Skeletonization of neuronal processes using Discrete Morse techniques from computational topology.

Area of Science:

• Computational neuroanatomy
• Neuroscience
• Computational topology

Background:

• Mapping vertebrate brain neuronal networks is crucial for understanding biological intelligence.
• Current methods using tracer injections quantify regional projections by label intensity, which lacks biological meaning.
• Difficulty in tracing individual axons within dense tracer labeling hinders detailed circuit analysis.

Purpose of the Study:

• To develop a novel computational method for more biologically meaningful quantification of neural projections from tracer data.
• To improve the mapping of mesoscale neural circuitry in vertebrate brains.
• To bridge the gap between single-axon tracing and bulk tracer injection data.

Main Methods:

• Skeletonization of labeled axon fragments.
• Estimation of volumetric length density using deep nets and Discrete Morse (DM) technique.
• Application of computational topology for noise-robustness and nonlocal connectivity analysis.

Main Results:

• Demonstrated utility and scalability on whole-brain tracer injected data.
• Developed an information-theoretic measure to quantify additional information from individual axon morphologies.
• The DM technique provides noise-robustness by considering nonlocal connectivity.

Conclusions:

• The proposed approach offers a more biologically meaningful quantification of neural projections compared to traditional methods.
• This is the first application of the DM technique in computational neuroanatomy.
• The method enhances the mapping of neural networks by integrating tracer injection and single-axon data.