Fig. 4: BayesDL-SIM for long-term live-cell SR imaging. | Nature Communications

Fig. 4: BayesDL-SIM for long-term live-cell SR imaging.

From: Bayesian deep-learning structured illumination microscopy enables reliable super-resolution imaging with uncertainty quantification

Fig. 4: BayesDL-SIM for long-term live-cell SR imaging.

a Time-lapse BayesDL-SIM SR imaging reveals F-actin dynamics over a three-h cell adhesion process after depositing a COS-7 cell onto a glass coverslip. b Representative SR images of F-actin reconstructed by Hifi-SIM (first column), DFCAN (second column), and BayesDL-SIM (third column) at two different time points. BayesDL AleaU (fourth column) and credibility (fifth column) maps are also shown. Intensity profiles along the lines in DFCAN and BayesDL-SIM reconstructions are presented at the far right, where BayesDL-SIM provides not only intensity values but also 3-σ CI centered on intensity values. c Comparison of SR-SIM images and corresponding EpisU predicted by BayesDL-SIM in the cases of KC (first column), KIC of subcellular structure (second column), KIC of sampling rate (third column), and KIC of fluorescence SNR (fourth column). d The time courses of the FG-EpisU over the three-h imaging duration in the KC and three different KIC scenarios. Scale bar, 8 μm (a), 1 μm (b, c). Gamma value, 0.8 for F-actin images in (b) and (c). Experiments were repeated independently for n = 3 COS-7 cells, all showing similar results.

Back to article page