IMPROVED BRAIN MRI CLASSIFICATION USING COMPUTATIONAL INTELLIGENCE APPROACHES

In the history of medical imaging various computer-aided diagnostic systems have been proposed to assist medical professionals for identifying the fatal conditions of brain tumor while analyzing MRI scans. In this context, the author had extended their earlier brain MRI segmentation model to offer high-end brain tumor classification in the proposed work. The already proved improved segmentation procedure based on k-means optimized Firefly Algorithm (FFA) is involved for brain MRI segmentation to identify Region-of-Interest (RoI) highlighting the tumor regions. The feature extraction of segmented RoI image is performed using Speeded Up Robust Features (SURF) followed by implementation of FFA for extracting the best feature set in order to reduce the dimensionality of the feature data that prove to be effective for accurate tumor classification. A hybrid of Support Vector Machine (SVM) and Deep Neural Network (DNN) is used at the training and classification stage in which trained support vectors are used for classification by DNN architecture. The performance of the proposed brain tumor classification work is evaluated using 500 MRI images in terms of precision, recall, f-measure, accuracy and execution time. Simulation analysis demonstrates the attainment of average classification accuracy of 99.08% and average precision of 94.22% with reduced classification time of 1.11%. The work proved to be very advantageous for medical professionals and radiologists involved in analyzing brain tumor using MRI scans.