Brain cell finder is a tool for fully automated localization of soma in 3D mouse brain images acquired by confocal light sheet microscopy. Given a 3D image as shown on the left, it first applies semantic deconvolution to enhance somata and reduce dendritic trees and axon bundles (center image). It then applies mean shift to locate soma centers (right image: red/orange are false positives, magenta false negatives). Furthermore, in the case of Purkinje soma in cerebellum images, it estimates manifold distances to filter false positives (orange points).
Images for this research were produced by the Biophisics lab of LENS. The software was developed by the AI Lab of DINFO, Università di Firenze in collaboration with Università Campus Biomedico di Roma.