No abstract

We present an ultra-fast, precise, parameter-free method, which we term Deep-STORM, for obtaining super-resolution images from stochastically-blinking emitters, such as fluorescent molecules used for localization microscopy. Deep-STORM uses a deep convolutional neural network that can be trained on simulated data or experimental measurements, both of which are demonstrated. The method achieves state-of-the-art resolution under challenging signal-to-noise conditions and high emitter densities, and is significantly faster than existing approaches. Additionally, no prior information on the shape of the underlying structure is required, making the method applicable to any blinking data-set. We validate our approach by super-resolution image reconstruction of simulated and experimentally obtained data.arXiv:1801.09631v3 [physics.optics]

Single-molecule super-resolution fluorescence microscopy and single-particle tracking are two imaging modalities that illuminate the properties of cells and materials on spatial scales down to tens of nanometers, or with dynamical information about nanoscale particle motion in the millisecond range, respectively. These methods generally use wide-field microscopes and two-dimensional camera detectors to localize molecules to much higher precision than the diffraction limit. Given the limited total photons available from each single-molecule label, both modalities require careful mathematical analysis and image processing. Much more information can be obtained about the system under study by extending to three-dimensional (3D) single-molecule localization: without this capability, visualization of structures or motions extending in the axial direction can easily be missed or confused, compromising scientific understanding. A variety of methods for obtaining both 3D super-resolution images and 3D tracking information have been devised, each with their own strengths and weaknesses. These include imaging of multiple focal planes, point-spread-function engineering, and interferometric detection. These methods may be compared based on their ability to provide accurate and precise position information of single-molecule emitters with limited photons. To successfully apply and further develop these methods, it is essential to consider many practical concerns, including the effects of optical aberrations, field-dependence in the imaging system, fluorophore labeling density, and registration between different color channels. Selected examples of 3D super-resolution imaging and tracking are described for illustration from a variety of biological contexts and with a variety of methods, demonstrating the power of 3D localization for understanding complex systems.

Abstract-We consider the problem of phase retrieval, namely, recovery of a signal from the magnitude of its Fourier transform, or of any other linear transform. Due to the loss of Fourier phase information, this problem is ill-posed. Therefore, prior information on the signal is needed in order to enable its recovery. In this work we consider the case in which the signal is known to be sparse, i.e., it consists of a small number of nonzero elements in an appropriate basis. We propose a fast local search method for recovering a sparse signal from measurements of its Fourier transform (or other linear transform) magnitude which we refer to as GESPAR: GrEedy Sparse PhAse Retrieval. Our algorithm does not require matrix lifting, unlike previous approaches, and therefore is potentially suitable for large scale problems such as images. Simulation results indicate that GESPAR is fast and more accurate than existing techniques in a variety of settings.Index Terms-Non-convex optimization, phase retrieval, sparse signal processing.

Deep Learning (DL) methods are powerful analytical tools for microscopy and can outperform conventional image processing pipelines. Despite the enthusiasm and innovations fuelled by DL technology, the need to access powerful and compatible resources to train DL networks leads to an accessibility barrier that novice users often find difficult to overcome. Here, we present ZeroCostDL4Mic, an entry-level platform simplifying DL access by leveraging the free, cloud-based computational resources of Google Colab. ZeroCostDL4Mic allows researchers with no coding expertise to train and apply key DL networks to perform tasks including segmentation (using U-Net and StarDist), object detection (using YOLOv2), denoising (using CARE and Noise2Void), super-resolution microscopy (using Deep-STORM), and image-to-image translation (using Label-free prediction - fnet, pix2pix and CycleGAN). Importantly, we provide suitable quantitative tools for each network to evaluate model performance, allowing model optimisation. We demonstrate the application of the platform to study multiple biological processes.