Brain Tumor Detection using Transfer Learning Technique with AlexNet and CNN

The segmentation, identification, and extraction of malignant tumor areas from magnetic resonance (MR) images is a major issue, yet it’s a monotonous and longterm operation that radiologists or clinical specialists must undertake, and their accuracy is entirely reliant on their experience. As a result, the deployment of computer-assisted technology becomes more important to overcome these obstacles. Numerous researchers have proposed several approaches for accurate brain tumor identification and segmentation. A comparison of several procedures has been conducted. After a detailed analysis, Image Rejuvenate and Intensification are revealed to be two major issues which are resolved and explained using appropriate methodologies. In this paper, Convolutional Neural Network (CNN) and AlexNet have been implemented to increase the performance and minimize the complexity involved in the magnetic resonance image segmentation process. Secondly, important features are retrieved from segmented tissues to increase the accuracy and performance efficiency using support vector machine (SVM) -histogram of oriented gradient (HOG). Based on accuracy, precision matrix, and confusion matrix, the experimental outcome of the proposed methodology have been verified and evaluated for quality assurance and efficiency on MR images. The research results proved to be 97% accurate indicating the hypothesized methods for detecting healthy and cancerous tissues