NEURD offers automated proofreading and feature extraction for connectomics

Area of Science:

• Neuroscience
• Computational Biology
• Bioinformatics

Background:

• Millimeter-scale electron microscopy volumes at nanometer resolution are now achievable.
• Machine learning advances enable dense reconstruction of cellular compartments.
• Automated segmentation requires post hoc proofreading for accurate connectomes.

Purpose of the Study:

• To develop a software package for decomposing meshed neurons into annotated graph representations.
• To automate the extraction of detailed morphological and connectivity information from 3D neuron meshes.
• To enhance accessibility of large-scale neural datasets for researchers.

Main Methods:

• Building on open-source mesh manipulation software.
• Developing the Neural Decomposition (NEURD) package.
• Decomposing meshed neurons into compact, annotated graph representations.

Main Results:

• NEURD generates feature-rich graphs from 3D neuron meshes.
• Automated proofreading of merge errors is enabled.
• Automated cell classification, spine detection, and axonal-dendritic proximity analysis are facilitated.

Conclusions:

• NEURD simplifies downstream analysis of neural morphology and connectivity.
• The software makes massive and complex neural datasets more accessible.
• This facilitates advanced research in neuroscience.