Universal super-resolution for subcellular fluorescence imaging

Super-resolution fluorescence microscopy has emerged as an essential instrument for investigating subcellular structures and monitoring intracellular physiology. Despite rapid advances driven by deep learning, most existing approaches are modality specific and exhibit limited generalizability, restricting their practical applicability. Here we introduce universal super-resolution (UniSR), a framework that operates across microscopes and subcellular structures without optical modifications or prior sample knowledge. UniSR is initially pre-trained on simulated image pairs to establish mappings from low-to high-resolution features and subsequently fine-tuned with only one experimental image pair, reducing data demand and acquisition time. Across wide-field and point-scanning modalities, UniSR enhances resolution to match structured illumination, single-molecule localization, or stimulated emission depletion microscopy. Applied to diverse biological samples, it reveals nanoscale structures, dynamic processes, and organelle interactions with high spatiotemporal fidelity. UniSR provides a practical and broadly generalizable tool for super-resolution imaging in live-cell research.