Hessian-based neck tracing of dendritic spines: a preliminary study on confocal images

Alterations in dendritic spines morphology and topology are known to be implied in several neurodegenerative diseases. However, visualizing and reconstructing the exact morphology of spine necks using conventional confocal imaging is still a challenge. In the light of this, we propose the extension of Hessian matrix eigenvalue analysis for the detection of tubular structures to the problem of 3D neck tracing. In this work we stress the relevance of the tracing procedure initialization, for the novel application of this standard approach to the specific scenario. Preliminary results on confocal microscopy images from ex-vivo human samples allowed to obtain almost 80% of necks keeping an average distance of less than one voxel from the reference segmentation and half of them fully overlapping with it. The preliminary results are promising, since they pave the way towards the development of this model-free solution to accurately trace spine neck.