Rapid robust high-fidelity 3D neuronal extraction from multiview calcium imaging datasets.

DeepWonder3D achieves high-fidelity 3D neuronal extraction from large-scale calcium imaging datasets, reducing computational costs tenfold and processing tens of thousands of neurons in hours.

• Why it matters: Accurate extraction of neuronal calcium dynamics from 3D imaging data is crucial for understanding neural coding but is hampered by noise, scattering, and computational demands, limiting large-scale analysis.
• What they did: The team developed DeepWonder3D, an end-to-end pipeline that processes multiview projections from various microscopy techniques, integrating denoising, registration, background removal, and fusion for high-resolution datasets.
• The result: DeepWonder3D outperforms existing methods in localization accuracy, enabling rapid, high-fidelity 3D calcium imaging of extensive neuronal populations, facilitating advanced neural circuit studies.