Efficient Brain Tumor Classification With Optimized Hybrid Deep Neural Networks

Abstract

Segmentation is an important stage in the processing of images. Following pre-processing, segmentation methods are used to isolate the tumor region from the MRI images. It's one of the most crucial CAD procedures from the perspective of medical imaging. The challenges in segmenting the tumor area is overcome by using the semantic segmentation method, in which each pixel in an image receives a name or classification. It is used to recognize collections of pixels that stand in for different categories. Semantic Segmentation is proposed which is used to separate the tumor region and then the deep learning classification is done using Augmented Radial Basis Function Network (ARBFNs) based deep learning, Long Short Term Based Recurrent Neural Network (LSTM-RNN) methodology and Regularized Convolutional Neural Network with Dimensionally Reduction Module (RCNN-DRM) architecture. The proposed algorithm providing 95% accuracy on training data.