Artificial Intelligence has been making impressive strides in the past year or so, with a number of medical applications utilizing AI to spot problems in medical imagery more effectively and efficiently than current methods.
For instance, we’ve had a couple of projects using the approach to better identify cancer, eye problems, and liver disease.
A recent study has set out to do a similar feat to help researchers detect epilepsy in children. The research, which was a collaborative project between Young Epilepsy, UCL Great Ormond Street Institute of Child Health and the University of Cambridge, focused on Focal Cortical Dysplasia, which is a major cause of epilepsy in children. It describes the way the brain fails to form normally, and because the abnormalities tend to be small, they tend to be very difficult to pick up on MRI scans.
What’s more is that because the child’s brain is rapidly developing, it’s very easy for such things to fly under the radar, even for the most highly trained radiologists.
So, the researchers turned to Machine Learning to help rapidly identify these abnormalities in children. Firstly, the subtle abnormalities in the brain were identified by a pediatric neuroradiologist before these were transformed into a range of features, including the thickness and folding of the brain, that could be used to train the algorithm.
When the algorithm was put through its paces, it was able to correctly identify the brain abnormality in 73% of patients. This is comparable to similar tests performed on adults, but there aren’t really any existing methods to compare against in children. The results are impressive because the adult brain is relatively stable compared to the rapidly developing brain of a child.
The hope is that it will ease the identification of more young people who could go on to receive life-changing treatments to ensure they don’t spend their adulthoods with epilepsy.
The next step for the project is to test out the algorithm on a number of more complex brain scans before then investigating how this technology can be effectively incorporated into current clinical practice. So, still very early days, but a promising start.