Purpose: The scoliosis diagnosing system needs radio-graphic information in terms of spinal curvature estimated using Cobb's definition. The evaluation process and treatment analysis depends on the reliability and reproducibility of the spine curvature in the frontal view. Methods: Manual identification of end vertebrae and other anatomical features required for the estimation of spinal curvature causes variability and unreliability at higher rate. This paper proposes an automated system to extract the required anatomical features using customized filter. The customized filter used in this paper is a combination of anisotropic, sigmoid and differential filter. Combination of these filters automatically extracts the anatomical features in terms of required vertebral endplates. Automatic identification of these endplates eliminates the human intervention involved in the quantification of Cobb angle. Results and Conclusions: Analysis of the results reveals significant difference between the observer variability between manual, computer assisted and computerized image understanding system in terms of inter and intra cross correlation coefficient ratio (ICCR).
All Science Journal Classification (ASJC) codes
- Biomedical Engineering