Real-time denoising of fluorescence time-lapse imaging enables high-sensitivity observations of biological dynamics beyond the shot-noise limit

Abstract

A fundamental challenge in fluorescence microscopy is the inherent photon shot noise caused by the inevitable stochasticity of photon detection. Noise increases measurement uncertainty, degrades image quality, and limits imaging resolution, speed, and sensitivity. To achieve high-sensitivity imaging beyond the shot-noise limit, we provide DeepCAD-RT, a versatile self-supervised method for effective noise suppression of fluorescence time-lapse imaging. We made comprehensive optimizations to reduce its data dependency, processing time, and memory consumption, finally allowing real-time processing on a two-photon microscope. High imaging signal-to-noise ratio (SNR) can be acquired with 10-fold fewer fluorescence photons. Meanwhile, the self-supervised superiority makes it a practical tool in fluorescence microscopy where ground-truth images for training are hard to obtain. We demonstrated the utility of DeepCAD-RT in extensive experiments, including in vivo calcium imaging of various model organisms (mouse, zebrafish larva, fruit fly), 3D migration of neutrophils after acute brain injury, and 3D dynamics of cortical ATP (adenosine 5'-triphosphate) release. DeepCAD-RT will facilitate the morphological and functional interrogation of biological dynamics with minimal photon budget.

Multi-focus light-field microscopy for high-speed large-volume imaging

“…We modeled the camera noise mathematically to solve the problem in our work, but the performance could be unstable for different noise types and levels. Pre-processing the input data using some selfsupervised denoising methods could be a more versatile way to address the issue [36][37][38] .…”