High-speed in vivo calcium recording using structured illumination with self-supervised denoising.

Area of Science:

• Neuroscience
• Biophysics
• Optical Imaging

Background:

• High-speed widefield fluorescence imaging of neural activity is crucial for understanding brain function.
• Signal fluctuations from background and noise limit current imaging techniques.

Purpose of the Study:

• To develop and evaluate novel computational image reconstruction methods for reducing noise in high-speed in vivo neural imaging.
• To improve the signal-to-noise ratio and data quality of ultrafast genetically encoded calcium indicators.

Main Methods:

• Implementation of a novel pseudo-HiLo (pHiLo) reconstruction method combining pseudo-widefield (pWF) and structured illumination (SI) data.
• Comparison of pHiLo with interleaved optical sectioning structured illumination microscopy (OS-SIM) and pWF reconstruction.
• Integration of a deep learning-based noise suppression method (DeepCAD-RT) for real-time processing.

Main Results:

• The denoised pHiLo reconstruction achieved a 75% increase in peak-to-noise ratio (PNR) compared to pWF at 500 Hz.
• DeepCAD-RT further enhanced PNR by 59% with OS-SIM reconstruction.
• Both pHiLo and OS-SIM reconstructions reduced background signal correlations by approximately 65%.

Conclusions:

• The developed pHiLo reconstruction method effectively reduces background and shot noise in high-speed neural imaging.
• Combining pHiLo or OS-SIM with DeepCAD-RT significantly improves image quality and data reliability.
• These advancements enable clearer visualization of neural activity and facilitate the study of neuronal dynamics.