Computer aided classification and segmentation of abnormal human brain magnetic resonance images using modified soft computing techniques

Abstract

Abnormality detection in abnormal Magnetic Resonance (MR) brain images is a challenging task. The abnormality detection process involves the concept of image classification and image segmentation techniques. Initially, the abnormal images are categorized into different categories since the treatment planning varies for different categories. Further, the effect of the treatment can be analyzed by the image segmentation process which performs the volumetric analysis on the size and shape of the abnormal tumor portion. The challenges and difficulties of the abnormality detection process are mainly due to the requirement of classification and segmentation techniques which can yield accurate results within less convergence time. Conventional abnormality detection procedures are based on manual interpretation which is highly prone to error. Incorrect identification may lead to different treatment planning which may lead to fatal results. Apart from the inaccurate results, the time involved for manual classification and segmentation techniques are very high. Hence, automated techniques without human intervention are highly essential for these applications. These automated techniques are mostly computer based and they must satisfy the requirements of high accuracy and less convergence time requirement. Several automated techniques are available in the literature for brain image analysis. These techniques can be broadly categorized into intelligence based techniques and non-intelligence based techniques. Among these techniques, the intelligence based techniques are found to be efficient than the non-intelligence based techniques. Artificial Neural Networks (ANN) and the fuzzy logic techniques are the prime constituent of the intelligence based approaches. These techniques are effective in terms of accuracy and the convergence rate.