Enhancing MRI Brain Tumor Classification with a Novel Hybrid PCA+RST Feature Selection Approach: Methodology and Comparative Analysis

Being a critical health challenge, understanding how to classify these brain tumours by using MRI will help the patient get proper treatment. In this work, a new hybrid model is proposed by integrating both Principal Component Analysis (PCA) and Rough Set Theory (RST) approaches for brain tumor classification from MRI images. Our strategy of PCA and RST aims at data dimensionality reduction and feature selection function for tumor classification fine. The hybrid method was also validated with ADNT and OASIS MRI datasets, demonstrating its effectiveness. The first step of image processing was segmentation to delimitate regions of interest, which were subsequently used for feature extraction by using the Discrete Wavelet Transform (DWT). Subsequently, PCA+RST simultaneous feature selection and reduced set thresholding algorithm was performed on these selected features to fine tune them for recognition. The comparison was performed using the following four classifiers: J48, Support Vector Machine (SVM), K-nearest neighbors(KNN) and Naive Bayes. The proposed hybrid PCA+RST approach was compared with the traditional approaches like DWT+SVM, and DWT+PCA+KNN in terms of their performance. The performance of the hybrid approach was found to be superior compared to these traditional methods in terms of classification accuracy. In this paper, a hybrid method of PCA and RST is proposed to improve the detection accuracy as well as classification performance in brain tumor diagnosis using MR images. Hopefully, this new tool will result in a visualization technique for more precise medical image analysis and timely diagnostics.