Semi-automatic spine extraction for disc space narrowing diagnosis

This paper describes the development of a semi-automatic system for detection and diagnosis of vertebrae condition, focuses on cervical area. The goal of this system is to facilitate medical community to make a faster pre-screening based on the imaging modalities, especially X-ray image. The main challenges in diagnosing a disease through X-ray image are the issue of blur and noise. Therefore, to achieve this goal, a semi-automatic spine extraction to detect disc space narrowing (DSN) condition has been developed that focused on patient with back pain history. In general, this system was developed on Matlab platform that consists of four major modules, which are image enhancement, image segmentation, feature extraction and classification. Image enhancement module utilized Contrast-limited Adaptive Histogram Equalization (CLAHE) and filtering technique to improve the image quality. After that, the second module is performed to extract the desired region from the original X-ray image. Feature extraction module is then implemented to extract unique signature of the vertebrae bones based on the bone's condition. For the last module, feed-forward backpropagation artificial neural network is used to classify the existence of DSN. It needs to be trained before testing is performed so that the parameters can be tuned for optimal classification. The quantitative performance proved that the X-ray image quality has been improved and the system has managed to classify the DSN condition. Simulation results show that the proposed system provides good performance of accuracy with average of 99% for the tested X-ray images. As for future work, the system can be further improved by using more measurement points between the two neighboring vertebras.