Single-molecule localization microscopy (SMLM) in combination with DNA barcoding (DNA-PAINT) enables easy-to-implement multi-target super-resolution imaging. However, image acquisition is slow because of the need to spatio-temporally isolate single emitters and to collect sufficient statistical data to generate a super-resolved image. Here, we bypass this limitation by utilizing a neural network, DeepSTORM, that can predict super-resolved SMLM images from high-emitter density data. This reduces the acquisition time 10- to 20-fold, enabling image acquisition as short as one minute. Integrating weak-affinity DNA labels allows precise control of single-molecule emitter densities, which enables recording of training, ground truth, and testing data from the same sample. Sequential imaging of multiple targets using different DNA barcodes with the same fluorophore enables aberration-free multi-target imaging (Exchange-PAINT). The constant exchange of fluorophore labels at target sites minimizes signal loss for long acquisition times, which allows imaging large samples in a matter of minutes. The concept is transferable to other weak-affinity, non-covalent fluorophore labels.

© (2023) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.

K. K. Narayanasamy, J. V. Rahm, S. Jang, and M. Heilemann, "Multi-color super-resolution microscopy accelerated by a neural network," Proc. SPIE 12386, Single Molecule Spectroscopy and Superresolution Imaging XVI, 123860B (Presented at SPIE BiOS: January 29, 2023; Published: 15 March 2023); https://doi.org/10.1117/12.2657442.