ViterBrain

ViterBrain reconstructs neuronal morphology from cleared, high-resolution brain imaging datasets by probabilistically tracing axonal and dendritic paths to support brain-wide atlas construction.

Key Features:

• Probabilistic reconstruction: Employs a hidden Markov state process to encode geometric neuron properties combined with a random field appearance model that captures neuron fluorescence.
• Path optimization: Uses dynamic programming to compute the most probable neuron path, yielding the global maximizer of path likelihood across images.
• Robustness to imperfect data: Designed to handle imperfect image segmentations and noise, enabling reliable following of axonal paths despite interference from nearby neurons.

Scientific Applications:

• Brain-wide atlas construction: Enables automated reconstruction of neuronal structures across entire mammalian brains at sub-micron resolution for atlas generation.
• Axonal and dendritic tracing: Facilitates tracing of axonal and dendritic paths in high-density neural networks subject to imaging noise and close-by neurites.
• Analysis of cleared, high-resolution imaging data: Applicable to datasets produced by brain clearing and high-resolution imaging technologies for large-scale morphology assembly.

Methodology:

Integrates geometric and appearance models in a probabilistic framework using a hidden Markov state process and a random field appearance model, and computes the global maximizer of the most probable neuron path via dynamic programming.