Special Issue Article
Comparative Study of De-Noising, Segmentation, Feature Extraction, Classification Techniques for Medical Images
Early identification and classification of brain tumors play a major role in the diagnosis of tumors. This paper attempts to take the study of various preprocessing, segmentation, feature extraction and classification techniques which are needed to efficiently extract the tumor region and classify them according to their grades from the MR brain images. Pre-processing of medical images is needed because the image may be degraded by noise either during transmission or acquisition. Filtering techniques are effective only if they preserve edges during pre-processing. Tumor region is extracted from the MR brain images using the various segmentation techniques. Segmentation is effective if it includes the spatial information as well as the global intensities of the image. Since the pixels in the images are highly correlated, statistical features are best suited for the optimistic classification. The various classifiers and their accuracy in terms of sensitivity, specificity for the classification of tumors are studied for the selection of appropriate classifier.