Abstract
The images of brain indicate the conditions of the brain. Numerous computer-aided approach to enhance brain images exist in literature. As an attempt to further contribute to the efforts in this direction, we aimed at designing a software by using MATLAB that would automatically classify the images of brain into their associated disorders. To achieve the objective of this research, a database for training and testing the software of brain images were used. We used a total about 120 images of brain from the database that constituted images from normal patients as well as abnormal patients. Features were extracted from these images based on which they were differentiated and classified. The images under study were categorized into two classes namely, normal, and abnormal (based on its binary study) by using Support Vector machine (SVM) and K-nearest neighbors (KNN) algorithms. Both SVM & KNN are commonly used classifiers for binary study. The extracted features from these algorithms were used in the classification process. These classifiers yielded good results with the SVM classifier performing better (accuracy reaching 86.7%). We also used Python and its libraries to further improve the results and optimize the classification. To start with, a custom convoluted neural network was created and run with four hidden layers. By using the same dataset, over the time the model gave an accuracy of 95 %. Further trials are underway.