Correcting drift in thick and opaque samples in optical microscopy
Super-resolution microscopy makes it possible to observe samples at the nanoscale, by precisely localizing individual molecules and tracking their movements within their environment. However, this level of precision makes acquisitions highly sensitive to even the smallest sample movements. A drift of only a few nanometers can lead to loss of focus, image blur, and misinterpretation of the actual position of the molecules being observed.
Several methods already exist to correct these drifts, but they are often limited to thin or transparent samples. Some rely on the reflection of infrared light at the interface between the sample and the coverslip, while others require the addition of fiducial markers. These approaches are more difficult to apply to thick, opaque, or highly scattering tissues, which cannot always be imaged using trans-illumination.
To address this challenge, the research teams led by Laurent Cognet at the LP2N (Bordeaux) and Laurent Groc at the IINS (Bordeaux) developed a stabilization method adapted to these types of samples1. Their approach combines homogenized differential phase contrast imaging, or hDPC, with cross-correlation-based analysis to automatically correct sample drift in three dimensions.
The principle is as follows: before the main experiment, the microscope rapidly records a reference stack of hDPC images around the desired focal plane. Each image corresponds to a precise depth position. During acquisition, hDPC images are regularly captured in parallel with fluorescence imaging. They are compared to the reference stack to determine whether the sample has drifted along the z-axis. If a shift is detected, the system automatically adjusts the microscope’s axial position to recover the correct focal plane.
Figure 2. (a) Immobilized fluorescent particle in a fixed brain slice and the expected behavior in the presence of focus drift during acquisitions. (b) Schematic of the optical setup using oblique back-illumination. (c) During autofocusing, the drift was calculated for every 5th frame acquired using the cross-correlation of gradients images (red curve, axial drift obtained right before its correction). Additionally, the current z-drive position of the microscope was recorded (blue curve). (d) The bead axial position recorded with and without active stabilization. The standard deviation was calculated using a sliding window of 100 frames.
One of the key strengths of this method is its use in oblique back-illumination. Unlike trans-illumination, this configuration illuminates the sample from the same side as detection, making it compatible with thick or opaque tissues. The label-free hDPC images provide enough structural detail to serve as stabilization references, without requiring the addition of fiducial markers.
The method was validated on several biological models. In fixed organotypic brain slices, it maintained the focal position with a precision of a few tens of nanometers. In live brain slices, it improved the quality of super-resolved maps of the extracellular space obtained through single-particle tracking. Finally, in liver slices, a particularly opaque and scattering tissue, active drift correction revealed more structural information than acquisitions performed without stabilization.
Figure 4. (a) Liver slices preparation. (b) hDPC image of a live liver slice tissue (scale bar = 20 μm). The liver slice ESC maps were obtained without (c) and with (d) active autofocusing (scale bars = 10 μm). The differences between the ECS maps are clearly visible, with several regions highlighted by circles. The area marked in yellow contains finer structural details in the ECS map reconstructed from the acquisition with active autofocusing. In contrast, the region indicated in green shows an area where the nanoparticles explored a different portion of the extracellular space. (e) Corresponding MSD curves. (f) Example of drift curves obtained in live liver slices, showing that the drift is nondirectional and exhibits strong variations over time.
By enabling stable imaging in thick, opaque, and living tissues, this approach opens new perspectives for high-resolution microscopy in complex biological environments. Compatible with back-illumination and requiring no additional labeling, it could be adapted to a wide range of microscopy configurations, both in biology and materials science.
(1) “Back-illumination Phase Imaging Enables Nanoscale Drift Stabilization in Non-transparent Biological Tissues” H. Manko, M. Tondusson, A. Boyreau, M. Meras, S. Bancelin, L. Groc, and L. Cognet*
ACS Photonics (2026) – https://doi.org/10.1021/acsphotonics.5c03066
* Laurent Cognet – Laboratoire Photonique Numérique et Nanosciences, Université de Bordeaux, Talence 33400, France
LP2N, Institut d’Optique Graduate School, CNRS UMR 5298, Talence 33400, France
Email: laurent.cognet@u-bordeaux.fr
