Real-time adaptive drift correction for super-resolution localization microscopy

Abstract

Super-resolution localization microscopy involves acquiring thousands of image frames of sparse collections of single molecules in the sample. The long acquisition time makes the imaging setup prone to drift, affecting accuracy and precision. Localization accuracy is generally improved by a posteriori drift correction. However, localization precision lost due to sample drifting out of focus cannot be recovered as the signal is originally detected at a lower peak signal. Here, we demonstrate a method of stabilizing a super-resolution localization microscope in three dimensions for extended periods of time with nanometer precision. Hence, no localization correction after the experiment is required to obtain super-resolved reconstructions. The method incorporates a closed-loop with a feedback signal generated from camera images and actuation on a 3D nanopositioning stage holding the sample.

Ultraprecise single-molecule localization microscopy enables in situ distance measurements in intact cells

“…Dynamic drift correction approaches (9,19,20,38) adjust the position of the sample during the acquisition. In vitro, active correction relies on the optical trapping of molecules, which limits throughput and/or sample formats (9).…”

“…In vitro, active correction relies on the optical trapping of molecules, which limits throughput and/or sample formats (9). For cellular imaging, active drift correction has been performed by monitoring the position of the fiducials markers (20,38) or the brightfield image of the sample (19). Ma et al (38) used the signal of gold nanoparticles arbitrarily embedded into the cover slide with correction rate of 4 s. Faster sample correction rates (e.g., 1 to 2 s) come at the expense of reduced accuracy (e.g., 10 to 20 nm) (19,20).…”

“…For cellular imaging, active drift correction has been performed by monitoring the position of the fiducials markers (20,38) or the brightfield image of the sample (19). Ma et al (38) used the signal of gold nanoparticles arbitrarily embedded into the cover slide with correction rate of 4 s. Faster sample correction rates (e.g., 1 to 2 s) come at the expense of reduced accuracy (e.g., 10 to 20 nm) (19,20). Until now, implementations of active stabilization have been performed at slower correction rates of 1 to 10 s (9,19,20,38).…”

“…Ma et al (38) used the signal of gold nanoparticles arbitrarily embedded into the cover slide with correction rate of 4 s. Faster sample correction rates (e.g., 1 to 2 s) come at the expense of reduced accuracy (e.g., 10 to 20 nm) (19,20). Until now, implementations of active stabilization have been performed at slower correction rates of 1 to 10 s (9,19,20,38). Slow and/or inaccurate corrections result in considerably worse in situ resolution, even after post-analysis treatment such as filtering or grouping.…”

“…1, C and D). Unlike other SMLM instruments, the accuracy of the Feedback SMLM correction is not limited by the sample properties (19), fiducial characteristics (11,13,20), or fluorophore density (9,21). An additional advantage of our autonomous optical feedback loop for stage positioning is that the sample position is maintained during buffer exchanges, for example, during multicolor DNA-PAINT experiments ( fig.…”