A Novel Technique for Brain Tumor Detection and Classification using T1-Weighted MR Image

Abstract

Brain tumors are particularly perilous because they form when cells in the brain multiply uncontrollably within the skull. Therefore, a fast and accurate method of diagnosing tumors is crucial for the patient’s health. This study proposes a method for evaluating brain cancer images. The phases of implementation for the proposed work are as follows: In the first phase, we compiled a set of specialized feature vector descriptions for advanced classification tasks by employing both deep learning (DL) and conventional feature extraction techniques. In the second phase, we employ a proposed convolutional neural network (CNN) approach and a traditional subset of features from a genetic algorithm (GA) to select our deep features. The third phase involves using the fusion method to merge the prioritized features. Finally, determine whether the brain image is normal or abnormal. The results showed that the proposed method successfully classified objects accurately and revealed their robustness across different ages and acquisition protocols. According to the results, the classification accuracy of the support vector machines (SVM) classifier has significantly improved by combining conventional features and deep learning features (DLF), achieving an accuracy of up to 86.50% using the T1 weighted brain MR image.