Multiple Sclerosis Identification Based on Ensemble Machine Learning Technique

The diagnosis of multiple sclerosis disease (MSD) is crucial because it is a neurological disease leading to communication failure between brain tissues and other parts of the body. Effective classification and segmentation of brain tissues are necessary for early detection of multiple sclerosis disease. In this proposed work, an ensemble learning-based classification technique is proposed to identify multiple sclerosis diseases from a database of healthy and unhealthy brain magnetic resonance (MR) images. Feature extraction from brain MR images is performed using an eighteen different Gray Level Co-occurrence Matrix (GLCoM) based features. Then, decision tree-based ensemble learning is accomplished on these features using three different boosting techniques for classification of healthy brain MR image from a weak brain MR image. Performance metrics like sensitivity ( PR T ), specificity ( NRT ), accuracy, precision (PPV), and F-score are utilized for MSD identification. It has been verified that the ensemble learning technique yielded higher accuracy of 94.91% from other states of the art techniques on the e-health dataset.