRI, researchers also noted how it could expand access to MRI exams, as the data, models, and code associated with it are available to external researchers as an open-source project. Researchers also noted that fastMRI is likely to perform examinations in a time similar to X-rays or CT scans but produces higher-quality information.

“This research represents an exciting step towards translating AI accelerated imaging to clinical practice,” says Patricia M. Johnson, PhD, assistant professor in the Department of Radiology, in the press release. “It truly paves the way for more innovation and advancements in the future.”

Similarly, a study published in March 2022 discussed the creation of a computer vision model that sped up and improved the quality of head MRI exams. Researchers decided to focus on axial T2-weighted scans, which occurred in more than 90 percent of MRIs. They created computer vision models through dataset labeling, image pre-processing, model interpretability, experiments, and simulation study.

Researchers found that the highest-performing model could assist in the production of faster and more accurate MRIs.