The Radiology Team at the Ulster Hospital has been working in partnership with King’s College, London on an innovative clinical study identifying abnormalities in MRI head scans.
MR Imaging Abnormality Deep Learning Identification (MIDI) study uses a new technique of ‘deep learning’ in computer science that automatically learns relevant features directly from MRI and CT images. This ‘deep learning’ algorithm will be trained on a large database of head MRI scans so that it will be able to pick up scans that have head abnormalities, by classifying as normal or abnormal. This technique will then be tested to determine its validity.
Leading the team on this ground breaking study, Principle Investigator & Radiology Consultant, Dr Ryan Wilson explained, “An increase in demand for MRI imaging coupled with a shortage of radiological expertise means increasing delays to reporting times. A triaging tool that identifies scans with abnormalities and prioritises them for reporting would be of great clinical benefit.
“It is our intention to use deep learning techniques to develop a decision-making tool that will flag abnormalities on head MRI scans for further attention. We are delighted to have been able to partner with King’s College London on this project.”
Deputy MRI Superintendent, Ross Mannus added, “The MIDI study will benefit patients by expediting the process of obtaining results, reducing patient anxiety and enabling healthcare providers to promptly initiate appropriate treatment plans.”
With a 4.6% increase in brain MRIs in one year and an increase in the time to report out-patient MRIs every year since 2012, an automated triage mechanism to identify abnormalities at the time of imaging and allow prioritised scan reporting, will be hugely beneficial in the clinical environment.