Register to receive the Zoom link:


Keywords: image segmentation, large-scale analyses, honey bee brain, deep learning


The analysis of volumetric medical and biological imaging data often requires isolating individual structures from the 3D volume by segmentation. In insects, analysis of large numbers of samples can reveal minor but statistically and biologically relevant variations in brain morphology and lateralization addressing major issues related to behaviour, ecology and evolution. However, the manual effort of conventional methods (e.g. histology, scanning electron or confocal laser scanning microscopy in combination with manual segmentation) limits the number of samples required for a large-scale analysis. Here we use micro-CT combined with automated 3D reconstruction to analyse the neuro-architecture of 110 honey bees. The reconstruction is achieved with Biomedisa, an online segmentation platform that utilizes deep neural networks for automated image segmentation. We analyse the inter-individual variability of brain morphologies and lateralization in honey bees and describe architectural asymmetries. In particular, we found a subsequent lateralization of antennal lobes and lobulae that may explain behavioral lateralizations previously reported for olfactory and visual learning.

Overall, Biomedisa enables easy access to large-scale quantitative comparative analyses. The platform is accessible via a web browser and does not require complex and tedious configuration of software and model parameters, thus addressing the needs of scientists without substantial computational expertise.


Short Biography of Dr. Philipp Lösel


Selected Publications