Machine Learning-Based System for Abnormality Detection and Classification in MRI with Advanced Techniques for Brain Tumor Identification

A brain tumor refers to an irregular proliferation of nerve cells in the human brain, disrupting its typical functioning. It has resulted in the loss of many lives. To rescue patients suffering from this condition, it is critical to discover it early and treat it properly. Identifying tumor-impacted cells in the brains of humans is a laborious and lengthy procedure. Nevertheless, the precision and time necessary for identifying tumors in the human brain provide a substantial hurdle in the arena of image investigations. This study designs and implements a brain tumor detection system by employing machine learning technique-Support Vector Machine and considering Magnetic Resonance Imaging brain images. The devised detection system serves multiple functionalities, which involves compilation of input data, pre-processing steps (encompassing k-means clustering for image segmentation endeavors), Histogram of Oriented Gradients (HOG) method-driven extraction of feature, SVM application, image classification, Convolutional Neural Network-based abnormality classification, Additionally, we propose treatment strategies for the three types of tumors: glioma, meningioma, and pituitary tumors. Our CNN-based framework for identifying brain abnormalities is currently being evaluated for its performance through accuracy metrics and confusion matrices, supplemented by simulation visuals. This thorough methodology aims to enhance the precision and dependability of brain tumor detection, while also offering valuable insights into customized treatment approaches tailored to the specific tumor types identified.