Facebook AI research team and NYU School of Medicine have collaborated to investigate the benefits of employing artificial intelligence (AI) technology on MRI (magnetic resonance imaging) scan. The mission of the team is to boost the speed of the imaging process up to ten times and revolutionize medical care department.
Larry Zitnick representing the Facebook Artificial Intelligence Research (FAIR) group and Daniel Sodickson, M.D., Ph.D., and Michael Recht, M.D., from NYU School of Medicine have associated on this project named fastMRI. The social media giant revealed the details about this artificial intelligence research project in a blog post.
“If this effort is successful, it will make MRI technology available to more people, expanding access to this key diagnostic tool,” as mentioned in the post.
Underlying Purpose To Use AI on MRI Scan
The researchers at Facebook and NYU School of Science’s Department of Radiology noted that MRI scan processes typically take longer time than an X-ray or CT scan. And this time duration can pose a challenge to people especially children, claustrophobic or those who have a problem while lying down.
On top of this, the patients also have to hold their breath while scanning heart, liver, or other organs in the abdomen and torso. The accelerated MRI scan devices could reduce the time of bearing this painful exercise.
Moreover, there is a shortage of MRI scan machines in rural areas and developing regions resulting in a stretched backlog. In that case, by integrating a faster technology into MRI scanners, the lab doctors can diagnose a number of people in a short time.
“Increased speed could let MRI machines fill the role of X-ray and CT machines for some applications, allowing patients to avoid the ionizing radiation associated with those scans,” said researchers in the blog.
Recht, Sodickson, and Lui examine MRI scans of a knee at NYU Langone Health in August 2018. Source: Facebook Blog
fastMRI Project
For the project purpose, the collaborating companies have access to 10,000 clinical comprising about 3 million MRI reports of the knee, brain and liver. The data was collected exclusively by NYU School of Medicine.
Facebook mentioned that the names along with critical health information of all participating patients have been removed from the data to protect their identity.
“The work is fully HIPAA-compliant and approved under NYU Langone’s Institutional Review Board, which oversees all human subject research at the medical center,” added in the post.
Employing AI on MRI Scan
The MRI scan machines currently capture data in a sequential pattern. This means if the area under the scanner is larger it will take more time to analyze the information. By using artificial intelligence, the fastMRI team determines to accelerate this time period.
The researcher team wants the scanning machines to capture less but relevant data and save on time for faster processing. It is similar to human behavior where, at times, it accurately infers the complete situation from the partial pictures. The basic principle working here is reinforced learning.
“The key is to train artificial neural networks to recognize the underlying structure of the images in order to fill in views omitted from the accelerated scan.”
