Deep Learning Is Revolutionising MRIs

Deep learning is one of the most exciting branches of computer science today. A subset of machine learning, deep learning works with artificial neural networks to create algorithms that can learn and improve on their own. Deep learning algorithms are designed to imitate how humans think and learn and handle complex subjects. Deep learning is now being used in the field of MRI imaging. Two researchers have been able to use deep learning to correct distortions in MRI images, in order to help researchers and radiologists better interpret brain scans. 

Bennett Landman, professor of electrical engineering and computer science and radiology and radiological sciences, and Kurt Schilling, research assistant professor of radiology and radiological sciences published an article, “Distortion correction of diffusion weighted MRI without reverse phase-encoding scans or field-maps” in the journal PLOS ONE, demonstrating just how far deep learning has gone in the field.

Landman and Schilling created a technique using deep learning algorithms that corrects MRI images in order to gain more accurate insights from brain scans. The march of science is possible only through the taking of measurements. Incorrect images distort not only what we should see, but what can be learnt. The understanding of the brain’s size/volume or our ability to interpret the connections of brain pathways, are compromised. It becomes impossible to measure, to observe, to describe and therefore, to conduct insightful and correct neurological research.”

As Medical Xpress notes, it is common to have distorted images of the brain, with images squashing or pulling three-dimensional objects in ways that take the image far from what the original object looks like. The stakes in understanding the image are considerably higher when we are dealing with the most important organ in the body, the human brain, and we are trying to understand the progress and nature of a disease or disorder. 

The new algorithm, Synb0-DisCo, that Landman and Schilling discuss, was developed by the core faculty at the Vanderbilt Institute for Surgery and Engineering and the Vanderbilt University Institute of Imaging Science. It uses anatomically correct images to determine what the MRI image should look like and uses that data to correct the MRI scan that was acquired. Much in the same way that the human brain can interpret what someone is trying to say even if they make grammatical or spelling errors, Synb0-DisCo is able to do something similar with MRI images. 

To achieve this, Synb0-DisCo was trained over thousands of images, including legacy images from the Human Connectome Project, a large-scale National Institutes of Health-funded project that constructed a complete map of the structural and functional neural connections of the brain, the Autism Brain Imaging Data Exchange and the Baltimore Longitudinal Study of Aging, the largest effort to understand how the brain develops in aging. These images are frequently used by researchers across the world who study the human brain. 

This is an exciting innovation which will someday be available in every MRI imaging center. The number and quality of insights which become possible with the application of deep learning to MRI imaging is huge and preoperatively, decisions on whether to operate, what to treat and how, can be made with much better information.