Semantic Segmentation of Brain MRI Images Using Squirrel Search Algorithm-Based Deep Convolution Neural Network

Abstract

AbstractIn recent years, brain tumor has become a severe threat to human lives. These tumors are so often inadequately contrasted and are inadequately dispersed. In recent days, brain tumor is automatically detected using semantic segmentation. However, the variability in the size of brain tumors and the low contrast of brain imaging are the two major problems affecting the performance of semantic segmentation. To address this problem, a squirrel search algorithm-based deep convolution neural network (SSA-DCNN) proposed for semantic segmentation of the medical images in this paper. The proposed method is a blend of deep convolution neural network (DCNN) and squirrel search algorithm (SSA). The SSA is used to fine-tune the performance of DCNN by optimizing the hyperparameters of the DCNN, which in turn enhances the accuracy of the semantic segmentation. The proposed method is implemented and validated by performance metrics such as accuracy, loss, IoU, and BF score. The performance of SSA-DCNN is compared with the jellyfish algorithm-based deep convolution neural network (JA-DCNN), oppositional-based seagull optimization algorithm (OSOA-3DCNN), and particle swarm optimization (PSO)-DCNN.KeywordsSquirrel searchDeep convolution networkBrain tumorSemantic segmentation