Looking at these examples one might say that in many cases the outline of the clusters is “counterintuitive” and contradicts our visual impression. This might be true; however the main reason for this discrepancy is the “preferred” orientation of visible neuronal clusters in the direction of the cortical column. So, to extract clusters outline which would better correspond to our perception we have to create distance map using non-isotropic structuring element elongated in the direction of the orientation of the cortical column. Technically it is possible, because it is easy to construct non-isotropic structuring element and use it for distance map construction. Unfortunately, this approach will give good results only in the regions where the section plane coincides with the orientation of cortical columns. In all other regions, where the orientation of the section is oblique, it will be difficult or even impossible to detect the required direction.
The ability to select the direction of sampling according to the orientation of cortical column in any given place would exist in isotropic 3D dataset. In isotropic 3D space we could detect this orientation and change the “preferred” direction of the distance map accordingly. Unfortunately, registered 3D section data are still not available at 1 mk resolution, and we still have to deal with problems and inconsistencies of processing 2D sections. So, for the purpose of analysis of clusters.through the whole 1 micron resolution section we have to use isotropic distance measure. Of course, this is a “work in progress”, and the situation might change in the future.
Fig. 4. “Clusterization signature” of the cortical area. Distances are color-coded.
The plot of the family of distributions of the number of clusters for each interval of cluster sizes and distances is presented in Fig. 4. This family of distributions can be called “clusterization signature” of the cortical area. The result of comparison of these signatures for different areas will follow shortly.
Copyright © 2017-2018, Brain Imaging Informatics, LLC