Semi-automated dendrogram generation for neural shape analysis

Abstract

This work introduces a new framework for shape characterization of neural cells by semi-automated generation of dendrograms, which are abstract structures describing different aspects associated to the branching structure of dendrites in neurons. Particularly, the dendrograms generated by the proposed method includes the length of each dendritic segment. The new method, which is based on multiscale curvature contour segmentation and syntactic shape analysis, is robust and greatly improves the tiresome task of dendrogram extraction, traditionally performed by human operators. Results corroborating the effectiveness of the technique are also presented.