Open-source 3D active sample stabilization for fluorescence microscopy

Abstract

Super-resolution microscopy has enabled imaging at nanometer-scale resolution. However, achieving this level of detail without introducing artifacts that could mislead data interpretation requires maintaining sample stability throughout the entire imaging acquisition. This process can range from a few seconds to several hours, particularly when combining live-cell imaging with super-resolution techniques. Here, we present a three-dimensional active sample stabilization system based on real-time tracking of fiducial markers. To ensure broad accessibility, the system is designed using readily available off-the-shelf optical and photonic components. Additionally, the accompanying software is open source and written in Python, facilitating adoption and customization by the community. We achieve a standard deviation of the sample movement within 1 nm in both the lateral and axial directions for a duration in the range of hours. Our approach allows easy integration into existing microscopes, not only making prolonged super-resolution microscopy more accessible but also allowing confocal and widefield live-cell imaging experiments spanning hours or even days.

Figure 1

Optical setup and axial calibration. (a) Optical setup of the stabilization system. The green part shows the stabilization setup, which is coupled to a microscope (partially shown in purple). The stabilization beam is coupled into the imaging path using a short-pass dichroic mirror, depicted in red in the imaging part. (b) Exemplary camera images of the XY and Z camera. The column on the right shows the astigmatism induced by the cylindrical lens, as imaged by the Z camera at different axial planes. (c) Autocorrelation of the images captured by the Z camera for the same z positions as in (b). The axial position of the sample is derived from the asymmetry of the autocorrelation function, calculated as the ratio of the summed intensities of the pixel values in H (green) and V (blue). (d) Measured autocorrelation asymmetry as a function of the $z$ position. The calibration measurement performed at the start of each experiment covers a smaller range (from −50 to +50 nm; see inset), where the asymmetry can be well approximated as a linear curve (dotted line). The calibration curve and the error bars are the mean and standard deviation over 10 iterations. The inset plot in the top left shows the horizontal and vertical components (shown in c) of the summed intensities over the green and blue boxes.

Figure 2

Stabilization code workflow. Flowchart with pseudocode describing the stabilization code that controls the hardware. The code is split into 10 different processes that are categorized into two types. The first type (green) represents the essential processes required for the stabilization code. The second type (pink) represents the nonessential part of the stabilization code. Similarly, the green arrows show essential data transfer, and the pink arrows show nonessential data transfer.

Figure 3

Stabilization results. (a) Sample position as monitored by the stabilization system and the corresponding histograms showing the distribution of the sample position across time. The parameter $\sigma$ is the standard deviation of the Gaussian fit to the histogram. (b) Corresponding position of the piezoelectric stage for each axis. (c) Power spectral density (PSD) of the stabilized sample position (colored) and the stage position (black).

Figure 4

Confocal images acquired with and without active stabilization. (a) Subset of a time series taken with the active stabilization on. The sample consisted of gold nanoparticles attached to the coverslip surface. Images were acquired every minute for 4 h. With the stabilization system on, there is no visible drift. (b) The same measurement with the stabilization system off shows a clear drift in both the lateral and axial directions over the course of 4 h (c and d) Measured drift in each image with respect to the start of the time lapse, calculated by phase correlation, with the stabilization system on (c) and off (d).