Computerised Segmentation of Medical Images Using Neural Networks and GLCM

This article proposes a novel method combining the Neural Networks (NN) and the features of Gray Level Co-Occurrence Matrix (GLCM) for segmenting Region of Interest (ROI) of multiple medical images. The proposed methodology combines the texture of different pixels of medical images with the Radial Bias Function Neural Networks (RBFNN) in order to increase the performance of ROI segmentation and to obtain an optimal segmented region. The proposed approach works in two steps. Initially, the image borders are detected in order to separate the background skin and the ROI. This starts by extracting the GLCM features by the process of texture analysis which represents the ROI border clearly. GLCM features such as energy, homogeneity, contrast and the correlation are extracted. Secondly, the extracted features are passed towards the RBFNN for generating the ROI as segmented area. 870 scans of multiple medical images stored in a database are used for analysing the accuracy of the proposed methodology. Analysis of accuracy shows that the proposed methodology segments the ROIs of multiple medical images more accurately.