Investigation on glioma detection and grading schemes by deep learning 202techniques

Abstract

The term quotcerebral gliomaquot refers to the form of primary brain tumor newlinethat occurs the most frequently. According to the degree of aggressiveness newlinethat these tumors display, the majority of the time, doctors classify them as newlineeither low grades or high grades. Patients who are found to have high-grade newlinegliomas typically have a survival time of less than eighteen months after the newlinediscovery of their tumors. This is because these tumors are highly malignant newlineand have a dismal prognosis. Low-grade gliomas progress more slowly than newlinehigher-grade gliomas, have a lower risk of being cancerous, and are more newlinelikely to respond favorably to treatment. newlineThe process of histological grading is the method that is currently newlineregarded as the technique that represents the gold standard for diagnosis, the newlineplanning of treatment, and the prediction of survival time. The primary newlineobjective of this thesis is to suggest innovative methods for the automatic newlinedetection, classification, and grading of gliomas utilizing conventional newlineMagnetic Resonance Imaging (MRI) modalities. To achieve the research goal, newlinethis work proposes four different research solutions, which are meant to detect newlineand grade gliomas. The initial research work detects brain tumors by newlineemploying hyperparameter-optimized Convolutional Neural Networks newline(CNN). The hyperparameter-like optimizers, momentum, and batch size were newlineall taken into consideration and compared with one another to determine newlinewhich one would produce the best results in terms of precision. It can be newlinededuced from the observations that to enhance the functionality of the system, newlinethe grid search tuning strategy was utilized to optimal hyperparameter settings newlinefor the dataset. newline