Anirejuoritse Egbe
Title: Automated neuron segmentation and connectivity analysis for investigating connectome-constrained subneuron computation
Abstract: In neuroscience and deep learning, the fundamental computational unit in the brain is often analyzed at the level of the individual neuron. However, the firing of the neuron is determined in part by its geometry and how current flows through individual branches. Recent synapse-level connectome data holds promise for investigating the principles of subneuron computation in more detail at the level of an entire neural circuit. Simulating the activation of each neuron using an entire skeleton with tens of thousands of segments each is computationally burdensome at the network level and yields challenges for interpretability of information processing at the subneuron level. We have developed methodologies to automate segmenting neuron skeletons into several functional compartments to simulate a network at the subneuron level, while still being computationally efficient. We demonstrate the neuron segmentation and subneuron connectivity analysis in the Navigation center of the fruit fly, Drosophila melanogaster.