Abstract:
Brain tumors can develop at any location of the brain with uneven boundaries and shapes. Typically, they were increasing rapidly due to which its size approximately doubles just in twenty-five days. If they were unrecognized in earlier phases, patients suffered from various medical problems, including death. So, the identification of brain tumors in the earlier stages is one of the critical aspects. In addition to that, an effective imaging sequence also plays a vital role in tumor diagnosis. Magnetic resonance (MR) imaging is widely used among the available scanning approaches. Therefore, this article develops a new methodology to classify MR-based brain images. The proposed methodology includes pre-processing, segmentation, feature extraction, and classification. In pre-processing, we enhance the brain MR images using a median filter and obtain the region-of-interest (ROI) by thresholding and morphological operations. Next, in feature extraction, we extracted relevant local textures and shaped informative features from ROI using Enhanced Gradient Local Binary Patterns (EGLBPs) and Modified Pyramid Histogram of Oriented Gradients (MPHOG). Finally, we perform classification by various supervised learning approaches, namely Support vector machine (SVM), K-nearest neighbors (KNN), and Ensemble learning. All these experiments are implemented on Harvard Medical School (HMS) database. From the simulation results, we identified that the implemented imaging system attained good performance on classification and segmentation tasks compared to the existing techniques. Hence, we conclude that our suggested framework can be utilized as a predictive tool during diagnosing patients who suffer from brain tumors.