Cost-function Optimized Maximal Overlap Drift Estimation for Single Molecule Localization Microscopy

Single-molecule localization microscopy (SMLM) methods enable fluorescence imaging of biological specimens with nanometer-scale resolution. Although fluorophore localization precision is theoretically limited only by photon statistics, in practice the resolution of SMLM images is often degraded by physical drift of the sample and/or the microscope during data acquisition. At present, correcting this effect requires either specialized stabilization systems or computationally intensive post-processing, and established drift correction algorithms based on image cross-correlation suffer from limited temporal resolution. In this study we introduce COMET, a new method for SMLM drift estimation which achieves a substantially higher precision, accuracy, and temporal resolution compared with existing algorithmic approaches. We demonstrate that improved drift estimation translates directly into higher SMLM image resolution, limited by localization precision rather than drift artifacts. COMET is applicable to all types of SMLM data, operating directly on 2D or 3D localization datasets, and is readily integrated into analysis workflows. We benchmark its performance using both simulations and experiments, including STORM, MINFLUX, and Sequential OligoSTORM measurements, where long acquisition times make drift correction particularly challenging. COMET is published as an open-source, Python-based software project and is also available on open cloud-computing platforms.