Optics Express

Super-resolution imaging has revolutionized the study of systems ranging from molecular structures to distant galaxies. However, existing super-resolution methods require extensive calibration and retraining for each imaging setup, limiting their practical deployment. We introduce a deviceagnostic deep-learning framework for super-resolution imaging of point-like emitters that eliminates the need for calibration data or explicit knowledge of optical system parameters. Our model is trained on a diverse, numerically simulated dataset encompassing a broad range of imaging conditions, enabling generalization across different optical setups. Once trained, it reconstructs superresolved images directly from a single resolution-limited camera frame with superior accuracy and computational efficiency compared to state-of-the-art methods. We experimentally validate our approach using a custom microscopy setup with ground-truth emitter positions. We also demonstrate its versatility on astronomical and single-molecule localization microscopy datasets, achieving unprecedented resolution without prior information. Our findings establish a pathway toward universal, calibration-free super-resolution imaging, expanding its applicability across scientific disciplines.

Molecular fluorescence microscopy is a leading approach to super-resolution and nanoscale imaging in life and material sciences. However, super-resolution fluorescence microscopy is often bottlenecked by system-specific calibrations and long acquisitions of sparsely blinking molecules. We present a deep-learning approach that reconstructs super-resolved images directly from a single diffraction-limited camera frame. The model is trained exclusively on synthetic data encompassing a wide range of optical and sample parameters, enabling robust generalization across microscopes and experimental conditions. Applied to dense terrylene samples with 150 ms acquisition time, our method significantly reduces reconstruction error compared to Richardson-Lucy deconvolution, ThunderSTORM multi-emitter fitting, and DECODE based on deep learning. The results confirm the ability to resolve emitters separated by 35 nm at 580 nm wavelength, corresponding to seven-fold resolution improvement beyond the Rayleigh criterion. Furthermore, we demonstrate strong generalization ability of the developed model and its resilience across a broad range of noise levels, numerical apertures, and optical aberrations. By delivering unprecedented details from a single short camera exposure without any prior information and calibration, our approach enables plug-and-play super-resolution imaging of fast, dense, or light-sensitive samples on common wide-field microscopy setups.

In single-molecule localization microscopy (SMLM), achieving precise localization hinges on obtaining an authentic point spread function (PSF) influenced by system and sample-induced aberrations. Here, we introduce VISPR (Vectorial in-situ PSF retrieval) retrieving precise 3D PSF models considering both system and sample-induced aberrations under SMLM conditions. By employing the theory of vectorial PSF model and maximum likelihood estimation (MLE) phase retrieval, VISPR is capable of reconstructing an accurate 3D PSF model achieving the theoretically minimum uncertainty and accurately reflecting three-dimensional information of single molecules. This capability empowers accurate 3D super-resolution reconstruction in 3D SMLM. Additionally, VISPR applies to low signal-to-noise ratio circumstances and is adept at retrieving high-frequency details of the experimental PSF across an extensive depth range--a challenging feat for alternative approaches. As an effective tool, VISPR enables the quantitative assessment of aberrations induced by the system and sample environment. From the simulations and experiments, we verified the superiority and effectiveness of VISPR. It is essential to highlight that VISPR applies to various SMLM microscope modalities.

Single-molecule localization microscopy methods such as dSTORM require specific buffer conditions to enable blinking and detection of individual emitters, making them incompatible with live cell imaging and expansion microscopy. An alternative approach to achieve super-resolution without blinking is to observe the fluctuations of the emitter intensity over time. Existing localization algorithms for high-emitter density make use of radial symmetry or use artificial neural networks trained on single high-density frames to predict emitter positions. Here, we aim to improve the resolution by using a larger temporal context. We combine the U-Net architecture used previously for image reconstruction with multi-head attention used in the Transformer architecture. We compare the results to DECODE and eSRRF as well as to traditional fitting algorithms on public benchmark data. A generic pre-trained model is provided together with a fast and robust simulator for training data and all scripts needed to train custom networks.

Single molecule fluorescence localization with minimum photon flux imaging (MINFLUX) can achieve localization precisions in the small nanometer range or better under suitable conditions. Potentially adverse conditions, such as a fixed fluorescence dipole or optical aberrations, that could cause systematic localization errors, have received little attention up to now. Here, we study these effects in simulation. We find that biases occur for fluorophores with a fixed absorption dipole tilted out of the imaging plane. These become larger (up to about 25% of the diameter of the circle spanned by the doughnut center positions) the larger the tilt angle gets. As a rule of thumb the spread in bias is smaller than 5 nm in case the dipole orientation is less than 30{degrees} out of plane for the typical case of a doughnut probing circle of diameter 100 nm. For freely rotating dipoles only the primary aberrations astigmatism and coma contribute to bias. This bias depends on the position of the fluorophore inside the circular probing area of MINFLUX and can be significantly larger than the localization precision. We show that increasing the number of measurements over the circle from a triangular to a hexagonal pattern is beneficial for reducing bias in all cases. Iterative shrinking of the probing area can eliminate the position dependent bias completely, but a strong dependence on dipole orientation of the bias at the center of the probing area remains.

Single-molecule localization microscopy (SMLM) enables super-resolution imaging on conventional fluorescent microscopes. Spectroscopic SMLM (sSMLM) further allows highly multiplexed super-resolution imaging. We report an easy-to-implement symmetrically dispersed dual-wedge prism (SDDWP)-sSMLM design that maximizes photon utilization. We first symmetrically dispersed photons to the -1st and +1st orders in an optical assembly using two identical dual-wedge prisms (DWPs). Then we computationally extracted the fluorophores spatial position and spectral characteristics using photons in both the -1st and +1st orders. Theoretical analysis and experimental validation showed lateral and spectral precisions of 10.1 nm and 0.3 nm, respectively, representing improvements of 28% and 48% over our previous DWP-based system, where emitted photons are divided separately for spatial and spectral analyses.

Single-molecule localization based on the concept of MINFLUX allows for molecular resolution imaging and tracking. However, MINFLUX has a limited field-of-view (FOV) and therefore requires a precise pre-localization step. We propose ISM-FLUX, a localization technique that combines structured illumination with structured detection. We show via simulations that by replacing the point-detector with a small single-photon detector array (e.g., of 5 x 5 elements) and sequentially exciting the sample with four spatially separated doughnut-shaped beams, a localization uncertainty between 1 and 15 nm can be obtained over a FOV of more than 800 nm with 100 photons. The large FOV and the extra spatial information induced by the detector array relax the requirements on prior information on the fluorophores position. In addition, ISM-FLUX allows the localization of multiple molecules simultaneously. We calculate the effect of different parameters, such as the relative position of the doughnut beams, the number of detector pixels, the number of photons and the signal-to-background ratio, on the localization uncertainty. We predict that the combination of a good localization precision and the experimental simplicity of ISM-FLUX will help the wide adoption of MINFLUX and other derived microscopy techniques.

According to Rayleighs criterion, two incoherent emitters with a separation below the diffraction limit are not resolvable with a conventional fluorescence microscope. One method of Super-Resolution Microscopy (SRM) circumvents the diffraction-limited resolution by precisely estimating the position of spatiotemporally independent emitters. However, these methods of SRM techniques are not optimal for estimating the separation of two simultaneously excited emitters. Recently, a number of detection methods based on modal imaging have been developed to achieve the quantum Cramer-Rao lower bound (QCRB) to estimate the separations between two nearby emitters. The QCRB determines the minimum achievable precision for all possible detection methods. Current modal imaging techniques assume a scalar field generated from a point source, such as a distant source from an optical fiber or a pinhole. However, for fluorescently labeled samples, point emitters are single fluorophores that are modeled as dipole emitters and, in practice, are often freely rotating. Dipole radiation must be described by vectorial theory, and the assumption of a scalar field no longer holds. Here, we present a method to numerically calculate the QCRB for measuring the separation of two dipole emitters, incorporating the vectorial theory. Furthermore, we propose a near-quantum optimal detection scheme based on one of the modal imaging techniques, super-localization by image inversion interferometry (SLIVER), for estimating the separation of two freely rotating dipoles. In the proposed method, we introduce a vortex wave plate before the SLIVER detection to separate the radial and azimuthal components of the dipole radiation. With numerical simulations, we demonstrated that our method achieves non-divergent precision at any separation between two dipole emitters. We investigated several practical effects relevant to experimental measurements in super-resolution microscopy, including numerical aperture, detection bandwidth, number of estimation parameters, background, and misalignment on separation estimation. Our proposed measurement provides a near quantum-limited detection scheme for measuring the separation of two freely-rotating dipole emitters, such as fluorescently tagged molecules, which are commonly used in super-resolution microscopy.

Single-molecule localization microscopy methods extensively leverage the microscope point spread function (PSF) for fitting the molecules. Calibrating an accurate PSF model is especially difficult in the presence of depth-dependent aberrations which alter the PSF shape depending on the imaging depth. The aberrations at depths of a few micrometers become substantial enough to considerably impoverish the conventional calibration methods performance. In our work, we propose a novel spline model which enables the depth-dependent PSF model calibration by interpolating between the beads at arbitrary depths. We show that diffspline reduces the PSF intensity overestimation by 67.8 percentage points and underestimation by 21.8 percentage points. Moreover, it eliminates the depth-dependent bias and improves the localization precision two-fold compared to previous approaches.

Localization microscopy techniques have overcome the diffraction limit, enabling nanoscale biological imaging by precisely determining the positions of individual emitters. However, the performance of deep learning methods commonly applied to these tasks often depends significantly on the quality of training data, typically generated through simulation. Creating simulations that perfectly replicate experimental conditions remains challenging, resulting in a persistent simulation-to-experiment (sim2exp) gap. To bridge this gap, we propose a physics-informed generative model leveraging self-supervised learning directly on experimental data. Our model extends the Deep Latent Particles (DLP) framework by incorporating a physical Point Spread Function (PSF) model into the decoder, enabling it to disentangle learned realistic environments from precise emitter properties. Trained directly on unlabeled experimental images, our model intrinsically captures realistic background, noise patterns, and emitter characteristics. The decoder thus acts as a high-fidelity generator, producing fully labeled, realistic training images with known emitter locations. Using these generated datasets significantly improves the performance of supervised localization algorithms, particularly in challenging scenarios such as complex backgrounds and low signal-to-noise ratios. Our results demonstrate substantial improvements in localization accuracy and emitter detection, underscoring the practical benefit of our approach for real-world microscopy applications. We will make our code publicly available.

We investigate the influence of different accuracy-detection rate trade-offs on image reconstruction in single molecule localization microscopy. Our main focus is the investigation of image artifacts experienced when using low localization accuracy, especially in the presence of sample drift and inhomogeneous background. In this context we present a newly developed SMLM software termed FIRESTORM which is optimized for high accuracy reconstruction. For our analysis we used in silico SMLM data and compared the reconstructed images to the ground truth data. We observe two discriminable reconstruction populations of which only one shows the desired localization behavior.

DNA fiber assays are powerful tools for investigating replication dynamics at the single-molecule level. However, their application and widespread adoption has been hampered by the labor-intensive and tedious nature of manual analysis of large numbers of images. Quantification of labeled DNA fibers typically depends on subjective examination, selection, and annotation of individual fibers from fluorescence microscopy images reducing inter-user consistency, reproducibility, and experimental throughput. To address these issues, we developed DNAi, a computer vision tool based on deep learning allowing automated detection and quantification of labeled DNA fiber length. DNAi was trained on a large and diverse dataset of manually annotated images of DNA fibers and matches human performance and accuracy in segmentation and length measurement across a wide range of experimental conditions. The open-source tool includes a user-friendly interface, which permits visual validation and manual selection of segmented fibers. Overall, DNAi enables robust, rapid, and reproducible DNA fiber analysis, and is freely available. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=113 SRC="FIGDIR/small/679603v1_ufig1.gif" ALT="Figure 1"> View larger version (29K): org.highwire.dtl.DTLVardef@1c8fd78org.highwire.dtl.DTLVardef@78f40corg.highwire.dtl.DTLVardef@de60e5org.highwire.dtl.DTLVardef@1ab7db5_HPS_FORMAT_FIGEXP M_FIG C_FIG

In the rapidly advancing field of biological imaging, there is a great need for high-resolution imaging techniques that are both cost-effective and accessible, for example to better observe and understand dynamics in intracellular processes. Structured illumination microscopy (SIM) is the method of choice to achieve high axial and lateral resolution in living samples due to its optical sectioning and minimal phototoxicity. However, the high cost and complexity of conventional SIM systems limit their wide application. In our work, we present an open-source, fully-automated, two-color structured illumination module that is compatible with commercially available microscope stands. The compact design, consisting of low-cost single-mode fiber-coupled lasers and a digital micromirror device (DMD), is integrated into the open-source acquisition and control software (ImSwitch) in order to realize real-time super-resolution imaging. This developed system achieves up to a 1.55-fold improvement in lateral resolution compared to conventional wide-field microscopy. To rationally design this module, we developed a model to ensure optimal DMD diffraction per-formance using tilt and roll pixels, thus covering a wide range of low-cost video projectors for use in coherent SIM setups. Our goal is to democratize SIM-based super-resolution microscopy by providing both comprehensive open-source documentation and a modular software framework that works with various hardware components (e.g. cameras, stages) and reconstruction algorithms. In this way, we try to upgrade as many devices as possible to the super-resolution realm.

We present a novel method to improve the imaging efficiency of tiling light sheet microscopy. In the method, scanning non-coaxial beam arrays synchronized with regional virtual confocal slits are used to illuminate imaging plane. There are two advantages. One is the imaging efficiency increases proportional to the number of excitation beams within the non-coaxial beam array. The other is the width of the regional virtual confocal slits could be very wide without admitting off-focus fluorescence generated by the non-coaxial beam array, which makes the method easy to adopt and very robust in practice. We describe the method in detail, characterize the method via numerical simulations. The results suggest that the imaging efficiency and feasibility of the tiling light sheet microscopy could be improved significantly without affecting the 3D imaging ability by using the method. In additions, we propose several configurations to implement the method in practice.

Spatial resolution is crucial for imaging subcellular structures. The advent of three-dimensional structured illumination microscopy (3D-SIM) greatly benefits the biology community, providing a powerful tool for imaging organelles with a two-fold resolution enhancement in all three dimensions. However, the axial resolution of 3D-SIM is limited to around 300 nm, which is inferior to its lateral resolution. Here, a novel method called image interference SIM (I2SIM) is reported, which utilizes two oppositely positioned objectives to detect fluorescence emission interference under three-beam excitation. By incorporating spectral modulation and spatial domain Frobenius-Hessian optimization, I2SIM achieves an axial resolution approximately twice that of 3D-SIM, reaching around 130 nm. Furthermore, the potential of I2SIM for imaging subcellular structures is demonstrated on various biological samples, including microtubules, actin filaments, and mitochondrial outer membranes. The enhanced optical sectioning capability can be utilized to resolve axial structures that are challenging to discern using ordinary 3D-SIM.

With a limited dynamic range of an imaging system, there are always regions with signal intensities comparable to the noise level, if the signal intensity distribution is close to or even wider than the available dynamic range. Optical brain/neuronal imaging is such a case where weak-intensity ultrafine structures, such as, nerve fibers, dendrites and dendritic spines, often coexist with ultrabright structures, such as, somas. A high fluorescence-protein concentration makes the soma order-of-magnitude brighter than the adjacent ultrafine structures resulting in an ultra-wide dynamic range. A straightforward enhancement of the weak-intensity structures often leads to saturation of the brighter ones, and might further result in amplification of high-frequency background noises. An adaptive illumination strategy to real-time-compress the dynamic range demands a dedicated hardware to operate and owing to electronic limitations, might encounter a poor effective bandwidth especially when each digitized pixel is required to be illumination optimized. Furthermore, such a method is often not immune to noise-amplification while locally enhancing a weak-intensity structure. We report a dedicated-hardware-free method for rapid noise-suppressed wide-dynamic-range compression so as to enhance visibility of such weak-intensity structures in terms of both contrast-ratio and signal-to-noise ratio while minimizing saturation of the brightest ones. With large-FOV aliasing-free two-photon fluorescence neuronal imaging, we validate its effectiveness by retrieving weak-intensity ultrafine structures amidst a strong noisy background. With compute-unified-device-architecture (CUDA)-acceleration, a time-complexity of <3 ms for a 1000x1000-sized 16-bit data-set is secured, enabling a real-time applicability of the same.

Structured illumination microscopy (SIM) is among the most widely used super-resolution fluorescence microscopy techniques for visualizing the dynamics of cellular organelles, such as mitochondria, the endoplasmic reticulum, or the cytoskeleton. In its most wide-spread implementation, SIM relies on the creation of an interference pattern at the diffraction limit using the coherent addition of laser beams created by a diffraction pattern.\n\nSpatial light modulators based on liquid crystal displays allow SIM micro-scopes to run at image rates of up to hundreds of super-resolved images per second. Digital micromirror devices are another natural choice for creating interference-based SIM patterns, but are not used to their fullest potential because of the blazed grating effect. This effect arises due to the fixed angles between which the mirrors can be switched, creating a sawtooth arrangement of mirrors and thus leading to a change in the intensity distribution of the diffracted beams. This results in SIM patterns with varying modulation contrast which are prone to reconstruction artifacts.\n\nWe have carefully studied the blazed grating effect of DMDs by simulations, varying a range of parameters and compared the simulation results with experiments. This allowed us to identify settings which result in very high modulation contrast across all angles and phases required to generate 2-beam SIM pattern. The use of inexpensive industry-grade CMOS cameras as well as low-cost lasers enabled us to construct a cost-effective, high-speed SIM system. Reconstruction of the super-resolved SIM images is achieved on a recently demonstrated parallel-computing platform, which allowed us to visualize living cells with super-resolution at multiple reconstructed frames per second in real time. We demonstrate the versatility of this new platform by imaging cellular organelle dynamics based on live-cell fluorescent stains as well as with fluorescent protein stained samples.

Polarization-resolved second harmonic generation microscopy provides structural information about non-centrosymmetric biological samples such as collagen. It involves illuminating the sample with a focused laser beam having a variable linear polarization angle and recording the second harmonic signal as a function of this angle. However, accurate linear polarization control is challenging due to ellipticity introduced by reflections from mirrors and dichroic mirrors in the optical path. Waveplate-based compensation has emerged as the standard approach to address these distortions, but its effectiveness for quantitative measurements remains incompletely characterized. Here, we attempt to fill this gap by implementing an established automated waveplate compensation method based on a rotating half-waveplate in combination with a compensating quarter-waveplate. This was done on a commercial Leica TCS SP8 MP multiphoton microscope, making various hardware improvements and carefully documenting important experimental details. Despite significant effort, we consistently observed substantial unwanted residual polarization ellipticity, with amplitudes up to 0.25, persisting under optimal waveplate configurations. Our simulation analysis provides evidence that this limitation may arise from wavelength-dependent dichroic mirror birefringence combined with the broad spectral bandwidth (10nm to 20nm full width at half maximum) of femtosecond laser pulses. While the approach investigated here can compensate a single wavelength, different spectral components within the pulse experience different phase retardations from wavelength-dependent optical elements, potentially resulting in residual ellipticity that cannot be eliminated. Our simulations qualitatively reproduced key features of the experimental observations. These findings have important implications for quantitative polarization-resolved second harmonic generation microscopy and suggest that alternative approaches, including specimen rotation or picosecond laser sources with narrower bandwidth, should be investigated for applications requiring precise polarization control. To facilitate community investigation of these effects, we provide open-source analysis code and simulation files.

Modulation enhanced single-molecule localization microscopy (meSMLM), where emitters are sparsely activated with patterned illumination, increases the localization precision over SMLM. Furthermore, meSMLM improves the resolution over structured illumination microscopy while reducing the required amount of illumination patterns. These factors motivate enabling meSMLM in existing systems which employ patterned illumination intensity. Here, we introduce SpinFlux: modulation enhanced localization for spinning disk confocal microscopy. SpinFlux uses a spinning disk with pinholes in its illumination and emission paths, to illuminate select regions in the sample during each measurement. The resulting intensity-modulated emission signal is analyzed to localize emitters with improved precision. We derive a statistical image formation model for SpinFlux and we quantify the theoretical minimum uncertainty, in terms of the Cramer-Rao lower bound, for various illumination pattern configurations. We find that SpinFlux requires multiple patterns to improve the localization precision over SMLM, with the maximum improvement being 1.17 when using a single pattern. When using two pinholes on opposing sides of the emitter position, the x-localization precision can locally be improved 2.62-fold over SMLM, whereas the y-precision is improved by maximally a factor 1.12. When using pinholes in a triangular configuration around the emitter position, the localization precision is balanced over the xand y-directions at approximately a twofold local improvement over SMLM, at the cost of suboptimal precision in each individual direction. When doughnut-shaped illumination patterns, created with a phase mask in the illumination and emission paths, are used for SpinFlux, the local precision improvement over SMLM is increased 3.5-fold in the x- and y-directions. While localization on ISM data ideally results in an average global improvement of 1.48 over SMLM, or 2.10 with Fourier reweighting, SpinFlux is the method of choice for local refinements of the localization precision. Why it matters: One of the main objectives of singlemolecule localization microscopy (SMLM) is to improve the precision with which single molecules can be localized. This has been successfully achieved through modulation enhanced SMLM, which uses patterned illumination to increase the information content of signal photons. However, this technique relies on setups with increased technical complexity over SMLM. With SpinFlux, we locally enable a twoto 3.5-fold precision improvement over singlemolecule localization microscopy, which can be achieved with only minor modifications to existing spinning disk confocal microscopy setups (e.g. a phase mask in the illumination and emission paths). In addition, our modeling framework enables evaluation of a wide variety of spinning disk setups and therefore paves the way for optimal spinning disk design.

Fluorescent structures are nowadays commonly used in the field of biological imaging. However the emission of fluorescence is always given within a finite spectrum. When the imaging systems point spread function is measured, this can be described as an incoherent summation of individual single-wavelength PSFs, weighted by the emission spectrum. As the full-width-half-maximum of the PSF scales inversely with the wavelength and most emission spectra exhibit a fluorescence tail at longer wave-lengths, this contributes significantly to lateral broadening. In our work we theoretically quantify this effect by deriving an analytic expression, which has been verified against some numerical simulations (rel. error on average 3 %). We report a FWHM-broadening on the order of 10 nm and additionally propose a way to overcome this broadening by splitting the emission spectrum into multiple wavelength segments. The corresponding image data is recombined in Fourier space by weighted averaging, leading to an improved signal-to-noise ratio at high spatial frequencies and a reduction of the FWHM of up to 8 nm (relative reduction of the broadening by {approx} 70 %). We also introduce a corrected wavelength, which in combination with already existing PSF calculation tools, describes a theoretical PSF which incorporates the aforementioned broadening effect.