High-speed low-light two-photon voltage imaging is an emerging tool to simultaneously monitor neuronal activity from a large number of neurons. However, shot noise dominates pixel-wise measurements and the neuronal signals are difficult to be identified in the single-frame raw measurement. We developed a self-supervised deep learning framework for voltage imaging denoising, DeepVID, without the need for any high-SNR measurements. DeepVID infers the underlying fluorescence signal based on independent temporal and spatial statistics of the measurement that is attributable to shot noise. DeepVID achieved a 15-fold improvement in SNR when comparing denoised and raw image data.
Paper
Full text
Two-photon voltage imaging denoising by self-supervised learning
Semantic Scholar · Computer Science · 2023
Abstract
High-speed low-light two-photon voltage imaging is an emerging tool to simultaneously monitor neuronal activity from a large number of neurons. However, shot noise dominates pixel-wise measurements and the neuronal signals are difficult to be identified in the single-frame raw measurement. We developed a self-supervised deep learning framework for voltage imaging denoising, DeepVID, without the need for any high-SNR measurements. DeepVID infers the underlying fluorescence signal based on independent temporal and spatial statistics of the measurement that is attributable to shot noise. DeepVID achieved a 15-fold improvement in SNR when comparing denoised and raw image data.