Methods for Evaluating Gray Matter Contour Similarity in Transverse Slices of the Mammalian Spinal Cord

The spinal cord may be divided into segments. The neural networks of different segment groups control, in particular, locomotion and visceral functions. Spinal cord segments serve as critical topographic landmarks for both experimental and therapeutic interventions. However, accurate identification of segment positions in vivo, particularly through automated methods, remains challenging: in mammals, some spinal cord segments are displaced rostrally (ascend) relative to their corresponding vertebrae, and the extent of this displacement varies even within a single species. One solution to this problem may be the use of reference images of slices of segments taken, for example, from histological atlases of the spinal cord. In this paper, we investigate various methods for analyzing the similarity of gray matter contours in transverse slices of the mammalian spinal cord, which allow us to determine whether a slice belongs to a certain segment. We consider the methods for analyzing slices obtained from one animal (based on the Jaccard coefficient, the metric of distances between contours, correlation analysis of R-φ curves, or Hu invariant moments), as well as the methods for comparing images of spinal cord segments with reference ones (correlation analysis of R-φ curves, Hu invariant moments). The results obtained allow us to assume that the method of determining segments by comparing tomographic or histological images of spinal cord transverse slices at different levels with a certain database containing a set of reference images of slices of specific segments, based on Hu invariant moments, is the most effective of those considered.