Multiple Sclerosis Identification by Grey-Level Cooccurrence Matrix and Biogeography-Based Optimization

This paper presents a new method amongst developing computer vision algorithms for the detection of multiple sclerosis (MS). Lesions caused by MS are detectable on MRI images. CV algorithms present subjective approaches in detection. In this study, we used the grey-level co-occurrence matrix to extract detailed texture features from the spatial distribution of greytone on MRI images. Multi-layered feedforward neural network was used as the classifier. Then, we selected biogeography-based optimisation algorithm to train this classifier. Through cross-validation, the method achieved sensitivity, specificity and accuracy of 92.75±1.31%, 92.76±1.65%, and 92.75±1.43% respectively. We validated the efficiency of the classifier, but overall, the method is inferior to state-of-art algorithms of MS lesion detection in all aspects.