Self-attention based progressive generative adversarial network optimized with momentum search optimization algorithm for classification of brain tumor on MRI image

Abstract

This manuscript proposes a self-attention based progressive generative adversarial network optimized with momentum search optimization algorithm for brain tumor classification on MRI image (SPGAN-MSOA-CBT-MRI). Initially, the data are gathered through the Brats 2019 dataset. Afterward, the data are given to pre-processing. In pre-processing, it reduces the noise and maximizes the superiority of input image utilizing anisotropic diffusion Kuwahara filtering (ADKF). Then the pre-processing output is fed to Feature extraction segment. Here, six texture features, like homogeneity (Angular Second Moment), contrast, Inverse difference moment, entropy, correlation and variance are extracted based on ternary pattern and discrete wavelet transform (TP-DWT). The extracted features are given to SPGAN for effectively classify the brain tumor on MRI image. The proposed SPGAN-MSOA-CBT-MRI approach is implemented in MATLAB. The performance metrics is evaluated to check the robustness of the proposed technique. The performance of proposed SPGAN-MSOA-CBT-MRI approach attains 6.45%, 9.45%, and 11.67% high accuracy;7.23%, 10.34%, and 12.56% high F-score compared with existing methods, such as GAM-SpCaNet: Gradient awareness minimization-based spinal convolution attention network for brain tumor classification (GAM-SpCaNet-CBT-MRI), Efficient and low complex architecture for detection and classification of Brain Tumor using RCNN with Two Channel (CNN RCNN-CBT-MRI) and Smart brain tumor detection scheme with the help of deep convolutional neural networks (DCNN-CBT-MRI) respectively.