RealNeuralNetworks.jl

RealNeuralNetworks.jl analyzes large-scale 3D neural segmentations derived from electron microscopy to generate neuron skeletons, quantify neuronal morphology, and analyze synaptic connectivity at terabyte scale.

Key Features:

• Sparse Skeletonization: Efficient transformation of voxel-based neuron segmentations into sparse skeletons for quantitative morphological analysis.
• Morphological Analysis: Computation of detailed neuronal morphology metrics from skeletonized reconstructions.
• Synaptic Connectivity Analysis: Tools to analyze synaptic connectivity and network relationships within EM-derived datasets.
• Distributed Computing (Google Cloud): Support for distributed processing of terabyte-scale datasets using cloud environments such as Google Cloud.
• Clustering with NBLAST: Integration of the NBLAST algorithm to cluster neurons by morphological similarity and synaptic connectivity.

Scientific Applications:

• Large-scale electron microscopy dataset analysis: Processing and analysis of terabyte-scale 3D EM segmentations, including datasets from Zebrafish.
• Quantitative neuron morphology: Measurement and comparison of neuronal shape and structural metrics across large populations.
• Synaptic circuit mapping: Characterization of synaptic connectivity patterns and network structure from EM-derived segmentations.
• Structure–function investigations: Exploring relationships between neuronal morphology and inferred functional roles.

Methodology:

Implemented in Julia, the package performs sparse skeletonization of voxel-based neuron segmentations, computes morphological metrics on skeletons, analyzes synaptic connectivity, supports distributed processing on Google Cloud for terabyte-scale datasets, and applies the NBLAST algorithm for neuronal clustering.