Perfusion MRI in automatic classification of multiple sclerosis lesion subtypes

This retrospective and exploratory study investigated the efficiency of 3T perfusion magnetic resonance imaging (MRI) at classification MS lesion subtypes. For subtype classification, firstly, it was necessary to segment all lesions. Therefore, a Bayesian classifier based on adaptive mixture method used lesions, an artificial neural network (ANN) employed multi-layer Perceptron as classifier. The accomplished segmentation lesions using Fluid Attenuated Inversion Recovery automatically, ANN part that worked extracted information from MRI (i.e. Mean Transit Time Cerebral Blood Volume maps) along with intensity conventional multi-channel in segmented Adding 3-Tesla proposed model for led increment about 7% 13% sensitivity acute chronic classifications, respectively. T2 did not meaningfully change. overall accuracy acute, chronic, classifications 96.1%, 90.5%, 92.9%, architectures reached high discrimination between subtypes when MRIs were used.