Three-dimensional localization microscopy using deep learning

Zelger, Philipp; Kaser, Keith A.; Rossboth, Benedikt K.; Veľas, Lukáš; Schütz, Gerhard J.; Jesacher, Alexander

doi:10.1364/oe.26.033166

Cited by 54 publications

(44 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The precisions obtained with SALM/DONALD+ will thus be generally better, but can only be achieved in practice when the localization estimator considers PSF shapes as well. A maximum-likelihood algorithm operating on an accurate PSF model is a suitable choice in this regard [30,31]. We note that SALM/DONALD+, i.e.…”

Section: Results From Precision Calculationsmentioning

confidence: 99%

“…This step is required to facilitate a meaningful comparison between the outputs of our single-emitter estimator and the theoretical precisions. A home-programmed maximum likelihood algorithm [31] was subsequently employed to estimate 3D positions, signals and background levels of each localization from molecule images measuring 13×13 pixels. The PSF used by the MLE algorithm was calculated according to Ref.…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Defocused imaging exploits supercritical-angle fluorescence emission for precise axial single molecule localization microscopy

Zelger¹,

Bodner²,

Veľas³

et al. 2020

Biomed. Opt. Express

Self Cite

View full text Add to dashboard Cite

Single molecule localization microscopy (SMLM) is one of the key techniques that break the classical resolution limit in optical imaging. It is based on taking multiple recordings of a sample, each showing only a sparse arrangement of spatially well separated fluorescent molecules which can be localized at nanometer precision. While localizing along the lateral directions is usually straightforward, estimating axial positions at a comparable precision is known to be much harder, which is due to the relatively large depth of focus provided by the microscope optics. Whenever a molecule is sufficiently close to the coverslip, it becomes feasible to draw additional information from near field coupling effects: super-critical angle fluorescence (SAF) appears and can be exploited to boost the axial localization precision. Here we propose defocused imaging as a SMLM strategy that is capable of leveraging the information contained in SAF. We show that, regarding axial localization precision, our approach is superior to established SAF-based approaches. At the same time it is simple and can be conducted on any research-grade microscope where controlled defocusing on the order of a few hundred nanometers is possible.

show abstract

Section: Results From Precision Calculationsmentioning

confidence: 99%

Section: Resultsmentioning

confidence: 99%

Defocused imaging exploits supercritical-angle fluorescence emission for precise axial single molecule localization microscopy

Zelger¹,

Bodner²,

Veľas³

et al. 2020

Biomed. Opt. Express

Self Cite

View full text Add to dashboard Cite

show abstract

“…For benchmarking, we compare our vectorial PR (VIPR) to two alternatives: (1) analytically optimizing (with our method) the scalar diffraction model and (2) adding Zernike polynomials to the initially designed phase mask using the vectorial model [21,34]. Note that using Zernike polynomials requires an additional step to address observed wobble, by fitting the z-stack with the initial guessed mask to center it; this only works when the initial guess is close to the actual mask.…”

Section: Comparison With Existing Methodsmentioning

confidence: 99%

“…in the typical case of PR from a z-stack of a bead on the coverslip surface, these correspond to different values of . Common examples of cost functions include the L1 [34] and L2 [24] norms. Other cost functions employ statistical estimation, e.g.…”

Section: Cost Functionsmentioning

confidence: 99%

“…Shows the average time required to calculate a single gradient as a function of number of pixels (binning) in the phase mask. The algorithm is orders of magnitudes faster than numerical pixel-wise gradient calculation [24], and much faster than fitting Zernike polynomials, which can take ~30 minutes when implemented using the MATLAB function fminunc [34]. The results were timed on a laptop with INTEL i7-8750H CPU and NVIDIA GeForce RTX 2070 GPU.…”

Section: Phase Mask Optimization Workflowmentioning

confidence: 99%

See 1 more Smart Citation

VIPR: Vectorial Implementation of Phase Retrieval for fast and accurate microscopic pixel-wise pupil estimation

Ferdman

Nehme

Weiss

et al. 2020

Preprint

View full text Add to dashboard Cite

In microscopy, proper modeling of the image formation has a substantial effect on the precision and accuracy in localization experiments and facilitates the correction of aberrations in adaptive optics experiments. The observed images are subject to polarization effects, refractive index variations and system specific constraints. Previously reported techniques have addressed these challenges by using complicated calibration samples, computationally heavy numerical algorithms, and various mathematical simplifications. In this work, we present a phase retrieval approach based on an analytical derivation of the vectorial diffraction model. Our method produces an accurate estimate of the system phase information (without any prior knowledge) in under a minute.

show abstract

Advancing Spectrally‐Resolved Single Molecule Localization Microscopy with Deep Learning

Manko

Godet

2023

Small

View full text Add to dashboard Cite

Spectrally‐resolved single molecule localization microscopy (srSMLM) is a recent technique enriching single molecule localization microscopy with the simultaneous recording of spectra of the single emitters. srSMLM resolution is limited by the number of photons collected per emitters. Sharing a photon budget to record the localization and the spectroscopic information results in a loss of spatial and spectral resolution—or forces the sacrifice of one at the expense of the other. Here, srUnet—a deep‐learning Unet‐based image processing routine trained to increase the spectral and spatial signals to compensate for the resolution loss inherent in additionally recording the spectral component is reported. Both localization and spectral precision are improved by srUnet—particularly for the low‐emitting species. srUnet increases the fraction of localization whose signal can be both spatially and spectrally characterized. It preserves spectral shifts and the linearity of the dispersion of light. It strongly facilitates wavelength assignment in multicolor experiments. srUnet is a simple post‐processing add‐on boosting srSMLM performance close to conventional SMLM with the potential to turn srSMLM into the new standard for multicolor single molecule imaging.

show abstract

Three-dimensional localization microscopy using deep learning

Cited by 54 publications

References 35 publications

Defocused imaging exploits supercritical-angle fluorescence emission for precise axial single molecule localization microscopy

Defocused imaging exploits supercritical-angle fluorescence emission for precise axial single molecule localization microscopy

VIPR: Vectorial Implementation of Phase Retrieval for fast and accurate microscopic pixel-wise pupil estimation

Advancing Spectrally‐Resolved Single Molecule Localization Microscopy with Deep Learning

Contact Info

Product

Resources

About