bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2025, Volume and Issue: unknown

Published: April 20, 2025

Abstract Optical calcium imaging is a powerful tool for recording neural activity across wide range of spatial scales, from dendrites and spines to whole-brain through two-photon widefield microscopy. Traditional methods analyzing functional data rely heavily on features, such as the compact shapes somas, extract regions interest their associated temporal traces. This dependency can introduce biases in time trace estimation limit applicability these different neuronal morphologies scales. To address limitations, Graph Filtered Temporal Dictionary Learning (GraFT) uses graph-based approach identify components based shared rather than proximity, enhancing generalizability diverse datasets. Here we present significant advancements GraFT algorithm, including integration more efficient solver L1 least absolute shrinkage selection operator (LASSO) problem application compressive sensing techniques reduce computational complexity. By employing random projections dimensionality, achieve substantial speedups while maintaining analytical accuracy. These significantly accelerate making it scalable larger complex Moreover, increase accessibility, developed graphical user interface facilitate running outputs GraFT. Finally, demonstrate utility beyond meso-scale imaging, vascular axonal imaging.

Language: Английский