Feature Extraction and Selection Techniques for Brain Tumor MRI Analysis Using Artificial Intelligence

Feature extraction and selection are the main elements which lead to the correct classification and diagnosis of brain tumors with the help of medical imaging techniques. The detailed method of getting the distinguishing features of the MRI brain images through the combination of statistical, texture, and deep learning-based methods is shown in the analysis of this chapter. At first, it is the tumor regions that are subjected to enhancement and segmentation, done in the preprocessing steps, which are then handled for feature computation via Gray Level Co-occurrence Matrix (GLCM), Histogram of Oriented Gradients (HOG), and Discrete Wavelet Transform (DWT). Then onward, feature reduction is done by applying Principal Component Analysis (PCA) and Genetic Algorithms (GA) techniques in order to take away non-informative features and lower dimensionality. The suggested method promotes the classification effectiveness as well as the diagnostic accuracy. Trials on MRI datasets with MATLAB 2017b exhibit an average accuracy of 96,8 % and F1-score of 0,94, thus validating that optimal feature extraction and selection practices are great supporters of AI-based medical image analysis frameworks.