Fusion of CT and MRI modalities for brain tumors classification using enhanced machine vision framework

This study focuses on a data fusion approach for classifying brain tumors utilizing an enhanced machine vision (MV) framework. The foundation of a dataset is based on the integration of CT and MRI. We utilized the proposed hybrid segmentation approach to extract the region of interest. The hybrid feature dataset was extracted from the segmented regions and optimized via a correlation-based approach for further analysis. MV-based six classifiers were deployed: weightless neural network (WNN), averaged dependence estimator (ADE), rough set, ForEx++, CS Forest, and Multilayer Perceptron (MLP), using a 10-fold validation method. The CT-scan-based experiments observed that the MLP gives the highest (97.80%) accuracy. Similarly, the MRI-based experiments observed that the ADE performs well compared to other implemented classifiers and provides 98.13% accuracy. Lastly, the fused optimized hybrid feature dataset was utilized for experiments. Among all deployed classifiers, WNN showed a promising higher accuracy of 99.66%, respectively.