Classification Of Spinal Disc Herniation Cases Using The Support Vector Algorithm (SVM) On Magnetic Resonance Imaging (MRI) Images

Abstract

Herniated disc disease is a common spinal disorder that results in compression of adjacent nerves, leading to severe pain and a range of neurological symptoms. Accurate diagnosis of this condition relies primarily on Magnetic Resonance Imaging (MRI), which is considered one of the most effective medical imaging modalities for providing high-resolution visualization of soft tissues, including intervertebral discs. This study aims to develop a Computer-Aided Diagnosis (CAD) system for the detection and classification of herniated discs using the MATLAB programming environment, with the objective of reducing human error and accelerating clinical decision-making. The proposed approach integrates the PCA-SIFT algorithm for precise segmentation and feature description, and the Gray-Level Co-occurrence Matrix (GLCM) for extracting statistical and textural features, followed by the application of a Support Vector Machine (SVM) algorithm for binary classification. The methodology is based on the analysis of 11 MRI images obtained from patients at Sebha Medical Center, from which 54 disc samples were extracted (34 normal discs and 20 herniated discs). The developed system demonstrated promising performance, achieving an overall classification accuracy of 70%. The results further indicate that extracted features such as kurtosis serve as significant biomarkers, showing a noticeable increase in herniated cases, thereby enhancing the system’s potential as a preliminary diagnostic support tool.