Experimental Analysis of ANN Based Tissues Segmentation and Classification of Brain MR Images

A fully automatic procedure for brain tissue classification of single channel magnetic resonance images (MRI) of human Brain is described. Two different ANN classifiers namely Learning Vector Quantization (LVQ), Multilayer Perceptron (MLP) are used for segmentation (classification) of tissues in brain MR images. Each tissue type are segmented and assigned with different gray shades for representation. Validation experiments were performed on real MRI data. Some similarity measure is (automatically) computed between the classification results and a reference gold standard classification results. For this dataset, the gold standard is the manual classification. The results obtained from these two classifiers are then compared with conventional K-means clustering. From this analysis, it is observed that LVQ ANN performs better and faster compared to MLP and K-means clustering.