Abstract

Brain tumors are among the most lethal and complex neurological disorders, often requiring precise diagnosis for effective treatment. Magnetic Resonance Imaging (MRI) plays a crucial role in detecting and classifying brain tumors due to its high-resolution imaging and contrast capabilities. However, accurate classification remains challenging due to overlapping tissue intensities, noise, and irrelevant features. Existing classification techniques often suffer from low precision due to redundant or non-discriminative features, especially when working with unsupervised clustering methods like K-Means. These shortcomings can result in misclassification, delayed treatment, and poor prognosis. This work proposes an improved method for MRI brain tumor classification by integrating K-Means clustering with an enhanced feature selection mechanism. Initially, preprocessing techniques such as grayscale conversion and histogram equalization are applied. K-Means clustering is used to segment the tumor region, followed by extraction of statistical and texture-based features (e.g., mean, entropy, contrast, GLCM). An enhanced feature selection approach based on Mutual Information and Principal Component Analysis (PCA) is used to reduce dimensionality and retain only the most relevant features. A classifier such as Support Vector Machine (SVM) is finally used for tumor type prediction. The proposed method was evaluated on a standard MRI brain tumor dataset. Experimental results showed improved classification accuracy (95.3%), precision (94.6%), sensitivity (95.1%), and F1-score (94.8%) compared to existing techniques such as basic K-Means and PCA-SVM.

Authors

K. Nagalakshmi, J. Rajalakshmi
Sethu Institute of Technology, India

Keywords

MRI, Brain Tumor, K-Means Clustering, Feature Selection, Classification