NeuRegenerate: A Framework for Visualizing Neurodegeneration.

Area of Science:

• Neuroscience
• Computational Biology
• Microscopy

Background:

• High-resolution microscopy advances brain connectivity understanding.
• Studying neural projection changes over time is limited by single-timepoint imaging and population analysis.

Purpose of the Study:

• Introduce NeuRegenerate, an end-to-end framework for predicting and visualizing neural fiber morphology changes across age-timepoints within subjects.
• Enable detailed longitudinal studies of brain connectivity.

Main Methods:

• Developed neuReGANerator, a cycle-consistent generative adversarial network (GAN) for translating neuronal structures across age-timepoints in large microscopy volumes.
• Implemented a density multiplier and hallucination loss to improve reconstruction quality.
• Introduced a spatial-consistency module to mitigate tiling artifacts.
• Integrated neuroCompare and neuroMorph visualization tools for analyzing projection differences and transformations.

Main Results:

• Successfully predicted and visualized changes in neural fiber morphology.
• Demonstrated the framework's capability on the mouse brain's cholinergic system, comparing young and old specimens.
• Showcased improved reconstruction quality and artifact reduction.

Conclusions:

• NeuRegenerate provides a robust framework for analyzing age-related structural changes in neural projections.
• The novel deep learning approach and visualization tools offer new insights into brain connectivity dynamics.
• Framework is optimized for wide-field microscopy data.