A Hybrid Optimal Feature Extraction for Brain Tumor Segmentation

Abstract

The brain is the central nervous system of a human being. Brain tumor disease is considered the significant cause of death in many people. The core idea of deep learning is the comprehensive feature representations that will be learned efficiently along with the deep architectures, which are composed of trainable non-linear operations. Learning effective feature representations directly from the MRI becomes harder. Therefore, in the present study, a hybrid and optimal method are proposed. Grey Wolf Optimization algorithm is used for feature selection which reduces the more numbered features and then the classification of an image with the tumor type is done by the classifier Recurrent Neural Networks. The segmentation process is performed after the classification process, here segmentation is done by the MRG method with threshold optimization. The performance analysis is performed in terms of sensitivity, specificity, and accuracy which is done for the proposed techniques. Performance accuracy is obtained from this study is 98.16% using the proposed GWO technique.