‘Machine learning’, a type of computer modelling, can detect areas of brain damage (lesions) associated with drug-resistant epilepsy. This is according to a new study published in the scientific journal PLOS One.During the study, researchers led by Dr Carole Lartizien, from the University of Lyon, developed a complex system that is able learn features associated with healthy brain MRI scans. It can then be used to assess other MRI scans for abnormalities (i.e. how much they deviate from the normal ‘patterns’ it has learned).The scientists focused on two parameters, detectable on MRI images, that are associated with lesions linked to epilepsy. These are the presence of grey matter within the white matter (heterotopia), and a blurred junction between grey and white matter.Dr Lartizien and her team tested the performance of the system on brain MRI images from 11 people with drug-resistant epilepsy, who had a total of 13 lesions in their brains, and 77 healthy volunteers.They found that the system was able to detect all of the lesions in the MRI positive cases (where the lesions on MRI images were detectable by experts) and 70% of lesions in MRI negative cases (where the lesions weren’t visible to experts). In addition, the rate of false positives (i.e. cases where the system “thought” there was a lesion when there was not) was low.The authors conclude that machine learning “may be a versatile system for unbiased lesion detection.” If used in clinical practice, it could potentially allow people more prompt access to the most appropriate treatment for their condition.Computer-aided diagnosis based on MRI has been suggested in the past few years to help screen lesions associated with epilepsy. The researchers propose that their system could also be used to detect other conditions where lesions are formed in the brain, such as multiple sclerosis and dementia; to study ageing; and to predict risk factors associated with stroke.Author: Dr Özge ÖzkayaClick here for more articles about brain science including genetics.
ERUK Team2019-10-26T22:52:17+01:00September 13th, 2016|