Drift correction for fluorescence live cell imaging through correlated motion identification

Fluorescence live cell imaging is an important experimental technique for visualizing and recording dynamic cellular processes under physiological conditions. However, a problem commonly encountered during imaging is sample drift. Without correction, such drift will cause bias or even error in subsequent image analysis of intracellular movement. Traditional area-based and feature-based image registration techniques often fail to correct such drift because, in addition to whole cell drift, intracellular features of interest also undergo separate and complex motion. To address this problem, we developed an image registration technique based on identifying clusters of features that undergo highly correlated motion. Sample drift is determined from movement of these features and corrected. Experiments confirmed that this technique can effectively remove translational drift with sub-pixel accuracy. For drift that also involves rotation, an extension of the technique is developed to determine rotation angles both within and out of the image plane so that rotational drift can be corrected. The technique is general and can be used for drift correction in a broad range of biological studies.