Enhancing Brain MRI Classification Through a Hybrid Machine Learning Methodology

Brain MRI classification is one of the most important problems nowadays. Many methods have been developed to tackle this problem, some of them provide good results but are computationally complex and some are simple, but the accuracy is very low. In this paper, we also proposed a novel method for brain MRI classification. The proposed method consists of three stages, namely preprocessing, feature extraction, and classification. In preprocessing stage, the median filter and transforms have been used to remove salt-and-pepper noise from brain MRI images and enhance the quality of images respectively. The grayscale images are converted to RGB images in this stage. In the feature extraction stage, the three channels namely red, green, and blue are extracted from RGB images. Nine statistical features have been calculated for each channel of RGB images. The extracted features are then stored in three different files and labeled. This stage provided the final dataset in the form of three files for red, green, and blue channels to be used in the classification stage. Different machine learning models are used in the classification stage, namely artificial neural networks, decision trees, naïve Bayes, k-NN, random forests, and support vector machines. These algorithms are applied separately on red channel features, green channel features, and blue channel features. The output classification results are then fed to the majority voting strategy where the images are classified as normal or abnormal. The performance of different machine learning algorithms is measured to find the best model for the extracted features. The decision tree accuracy which is 97.04% is higher than other algorithms. The accuracy of the proposed method is compared with some well-known methods for brain MRI classification, and the result indicates that the proposed method performs better as compared to other methods.