RealNeuralNetworks.jl: An Integrated Julia Package for Skeletonization, Morphological Analysis, and Synaptic...

Area of Science:

• Neuroscience
• Computational Biology
• Bioinformatics

Background:

• Electron microscopy and deep learning generate large brain image datasets.
• Current automated segmentation methods require a separate skeletonization step for analysis.
• Existing tools for skeletonization and connectivity analysis are fragmented across different platforms.

Purpose of the Study:

• Introduce RealNeuralNetworks.jl, a Julia package for efficient neural network analysis.
• Provide tools for sparse skeletonization, morphological analysis, and synaptic connectivity.
• Demonstrate applicability to terabyte-scale electron microscopy datasets.

Main Methods:

• Developed a Julia package, RealNeuralNetworks.jl.
• Utilized distributed computing on Google Cloud for large-scale skeletonization.
• Employed the NBLAST algorithm for neuron clustering and connectivity analysis.

Main Results:

• Successfully performed efficient sparse skeletonization on large-scale Zebrafish brain data.
• Integrated morphological similarity and synaptic connectivity analysis.
• Demonstrated the package's capability to handle terabyte-scale datasets.

Conclusions:

• RealNeuralNetworks.jl offers an efficient solution for neural network skeletonization and analysis.
• The package facilitates comprehensive morphological and synaptic connectivity studies.
• It is well-suited for analyzing massive electron microscopy datasets in neuroscience research.