Automatically tracking neurons in a moving and deforming brain

Area of Science:

• Neuroscience
• Computational Biology
• Biophysics

Background:

• Optical neuroimaging enables cellular-resolution recordings in behaving animals.
• Tracking neuron locations over time is crucial for accurate activity extraction.
• Large brain motion and deformation in C. elegans present significant tracking challenges.

Purpose of the Study:

• To present an automated computer vision pipeline for tracking neuron populations in freely moving C. elegans.
• To address the challenges of large brain motion and deformation in neuroimaging data.
• To enable reliable single-neuron resolution tracking in dynamic biological systems.

Main Methods:

• Developed an automated pipeline involving straightening, alignment, and registration of 3D volumetric fluorescent images.
• Utilized neuron segmentation to identify individual neuronal locations.
• Introduced Neuron Registration Vector Encoding (NRVE), a machine learning approach for time-independent neuron identification using non-rigid point-set registration.
• Employed thin-plate spline interpolation for error correction and identity consistency checks.

Main Results:

• Successfully tracked populations of neurons with single-neuron resolution in freely moving C. elegans.
• The NRVE approach demonstrated suitability for tracking neurons in brains with large deformations.
• The pipeline located 156 neurons over an 8-minute recording.
• Achieved faster and more consistent neuron identification compared to manual or semi-automated methods.

Conclusions:

• The automated pipeline reliably tracks neurons in C. elegans brains undergoing significant motion and deformation.
• Neuron Registration Vector Encoding is an effective method for neuron identification in dynamic neuroimaging data.
• This approach enhances the analysis of whole-brain calcium imaging in freely moving small invertebrates.