A Conceptual Model for Segmentation of Multiple Scleroses Lesions in Magnetic Resonance Images Using Massive Training Artificial Neural Network

Detecting abnormalities in medical images is one application of image segmentation. MRI as an imaging technique sensitive to soft tissues such as brain shows Multiple Scleroses lesions as hyper-intense or hypo-intense signals. As manual segmentation of these lesions is a laborious and time consuming task, many methods for automatic brain lesion segmentation have been proposed. To tackle difficulties of Multiple Scleroses lesion segmentation we have proposed a conceptual model based on MTANN, as a method for training artificial neural networks to detect abnormalities in medical images. The proposed model has three main phases namely, Pre-Processing, Segmentation, and False Positive/Negative Reduction. In the segmentation phase, feature extraction and selection are done automatically using MTANN. The Fuzzy Inference System reduce false positives/negatives in the last phase. As advantage of proposed model, it is supposed to produce accurate lesion mask using just FLAIR MRI that reduce computational time and brings comfort for patients.