<b>Self-adaptive fine-tuning of deep learning super-resolution microscopy for artifact suppression in live-cell imaging</b>

Existing pre-trained super-resolution models often struggle to adapt to changes in imaging conditions, leading to artifacts in live-cell imaging. In this study, we introduce an innovative online fine-tuning method specifically designed for deep-learning super-resolution microscopy (DSRM). This approach minimizes artifacts in live-cell imaging by dynamically adjusting model parameters based on data obtained directly from live-cell observations. We applied this method in single-molecule localization microscopy (SMLM)-enabled DSRM, successfully visualizing nanoscale organelle interactions with significantly reduced artifacts.