Abstract

Understanding how cells form tissues is an essential component in systems biology that involves the generation of tissue models. Generating such a tissue model requires a proper reconstruction of the different cells forming the tissue visualized as fluorescent objects in microscopy images. This is limited by the number of fluorescent markers that can be simultaneously imaged in a tissue sample (up to 4-5 by confocal microscopy). This limitation can be overcome by using automatic algorithms for the recognition of the different cell types without the use of specific markers. In this study, we propose a toolbox of algorithms for an accurate identification, reconstruction and characterization of different cells types in 3D tissue images. We applied our toolbox to the recognition of sinusoidal endothelial cells (SECs), Kupffer and Stellate cells in adult mouse liver tissue. The cell recognition algorithm was based on the morphology, texture and relative localization of the nuclei. The analysis of the most relevant parameters used for cell classification gave new insights into liver cell structure and function. In particular, nuclear shape, distance to cell borders, chromatin texture and proximity to sinusoids were the most important parameters for the non-paraenchymal liver cells classification.