An Optimal Feature Extraction Technique for Glioma Tumor Detection from Brain MRI

A brain tumor is defined by the uncontrolled proliferation of brain tissue cells, defying the typical cellular regulation mechanisms governing growth. The most significant challenge associated with brain tumors lies in their timely diagnosis and accurate stage determination. Accurate detection of tumors from MRI scans can not only assist doctors in their examination but also provide crucial information for appropriate and timely treatment decisions. In this paper, a comprehensive analysis is presented based oncomparisons between state-of-the-art dimensionality reduction and classification algorithms. We useda dataset containing brain MRI scans, including both tumor and non-tumor cases, which wassplit into training and testing sets. After preprocessing the data, we implementedfour feature extraction algorithms to obtain different sets of features. Consequently, these sets of features were used to train five classifiers to analyze the accuracy. Based on these results, optimal feature extraction and brain tumor classification technique is selected. The results indicate that the Linear Discriminant Analysis (LDA) technique extracted highly informative features, leading to an impressive accuracy of 92.84%. This highlights the effectiveness of LDA in significantly enhancing the performance of the brain tumor classification process, making it the prime choice for feature extraction that aligns seamlessly with the research's intuition. It has higher accuracy with all the classifiers.