dNMF

dNMF performs joint image registration and non-negative matrix factorization to extract and demix neural activity from calcium imaging microscopy videos with non-rigid motion.

Key Features:

• Joint Optimization: Simultaneously optimizes image registration and signal demixing to correct motion-induced deformations while separating overlapping neural signals.
• Handling Non-Rigid Motion: Explicitly models and compensates for non-rigid motion, enabling analysis of datasets from freely moving animals.
• Improved Signal Extraction: Concurrent registration and demixing enhances accuracy of calcium trace extraction from populations of neurons.
• Versatility Across Experimental Conditions: Validated on simulated data and microscopy videos of semi-immobilized C. elegans, demonstrating robustness across conditions.
• Quantile Regression Time-Series Normalization: Applies quantile regression to normalize time-series traces and adjust for baseline intensity variations across animals or imaging conditions.

Scientific Applications:

• Neural population activity analysis: Extraction and demixing of calcium traces for studying population-level neural dynamics.
• Behavioral neuroscience: Enables investigation of large-scale neural dynamics underlying behavioral control by providing accurate activity estimates.
• Calcium imaging in moving specimens: Applicable to calcium imaging datasets with non-rigid motion, including freely moving animals and semi-immobilized C. elegans.

Methodology:

Simultaneous optimization of image registration and non-negative matrix factorization, followed by quantile regression time-series normalization.