TAG-SPARK: Empowering High-Speed Volumetric Imaging With Deep Learning and Spatial Redundancy

Area of Science:

• Neuroscience
• Biophysics
• Optical Imaging

Background:

• Two-photon high-speed fluorescence calcium imaging is crucial for neuroscience but faces SNR limitations due to speed-quality tradeoffs.
• Low signal photon flux in current techniques compromises the resolution of neural activity capture.

Purpose of the Study:

• To develop a contrast-enhanced, video-rate volumetric imaging system to overcome SNR limitations in high-speed calcium imaging.
• To improve the spatiotemporal resolution and image quality of neural activity recordings.

Main Methods:

• Integration of a tunable acoustic gradient (TAG) lens-based high-speed microscopy with a TAG-SPARK denoising algorithm.
• Utilizing spatial redundancy in dense z-sampled 4D (xyzt) datasets for self-supervised model training.
• High-speed dense z-sampling at sub-micrometer intervals.

Main Results:

• >700% enhancement in signal-to-noise ratio (SNR) was achieved.
• Fast-spiking functional profiles of neuronal activities were preserved.
• In vivo imaging of Purkinje cells revealed novel dendritic-to-somatic signal convolution.

Conclusions:

• The developed system provides high-speed, high-SNR calcium imaging, advancing neural activity capture.
• This technique facilitates a deeper understanding of neuronal transduction pathways within complex 3D neuronal architectures.
• The findings enable more detailed observation of neural dynamics and signal processing in the brain.