SpecSeg

SpecSeg segments neuronal regions of interest (ROIs) in calcium imaging datasets to detect and extract activity from cell bodies, dendrites, and axons in chronic recordings of awake, behaving animals.

Key Features:

• Cross-spectral power-based segmentation: Employs cross-spectral power analysis across multiple frequencies to detect ROIs optimized for neuronal compartments such as cell bodies, dendrites, and axons.
• Imaging modality compatibility: Supports data from chronic glass windows, GRIN lenses, single-photon (1P) miniscope and two-photon (2P) microscopy.
• Motion correction integration: Incorporates motion correction by adapting NoRMCorre (Pnevmatikakis & Giovannucci 2016) and processes sbx file formats.
• Neuropil correction and signal extraction: Performs neuropil correction and extracts fluorescence signals to account for background fluorescence contributions.
• ROI matching across sequential recordings: Matches ROIs across sequential recordings to enable longitudinal tracking of identified neuronal structures.

Scientific Applications:

• Population-level neural coding: Enables segmentation and extraction of calcium signals from large neural populations to study coding mechanisms.
• Longitudinal studies in behaving animals: Facilitates tracking of the same ROIs across chronic recordings in awake, behaving animals for longitudinal analyses.
• Multi-modal calcium imaging analysis: Applies to both single-photon (miniscope) and multi-photon datasets for comparative or combined analyses.

Methodology:

Segmentation uses cross-spectral power analysis and Fourier-derived low-frequency temporal components with temporal cross-correlations between pixels and their neighbors; motion correction adapts NoRMCorre; neuropil correction and signal extraction are applied; sbx file support is included.