Improving STED microscopy with SUPPOSe: enhancing resolution from a single-image.

Here we apply the SUPPOSe algorithm on images acquired using Stimulated Emission Depletion (STED) microscopy with the aim of improving the resolution limit achieved. We processed images of the nuclear pore complex (NPC) from cell lines in which the Nup96 nucleoporin was endogenously labeled. This reference protein forms a ring whose diameter is ∼107 nm with 8 corners ∼42 nm apart from each other. The stereotypic arrangement of proteins in the NPC has been used as reference structures to characterize the performance of a variety of microscopy techniques. STED microscopy images resolve the ring arrangement but not the eightfold symmetry of the NPC. After applying the SUPPOSe algorithm to the STED images, we were able to solve the octagonal structure of the NPC. After processing 562 single NPC, the average radius of the NPC was found to beR= 54.2 ± 2.9 nm, being consistent with the theoretical distances of this structure. To verify that the solutions obtained are compatible with a NPC-type geometry, we rotate the solutions to optimally fit an eightfold-symmetric pattern and we count the number of corners that contain at least one localization. Fitting a probabilistic model to the histogram of the number of bright corners gives an effective labeling efficiency of 31%, which is in agreement with the values reported in for other cell lines and ligands used in Single Molecule Localization microscopy, showing that SUPPOSe can reliably retrieve sub-resolution, nanoscale objects from single acquisitions even in noisy conditions.

Combining deep learning with SUPPOSe and compressed sensing for SNR-enhanced localization of overlapping emitters.

We present gSUPPOSe, a novel, to the best of our knowledge, gradient-based implementation of the SUPPOSe algorithm that we have developed for the localization of single emitters. We study the performance of gSUPPOSe and compressed sensing STORM (CS-STORM) on simulations of single-molecule localization microscopy (SMLM) images at different fluorophore densities and in a wide range of signal-to-noise ratio conditions. We also study the combination of these methods with prior image denoising by means of a deep convolutional network. Our results show that gSUPPOSe can address the localization of multiple overlapping emitters even at a low number of acquired photons, outperforming CS-STORM in our quantitative analysis and having better computational times. We also demonstrate that image denoising greatly improves CS-STORM, showing the potential of deep learning enhanced localization on existing SMLM algorithms. The software developed in this work is available as open source Python libraries.

Device for real-time monitoring of oil-in-water and suspended solids based on thermal lens spectrometry and light scattering.

A novel system suitable for simultaneous monitoring of both oil-in-water and suspended solids based on thermal lens spectroscopy and forward light scattering is presented. The technique measures the concentration of dissolved hydrocarbons and simultaneously detects single oil droplets and suspended particles separately. The device was tested with injection water samples from an on-field water treatment plant, and hydrocarbon concentrations were measured with a precision better than 5% in the range of up to 100 ppm, reaching resolutions as low as 0.03 ppm. Particle detection was tested with model samples of dyed and undyed polystyrene spheres acting as absorption and scattering centers, which simulated oil droplets and suspended solids, respectively. We show that particles of different sizes are distinguished by the magnitude of the perturbations introduced in the signals, and their concentrations can be measured independently of dissolved components.