Natural variability in bee brain size and symmetry revealed by micro-CT imaging and deep learning

Abstract

Analysing large numbers of brain samples can reveal minor, but statistically and biologically relevant variations in brain morphology that provide critical insights into animal behaviour, ecology and evolution. So far, however, such analyses have required extensive manual effort, which considerably limits the scope for comparative research. Here we used micro-CT imaging and deep learning to perform automated analyses of 3D image data from 187 honey bee and bumblebee brains. We revealed strong inter-individual variations in total brain size that are consistent across colonies and species, and may underpin behavioural variability central to complex social organisations. In addition, the bumblebee dataset showed a significant level of lateralization in optic and antennal lobes, providing a potential explanation for reported variations in visual and olfactory learning. Our fast, robust and user-friendly approach holds considerable promises for carrying out large-scale quantitative neuroanatomical comparisons across a wider range of animals. Ultimately, this will help address fundamental unresolved questions related to the evolution of animal brains and cognition.

Automated 3D cytoplasm segmentation in soft X-ray tomography

“…Previous work on U-Net segmentation showed that the number and quality of training datasets are crucial to achieving high accuracy of segmentation. 24 , 25 Therefore, we have measured the accuracy of the ACSeg by computing Dice coefficients 26 , 27 for U-Net trained on various SXT tomograms. We trained the ACSeg on 5, 10, 20, 30, and 43 datasets, see Figure 2 .…”

130 years from discovery to description: micro‐CT scanning applied to construct the integrative taxonomy of a forgotten moth from Southern Africa (Lepidoptera: Geometridae)

“…Third, the rendering software and protocols should further be developed particularly to shorten the arduous and still largely manual segmentation of target structures. Rather promising attempts have lately been made to cut the manual segmentation time from several hours to a few minutes using a deep learning-based fully automated segmentation of micro-CT images from ant (Toulkeridou et al, 2023) and bee (Lösel et al, 2023) brains. To take the automation a step further, an entire automatic rendering process from the scan to a standardised set of images should be available at least for certain diagnostic structures mentioned above.…”

How many specimens make a sufficient training set for automated 3D feature extraction?

“…To train our models in Biomedisa, we used a stochastic gradient descent with a learning rate of 0.01, a decay of 1 × 10 -6 , momentum of 0.9, and Nesterov momentum enabled. A stride size of 32 and a batch size of 24 samples per epoch were used alongside an automated cropping feature, which has been demonstrated to enhance accuracy [18]. The training of each network was performed on a Tesla V100S-PCIE-32GB graphics card with 30989 MB of available memory.…”