Paper ID: 2207.13021

Topological Optimized Convolutional Visual Recurrent Network for Brain Tumor Segmentation and Classification

Dhananjay Joshi, Kapil Kumar Nagwanshi, Nitin S. Choubey, Naveen Singh Rajput

In today's world of health care, brain tumor (BT) detection has become a common occurrence. However, the manual BT classification approach is time-consuming and only available at a few diagnostic centres. So Deep Convolutional Neural Network (DCNN) is introduced in the medical field for making accurate diagnoses and aiding in the patient's treatment before surgery. But these networks have problems such as overfitting and being unable to extract necessary features for classification. To overcome these problems, we developed the TDA-IPH and Convolutional Transfer learning and Visual Recurrent learning with Elephant Herding Optimization hyper-parameter tuning (CTVR-EHO) models for BT segmentation and classification. Initially, the Topological Data Analysis based Improved Persistent Homology (TDA-IPH) is designed to segment the BT image. Then, from the segmented image, features are extracted simultaneously using TL via the AlexNet model and Bidirectional Visual Long Short Term Memory (Bi-VLSTM). Elephant Herding Optimization (EHO) is used to tune the hyper parameters of both networks to get an optimal result. Finally, extracted features are concatenated and classified using the softmax activation layer. The simulation result of this proposed CTVR-EHO and TDA-IPH method is analysed based on some metrics such as precision, accuracy, recall, loss, and F score. When compared to other existing BT segmentation and classification models, the proposed CTVR-EHO and TDA-IPH approaches show high accuracy (99.8%), high recall (99.23%), high precision (99.67%), and high F score (99.59%).

Submitted: Jun 6, 2022