Fast analysis method for stochastic optical reconstruction microscopy using multiple measurement vector model sparse Bayesian learning

Compressed sensing (CS) can be used in fluorescence microscopy to improve the temporal resolution of stochastic optical reconstruction microscopy (STORM). Currently, most algorithms used in CS-STORM belong to the single measurement vector (SMV) model, where each super-resolution image is recovered individually from a raw frame, thereby prolonging the computational time. Here, we apply the multiple measurement vector (MMV) model CS algorithm to STORM, wherein all raw images are converted into a matrix and recovered by solving the simultaneous sparse recovery problem. We use the MMV model-based sparse Bayesian learning (SBL) algorithm to reconstitute the raw images of STORM, then compare its imaging resolution and run time with the SMV model CS algorithms. The simulated and experimentally recovered super-resolution images prove that the resolution of MMV model SBL (M-SBL) is comparable with the SMV model algorithm, while the run time is far less and decreases from several hours to several minutes. The high resolution and shorter reconstitution time make M-SBL a promising real-time image reconstruction method for CS-STORM.