Modality Based Medical Image Classification

Abstract

This study is conducted to evaluate the performance of fuzzy systems and supervised classification techniques for the medical image modality classification problem. Due to the increase in medical image acquisition medical image classification is essential in computer-aided diagnosis. The contribution of research includes an introduction to the classification technique using a minimum number of intensity-based features (texture features) and the comparisons of the supervised and fuzzy approaches and obtaining the maximum classification rate with minimum risk. This paper describes SVM (linear and RBF kernel) and k-NN supervised techniques and a novel fuzzy rule-based system approach for medical image classification. The experiment was conducted on real images of five types of modalities (CT, MRI, X-Ray, Ultrasound, and Microscopic) and results show that every classification technique differs and shows different classification results for medical images.