Extended Data Fig. 1: The forward imaging model of fluctuation-based samples and reconstruction workflow of the FRC-assisted parameterless SACD.

(a) If it is in real experiments, the real fluorescent content we imaged includes not only in-focus fluorescent molecules (‘Desired fluctuation’), but also out-of-focus and cytosol fluorescence background. Then the fluorescence signal accompanied with baseline signal (‘Background’) is encoded by the microscopy PSF, and sampled by the sensor (‘Sampling & Noise’). The actual measurements will always deviate from the desired fluctuation model from SOFI (‘Unbiased fluctuation’, see also Methods). (b) Workflow and representative example (a COS-7 cell labelled with QD₅₂₅) of the FRC-assisted parameterless SACD. Left insets of top panel: Fourier transforms, and the corresponding FRC resolutions of different stages are labelled on the top right; Right insets of top panel: Magnified views from the white box. In Step1, the image sequence for reconstruction is captured. In Step 1.5, the PSF of raw data for the pre-deconvolution is calculated by the theoretical model. In Step2, we perform RL deconvolution for each image and it will be stopped at half of the iteration of the maximal resolution (lowest FRC value). In Step3, we calculate 2^nd order autocorrelation cumulant. Then in Step 3.5, we estimate the FRC resolution of the resulting image to calculate PSF for post-deconvolution. In Step4, we employ post-RL-deconvolution for the autocorrelation image and it will be stopped when reaching the maximum resolution. (c) Comparisons of the pre-deconvolution using different iteration times. Microtubule filaments in a COS-7 cell labelled with QD₅₂₅ imaged by SD-confocal (averaged by 20 frames) and SACD under 9, 20, 50, and 100 iteration times, respectively, of pre-deconvolution configurations. Experiments were repeated ten times independently with similar results. Scale bars: (b, main panel and left inset) 5 μm; (b, right inset; c) 1 μm.