Extracting spinal cord contours from transaxial MR images using computer vision techniques

Abstract

Computer vision techniques were applied to finding spinal cord contours in MR (magnetic resonance) images of the lumbar spine. For each slice, the method starts with an approximate spinal cord center supplied by the user. A search was then made for significant edges by moving outwards along 16 equally spaced radii. Edges were interpolated on radii for which no significant gradient existed. The contour was smoothed by adjusting the position of each edge with respect to the edges found on neighboring radii. The results of the experiments are reported. The research demonstrates that it is feasible to use a fairly simple image processing algorithm on a personal computer to automatically extract the spinel cord contours from transaxial MR images.<>