August 18, 2020

expert reaction to a study looking at using deep learning and MRI scans

A study, published in the American Journal of Roentgenology*, has investigated using deep learning to algorithms to improve the resolution of MRI scans in a shorter period of time.

Prof Paul Leeson, Professor of Cardiovascular Medicine, University of Oxford, said:

“MR imaging is incredibly powerful for studying changes in body tissues but the patient can sometimes need to stay in the scanner for several minutes to get the picture. One way of shortening this time is to collect less datapoints but images can become blurry – a bit like opting for a lower resolution on a camera.

“This research has studied whether a deep learning AI algorithm, combined with more traditional image processing techniques, can create precise-looking medical images from ‘lower resolution’ MR scans. They have been successful. Scan time has been dramatically shortened and images created that the radiographers thought looked very detailed.

“A downside of this approach is that, although the images look real and precise, the amount of medically relevant information in the picture has been reduced. Essentially, gaps in the picture have been ‘filled in’ with realistic looking content to fool the viewer.

“However, this may not matter. For many diseases, small changes are not necessarily important to know about for the health of the patient. Also, a technique that increases use of MR scans might actually help identify more disease as less reliance is placed on simpler, less precise, imaging tests.

“Whether to introduce this AI method into regular use to image the knee, or apply it to other parts of body, needs careful thought. Ultimately, proper randomised trials, focused on the patient, will help to prove whether these type of medical image enhancement tools enhance health or just the images!”

* ‘Using Deep Learning to Accelerate Knee MRI at 3T: Results of an Interchangeability Study’ by Recht et al was published in American Journal of Roentgenology on Tuesday 18^th August.

DOI: 10.2214/AJR.20.23313

Declared interests

Professor Paul Leeson: “Prof Leeson is a founder and non-executive director of Ultromics Ltd a medical imaging AI company”