Genetic Algorithm and Deep Learning feature based Tumor Detection

Abstract

In medical diagnostics, brain tumour segmentation is a difficult process. The basic goal of brain tumour segmentation is to provide accurate characterizations of brain tumour regions by employing masks that are appropriately positioned. Deep learning approaches have shown a lot of promise in recent years for tackling issues like object recognition, picture classification, and semantic segmentation in computer vision. To obtain outstanding system performance in brain tumour segmentation, a variety of deep learning-based methods have been used. This article seeks to examine the recently created deep learning-based brain tumour segmentation technology in light of the most sophisticated technology and its performance. This work utilised a genetic algorithm based on fuzzy C-means (FCM-GA) to separate tumour areas from brain scans. During the preprocessing stage, the input picture is scaled to 256x256 pixels. A preprocessed MRI picture was segmented using FCM-GA. This is a sophisticated machine learning (ML) approach for finding items in huge datasets that is extremely flexible. To enhance the feature subset, the segmented picture is next subjected to hybrid feature extraction (HFE).Kernel Nearest Neighbor with a Genetic Algorithm (KNN-GA) is utilized in the feature selection process to achieve the optimum feature value. The RESNET classifier uses the best feature value to split the MRI picture into three regions: meningioma, glioma, and pituitary gland. The suggested hybrid method's performance is validated using real-time data sets. In comparison to current Convolutional Neural Networks (CNN) and Support Vector Machines (SVM) classification methods, the suggested technique enhances average classification accuracy by 7.99 percent.