For a long time, the resolution of light microscopy was restricted to approximately 200 nm, as described by Abbe’s diffraction limit. Single-Molecule Localization Microscopy (SMLM) overcomes this limit by capturing many frames of a sample labeled with blinking fluorophores, where each frame shows a different subset of molecules. The fluorophores are located by fitting a model of their point spread function (PSF) and these localizations are combined to create an image with 20-50 nm resolution. Localizing a fluorophore is only possible when no other fluorescent molecules are active in the surrounding diffraction-limited area so that its PSF appears as an isolated blob. However, high-density regions unavoidably contain overlapping PSFs, which lead to localization errors as the number of emitters in the region of interest is unknown. To prevent this, the density of active fluorophores is typically kept between 0.01-0.1 μm−2 and data acquisition can take up to days, making dynamic imaging impossible. This theoretical research explores the possibility of enabling high-density SMLM by using Single-Photon Avalanche Diode (SPAD) arrays, detecting every incident photon with picosecond timing precision, instead of capturing the total intensity in a 10-100 ms interval like the conventionally used sCMOS cameras. The photon arrival times from simulated SPAD measurements are used to determine the number of emitters in the field of view, which is directly related to the second-order quantum coherence of the signal. This coherence can be calculated by dividing the arrival times into discrete time bins and convoluting the signal with itself. To correct the number of emitters for the bias that is introduced by the discretization of the data, an analytical expression is derived and validated with simulations. Using this correction, the number of Alexa647 fluorophores can be determined from a 0.1 ms simulated measurement with a relative standard error of 1% independent of emitter count, using a laser intensity of 330 kWcm−2 and 100% detection efficiency. Considering an experimental setting in which 10 kWcm−2 intensity and 10% detection efficiency are more realistic, a 1.5 s measurement is needed to obtain the same accuracy, and a 15 ms interval is required to obtain a standard error of 10%. The newly acquired information about the number of emitters will make it possible to locate fluorophores with overlapping point spread functions for multi-emitter fitting. Consequently, SPAD arrays will provide the ability to image high emitter densities, which enables faster data acquisition and dynamic imaging.

Live-cell imaging captures dynamic cellular behaviors and aims to maximize both spatial and temporal resolution while minimizing sample damage, enabling advancements in fundamental cell biology. However, spatial resolution is limited by the diffraction barrier of optical lenses, which prevents the visualization of many subcellular structures. Single-molecule localization microscopy (SMLM) overcomes this barrier, achieving resolutions as fine as 10 nm, but it typically requires millions of frames and higher illumination, which reduces temporal resolution and can damage the sample. Super-resolution Optical Fluctuation Imaging (SOFI) operates at lower illumination levels and requires hundreds of frames, leveraging the statistical relationships of blinking fluorophores to achieve n-fold spatial resolution based on the nth-order SOFI calculation. Despite its benefits, SOFI still demands too many frames and involves extensive post-processing, making it impractical for real-time live-cell imaging. Without real-time imaging, researchers are unable to make immediate decisions, ultimately costing valuable time for the researchers.

To address this limitation, we introduce a supervised deep learning model designed to accelerate second-order SOFI. Our model reconstructs super-resolved second-order SOFI images using just 20 frames, compared to the hundreds typically required, while maintaining a 2-fold improvement in spatial resolution and showing minimal background artifacts. We demonstrate that after being trained on real fixed-cell (static) mitochondria data, the model is able to reconstruct super-resolved images in a dynamic environment by moving the microscope stage. The model achieves real-time temporal resolutions of up to 4.85 fps, unlocking new possibilities for real-time studies of live-cell dynamics.