An Efficient Classification of MRI Brain Images

The unprecedented improvements in computing capabilities and the introduction of advanced techniques for analysis, interpretation, processing, visualization images have greatly diversified domain medical sciences resulted field imaging. Magnetic Resonance Imaging (MRI), an imaging technique, is capable producing high quality human body including brain diagnosis purposes. This paper proposes a simple but efficient solution classification MRI into normal, abnormal containing disorders injuries. It uses with tumor, acute stroke alzheimer, besides normal images, from public dataset developed by harvard school, evaluation proposed model four step process, which steps are named: 1). Pre-processing, 2). Features Extraction, 3). Reduction, 4). Classification. Median filter, being one best algorithms, used removal noise such as salt pepper, unwanted components scalp skull, pre-processing step. During this stage, converted gray scale to colored further processing. In second step, it Discrete Wavelet Transform (DWT) technique extract different features images. third Color Moments (CMs) reduce number get optimal set characteristics. Images passed classifiers Feed Forward - ANN (FF-ANN), individual classifier, was given 65% 35% split ratio training testing, hybrid called: Random Subspace Forest (RSwithRF) Bayesian Network (RSwithBN), 10-Fold cross validation 95.83%, 97.14% 95.71% accurate classification, corresponding order. These promising results show that method robust efficient, comparison with, existing methods terms accuracy smaller features.