Improved superresolution microscopy imaging by sparse deconvolution with an interframe penalty

Hugelier, Siewert; Eilers, Paul H.C.; Devos, O.; Ruckebusch, Cyril

doi:10.1002/cem.2847

Cited by 18 publications

(12 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The resolved low spatial resolution distribution maps are combined to obtain a single superresolved map for each compound. Graphical abstract reproduced from Piqueras et al B, Plot 1: Image obtained from the average of 300 time frames of a single‐molecule fluorescence image; plot 2: Superresolved image obtained by averaging deconvolved frames constrained with sparsity; plot 3: Superresolved image obtained by averaging deconvolved frames constrained with sparsity and applying penalized interframe changes (partially reproduced from Hugelier et al). MCR‐ALS, multivariate curve resolution–alternating least square…”

Section: Going Beyond the Limits Of The Measurement: Superresolution mentioning

confidence: 99%

“…Sparsity is induced based on penalized deconvolution approaches tending to minimize either the L 1 or the L 0 norm, the latter option seeming to provide better results . An interesting way to improve even more the spatial definition passes through incorporating the interframe time‐based correlation . This correlation indicates that an emitter should stay present during a certain number of consecutive frames.…”

Section: Going Beyond the Limits Of The Measurement: Superresolution mentioning

confidence: 99%

“…56,57 An interesting way to improve even more the spatial definition passes through incorporating the interframe time-based correlation. 54 This correlation indicates that an emitter should stay present during a certain number of consecutive frames. Therefore, the deconvolution approach mentioned above is improved by including an additional condition that penalizes changes of emitter positions among consecutive frames.…”

Section: Going Beyond the Limits Of The Measurement: Superresolutiomentioning

confidence: 99%

See 2 more Smart Citations

Hyperspectral image analysis. When space meets Chemistry

Juan

2017

Journal of Chemometrics

View full text Add to dashboard Cite

Hyperspectral images provide spatial and structural information on samples.This article aims at providing an overview on hyperspectral image analysis research trends, mainly focusing on specific aspects related to this kind of analytical measurement, such as the global/local and the spectral/spatial dualities, the hurdles associated with image fusion strategies, and the design of tools devoted to enhance the spatial definition provided by the instrumental measurement. The complexity of the measurement and the wealth of possibilities to use image properties and information ensure a very lively field of research for the coming years. | INTRODUCTIONHyperspectral images are beautiful measurements exploiting at the most the spatial and structural information of samples. Images can nowadays focus on the smallest detail (spatial resolutions of nanometer in single-molecule fluorescence images) or on the largest element (pixels covering squared kilometer in remote sensing), and the chemical structure of compounds in a sample can be gathered using almost all imaginable spectroscopies and spectrometries, providing all possible facets from macromolecular to micromolecular structure of the constituents/elements in a sample.Hyperspectral image analysis needs to take into account the spatial/spectral duality, although the research tendency on image analysis has been for a long time to give a larger weight to a particular side of the coin, depending on the field of knowledge and the related applications. Thus, the image processing community has put the accent on the spatial side, whereas classical chemometrics has longer considered hyperspectral images as nothing else than another kind of spectroscopic data set. Such a bias towards space or spectroscopy is not a coincidence. Image processing was born to deal with gray or RGB images, where the small number of channels per pixel (either a gray scale value or 3 color coordinates) did not give much room to exploit information other than the one derived from spatial characteristics, such as textures and edges. On the other hand, chemometrics developed a wide span of multivariate analysis tools focused on interpreting the large amount of high quality structural and chemical information provided by spectroscopic data, and the sole use of this knowledge was often sufficient to give a satisfactory interpretation of the hyperspectral image measurement. Recent research efforts on hyperspectral image analysis point to the development or adaptation of data analysis tools that may benefit from the synergy of using simultaneously properties and information encoded in both spatial and spectral dimensions to improve the final results obtained and take the most of the original measurement.

show abstract

Section: Going Beyond the Limits Of The Measurement: Superresolution mentioning

confidence: 99%

Section: Going Beyond the Limits Of The Measurement: Superresolution mentioning

confidence: 99%

Section: Going Beyond the Limits Of The Measurement: Superresolutiomentioning

confidence: 99%

See 1 more Smart Citation

Hyperspectral image analysis. When space meets Chemistry

Juan

2017

Journal of Chemometrics

View full text Add to dashboard Cite

show abstract

“…To implement sparsity in super‐resolution imaging deconvolution, an L 0 ‐norm penalization was proposed in a recent algorithm (SPIDER) . L 0 ‐norm regularization strongly forces coefficients in x towards zero unless their contribution to the signal is very strong.…”

Section: Introductionmentioning

confidence: 99%

A criterion for automatic image deconvolution with L₀‐norm regularization

Ahmad

Hugelier

Vitale

et al. 2020

Journal of Chemometrics

View full text Add to dashboard Cite

Automatic penalty adjustment in sparse deconvolution with penalized least squares is required for improved reliability and broader applicability. In sparse deconvolution with an L0‐norm penalty, the latent signal is by nature discontinuous, and the magnitudes of the residuals and sparsity regularization terms are of different order of magnitude. This makes approaches such as generalized cross validation or L‐curve unsuitable in practice. The criterion proposed in this paper is based on the representation of the sum of the normalized residuals and regularization terms (SNT) as a function of the penalty parameter. We observed that the minimum of the SNT corresponds to the optimal value of the penalty parameter. This approach was tested in the context of super‐resolution fluorescence microscopy imaging. Both simulated and real live‐cell images characterized by different complexities and emitter densities were analyzed to assess the performance of the developed optimization strategy and to demonstrate its usefulness over manual tuning.

show abstract

“…As the localization method requires that two molecules are not simultaneously ignited with overlapping PSF, this requires the acquisition of thousands or even tens of thousands of images for a single reconstruction. Using compressed sensing schemes added to STORM with different penalty schemes to promote sparsity (Zhu et al ., ; Min et al ., ; Hugelier et al ., , ), or locating several sources simultaneously by direct computation (Huang et al ., ) or deep learning methods (Nelson & Hess, ), image acquisition has been sped by localizing simultaneously several molecules within the point spread function. Still we are dealing with extremely sparse individual images requiring hundreds or thousands of images to complete the restoration.…”

Section: Introductionmentioning

confidence: 99%

Superresolution method for a single wide‐field image deconvolution by superposition of point sources

Martínez

Toscani

Martı́nez

2019

Journal of Microscopy

View full text Add to dashboard Cite

Summary In this work, we present a new algorithm for wide‐field fluorescent micrsocopy deconvolution from a single acquisition without a sparsity prior, which allows the retrieval of the target function with superresolution, with a simple approach that the measured data are fit by the convolution of a superposition of virtual point sources (SUPPOSe) of equal intensity with the point spread function. The cloud of virtual point sources approximates the actual distribution of sources that can be discrete or continuous. In this manner, only the positions of the sources need to be determined. An upper bound for the uncertainty in the position of the sources was derived, which provides a criteria to distinguish real facts from eventual artefacts and distortions. Two very different experimental situations were used for the test (an artificially synthesized image and fluorescent microscopy images), showing excellent reconstructions and agreement with the predicted uncertainties, achieving up to a fivefold improvement in the resolution for the microscope. The method also provides the optimum number of sources to be used for the fit. Lay Description A new method is presented that allows the reconstruction of an image with superresolution from a single frame taken with a standard fluorescent microscope. An improvement in the resolution of a factor between 3 and 5 is achieved depending on the noise of the measurement and how precisely the instrument response function (point spread function) is measured. The complete mathematical description is presented showing how to estimate the quality of the reconstruction. The method is based in the approximation of the actual intensity distribution of the object being measured by a superposition of point sources of equal intensity. The problem is converted from determining the intensity of each point to determining the position of the virtual sources. The best fit is found using a genetic algorithm. To validate the method several results of different nature are presented including an artificially generated image, fluorescent beads and labelled mitochondria. The artificial image provides a prior knowledge of the actual system for comparison and validation. The beads were imaged with our highest numerical aperture objective to show method capabilities and also acquired with a low numerical aperture objective to compare the reconstructed image with that acquired with a high numerical aperture objective. This same strategy was followed with the biological sample to show the method working in real practical situations.

show abstract

Improved superresolution microscopy imaging by sparse deconvolution with an interframe penalty

Cited by 18 publications

References 20 publications

Hyperspectral image analysis. When space meets Chemistry

Hyperspectral image analysis. When space meets Chemistry

A criterion for automatic image deconvolution with L₀‐norm regularization

Superresolution method for a single wide‐field image deconvolution by superposition of point sources

Contact Info

Product

Resources

About

Improved superresolution microscopy imaging by sparse deconvolution with an interframe penalty

Cited by 18 publications

References 20 publications

Hyperspectral image analysis. When space meets Chemistry

Hyperspectral image analysis. When space meets Chemistry

A criterion for automatic image deconvolution with L0‐norm regularization

Superresolution method for a single wide‐field image deconvolution by superposition of point sources

Contact Info

Product

Resources

About

A criterion for automatic image deconvolution with L₀‐norm regularization