Drift Correction
Sample drift due to thermal fluctuations is very common and can hardly be avoided. Hence it needs to be corrected in order to obtain a high quality dataset. While axial drift can be efficiently be avoided using focus-lock systems, lateral drift needs to be corrected post-acquisition. There are two popular ways for drift correcting SMLM datasets.
- Fiducial-based drift correction (FBDC)
Fiducial markers are placed on the sample and imaged along with the sample. Fiducial markers need to be diffraction limited and clearly visible without blinking or bleaching. Often gold beads are used here, but fluorescent beads also provide reliably fiducial markers. The fiducial are localized as any other molecule, but since they dont bleach and are thus visible throughout the entire image stack, their position can be tracked and gives information about the sample drift that can be used to correct for it.
- Redundant cross-correlation (RCC) [REF]
RCC finds the drift by comparing substacks of the dataset using cross-correlation. The resulting XY shift is then corrected using interpolation between the different coordinates.
In contrast to RCC, FBDC is independent of the sample and thus tends to give a better results on samples that dont allow robust cross-correlation, like point-like structures. On datasets that contain visible structures, like microtubules, RCC performs very well.
We implemented a simple viewer to visually check the localization output.
-
Input formats are ThunderStorm and "splineFitter", i.e. the output data generated by the splineFitter module.
-
This module further allows to filter the localization data using common filtering parameters (frame, sigma, photons, uncertainty).
-
The RCC drift correction was developed by the Huang lab and is available via their lab website.
Wang, Yina, et al. "Localization events-based sample drift correction for localization microscopy with redundant cross-correlation algorithm." Optics express 22.13 (2014): 15982-15991.
Select Localize/Register > Viewer / CC drift correction from the SPARTAN menu.
- Select the input data format and navigate to the localization file. Default format: ThunderStorm.
- Select the drift correction method (RCC, DCC, MCC). Default: RCC.
- Run drift correction. The solver iterations are displayed in the MatLab main window. A new figure will show the detected drift for the xy axis.
- Filter the data according to the parameters: frame, sigma, photons, uncertainty.
- Visualize the localization (Gaussian blurred 2D histogram) after selecting a suitable pixel size.
