Online analysis of microendoscopic 1-photon calcium imaging data streams

Area of Science:

• Neuroscience
• Computational Biology
• Bioimaging

Background:

• In vivo calcium imaging allows observation of deep brain activity in freely moving animals.
• Constrained non-negative matrix factorization (CNMF-E) is an offline method for extracting neuronal signals, but is computationally intensive and memory-demanding.
• Existing methods hinder analysis of large datasets and real-time applications like closed-loop experiments.

Purpose of the Study:

• To develop efficient online algorithms for extracting single-neuronal activity from microendoscopic calcium imaging data.
• To overcome the computational and memory limitations of offline processing methods.
• To enable real-time analysis of neuronal activity for advanced experimental designs.

Main Methods:

• Developed OnACID-E, an online adaptation of CNMF-E, reducing memory and computation needs.
• Introduced a convolution-based background model for real-time microendoscopic data processing.
• Integrated modular algorithms with existing online motion correction and deconvolution techniques for a scalable analysis pipeline.

Main Results:

• OnACID-E and the convolution-based method achieve high-quality neuronal activity extraction comparable to offline CNMF-E.
• The online algorithms significantly outperform CNMF-E in terms of computing speed and memory efficiency.
• Demonstrated successful application on typical experimental datasets, validating their performance.

Conclusions:

• The developed online algorithms provide a faster and more memory-efficient alternative to CNMF-E for microendoscopic data analysis.
• These methods enable, for the first time, online analysis of live-streaming neuronal activity data, even on standard laptops.
• The scalable pipeline supports advanced research, including closed-loop experiments and large-scale data processing.