Effective feature extraction for Cerebral Microbleed detection using Edge Emphasized Weber Maximum Directional Co-occurance Matrix

CMBs are the accumulations in the brain vessels of many aged and stroke-affected persons. The presence of CMBs may lead to dementia, traumatic brain injury, and other physiological complications leading them to confusing behavior. Tracking of CMBs from the human brain is challenging due to their small size, manual detection by neurologists may lead to delusions. In this paper, we propose a newfangled technique for feature extraction in the detection of CMB, namely the Edge Emphasized Weber Maximum Directional Co-occurance Matrix (EEWMDCM). Our proposed methodology efficiently recognizes the CMB from magnetic resonance images by integrating the WLD with the notions of the Directional Co-occurance Matrix. When compared with the previous works that occupied many handpicked ROIs selections, more useful features can be extracted in our proposed work from the segmented images that progress the accuracy rate in the detection process. Noise removal filters are utilized and the segmentation is done to identify the candidate in the preprocessing stage which strengthens the feature extraction stage. The features that are extracted from the feature extraction stage are classified and experimented with a set of classifiers, the Support Vector Machine (SVM), Cosine Distance (CO) classifier, Chi-square (CS) classifier, Extreme Learning Machine (ELM), and Convolutional Neural Network (CNN) classifier. The proposed work has been verified and validated on an SWI-CMB dataset that was captured from 320 subjects, in which the images from 230 subjects are for training and images from 90 subjects are for testing purposes. The experiential results specify that our proposed work gives the best sensitivity of 97.11%, the precision of 97.31%, specificity of 97.24%, and an accuracy of 98.06%.