A recently published study shows how an AI algorithm could be used to provide accurate MRI results at a faster rate with 4 times fewer data. To top it all off, the existing imaging hardware does not need to be updated for the AI algorithm to work. This new study is the result of a collaboration between a group of radiologists from New York University’s Grossman School of Medicine and Facebook’s Artificial Intelligence Research (FAIR) team.
The current MRI scan method is the best one available but not without its disadvantages, such as the small space could trigger claustrophobia, and it is time-consuming, limiting the scans done per day. The new study could bridge that gap.
The new project, entitled fast MRI brings an AI model that can effectively reduce the data required for scanning to ¼ of the original.
According to the Facebook team, “Generating accurate images is not the only challenge. This AI model must also generate images that are visually indistinguishable from traditional MRI images. Radiologists will spend a lot of time carefully analyzing these images instead of The familiar look and feel may make radiologists less likely to adopt fast MRI in their practice.”
6 independent musculoskeletal radiologists tested out the two MRI images chosen from 108 test patients. The radiologists evaluated the results of both the data sets and came to the conclusion that essentially both the data sets displayed the identical results. In fact, the study also pointed out that radiologists believe that the overall quality of AI images is better than traditional MRI images.
The results of the fast MRI project are open source so that the MRI hardware vendors can integrate the new set of algorithms into their products as quickly as possible for them. The research team is hoping that in the next few years, this innovation could be easily integrated into the existing MRI hardware, making it easier for both the patients and the hospital.
