Deep-STORM: super-resolution single-molecule microscopy by deep learning

Abstract

We present an ultra-fast, precise, parameter-free method, which we term\nDeep-STORM, for obtaining super-resolution images from stochastically-blinking\nemitters, such as fluorescent molecules used for localization microscopy.\nDeep-STORM uses a deep convolutional neural network that can be trained on\nsimulated data or experimental measurements, both of which are demonstrated.\nThe method achieves state-of-the-art resolution under challenging\nsignal-to-noise conditions and high emitter densities, and is significantly\nfaster than existing approaches. Additionally, no prior information on the\nshape of the underlying structure is required, making the method applicable to\nany blinking data-set. We validate our approach by super-resolution image\nreconstruction of simulated and experimentally obtained data.\n