Super-resolution fluorescence microscopy improves spatial resolution, but this comes at a loss of image throughput and presents unique challenges in identifying optimal acquisition parameters. Microscope automation routines can offset these drawbacks, but thus far have required user inputs that presume a priori knowledge about the sample. Here, we develop a flexible illumination control system for localization microscopy comprised of two interacting components that require no sample-specific inputs: a self-tuning controller and a deep learning molecule density estimator that is accurate over an extended range. This system obviates the need to fine-tune parameters and demonstrates the design of modular illumination control for localization microscopy.
Applications of a deep learning method for anti-aliasing and super-resolution in MRI