CaltechAUTHORS
  A Caltech Library Service

Real-time image denoising of mixed Poisson–Gaussian noise in fluorescence microscopy images using ImageJ

Mannam, Varun and Zhang, Yide and Zhu, Yinhao and Nichols, Evan and Wang, Qingfei and Sundaresan, Vignesh and Zhang, Siyuan and Smith, Cody and Bohn, Paul W. and Howard, Scott S. (2022) Real-time image denoising of mixed Poisson–Gaussian noise in fluorescence microscopy images using ImageJ. Optica, 9 (4). pp. 335-345. ISSN 2334-2536. doi:10.1364/OPTICA.448287. https://resolver.caltech.edu/CaltechAUTHORS:20211116-153719213

[img] PDF - Published Version
See Usage Policy.

19MB
[img] PDF (February 8, 2022) - Submitted Version
See Usage Policy.

2MB
[img] PDF - Supplemental Material
See Usage Policy.

6MB

Use this Persistent URL to link to this item: https://resolver.caltech.edu/CaltechAUTHORS:20211116-153719213

Abstract

Fluorescence microscopy imaging speed is fundamentally limited by the measurement signal-to-noise ratio (SNR). To improve image SNR for a given image acquisition rate, computational denoising techniques can be used to suppress noise. However, common techniques to estimate a denoised image from a single frame either are computationally expensive or rely on simple noise statistical models. These models assume Poisson or Gaussian noise statistics, which are not appropriate for many fluorescence microscopy applications that contain quantum shot noise and electronic Johnson–Nyquist noise, therefore a mixture of Poisson and Gaussian noise. In this paper, we show convolutional neural networks (CNNs) trained on mixed Poisson and Gaussian noise images to overcome the limitations of existing image denoising methods. The trained CNN is presented as an open-source ImageJ plugin that performs real-time image denoising (within tens of milliseconds) with superior performance (SNR improvement) compared to conventional fluorescence microscopy denoising methods. The method is validated on external datasets with out-of-distribution noise, contrast, structure, and imaging modalities from the training data and consistently achieves high-performance (>8dB) denoising in less time than other fluorescence microscopy denoising methods.


Item Type:Article
Related URLs:
URLURL TypeDescription
https://doi.org/10.1364/OPTICA.448287DOIArticle
https://doi.org/10.1101/2021.11.10.468102DOIDiscussion Paper
https://doi.org/doi:10.7274/r0-ed2r-4052DOIFMD dataset
https://github.com/ND-HowardGroup/Instant-Image-DenoisingRelated ItemCode
https://github.com/ND-HowardGroup/Instant-Image-Denoising/tree/master/Plugins/Model_validationRelated ItemCode
ORCID:
AuthorORCID
Mannam, Varun0000-0002-1866-6092
Zhang, Yide0000-0002-9463-3970
Zhu, Yinhao0000-0002-9435-4576
Nichols, Evan0000-0001-9835-9107
Wang, Qingfei0000-0003-2503-0639
Sundaresan, Vignesh0000-0001-9390-1681
Zhang, Siyuan0000-0003-0910-3666
Smith, Cody0000-0002-9831-1514
Bohn, Paul W.0000-0001-9052-0349
Howard, Scott S.0000-0003-3246-6799
Additional Information:© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement. Received 10 November 2021; revised 1 February 2022; accepted 6 February 2022; published 22 March 2022. We thank Prof. Joshua D. Shrout, Department of Civil and Environmental Engineering and Earth Sciences, University of Notre Dame, for providing M. xanthus cells for our experiments. Yide Zhang’s research was supported by the Berry Family Foundation Graduate Fellowship of Advanced Diagnostics & Therapeutics (AD&T), University of Notre Dame. The authors further acknowledge Notre Dame’s Center for Research Computing (CRC) for the use of GPUs to train the ML models and the Integrated Imaging Facility (NDIIF) for the use of the Nikon A1R MP confocal microscope and Nikon Eclipse 90i widefield microscope in NDIIF’s Optical Microscopy Core. Funding: Office of Science (0019312); Division of Chemical, Bioengineering, Environmental, and Transport Systems (1554516). The authors declare no conflicts of interest. Data availability: Data and Code Availability. The FMD dataset mentioned in this paper is publicly available in [25]. The code for training the Noise2Noise plugin and DnCNN plugin architectures and underlying the results presented in the paper are publicly available in [43]. Also, this repository includes the estimation of noise parameters using MATLAB codes, Noise2Noise plugin validation on the W2S dataset, Noise2Noise plugin validation on out-of-distribution samples, and plugin source code in java with ImageJ plugins. Additional details of our Noise2Noise plugin denoising validation on the large datasets are available in [50]. The out-of-distribution structure dataset (3D RCAN dataset) used to generate underlying denoising results presented in this paper is available in [33].
Funders:
Funding AgencyGrant Number
Berry Family FoundationUNSPECIFIED
University of Notre DameUNSPECIFIED
Department of Energy (DOE)DE-SC0019312
NSFCBET-1554516
Issue or Number:4
DOI:10.1364/OPTICA.448287
Record Number:CaltechAUTHORS:20211116-153719213
Persistent URL:https://resolver.caltech.edu/CaltechAUTHORS:20211116-153719213
Official Citation:Varun Mannam, Yide Zhang, Yinhao Zhu, Evan Nichols, Qingfei Wang, Vignesh Sundaresan, Siyuan Zhang, Cody Smith, Paul W. Bohn, and Scott S. Howard, "Real-time image denoising of mixed Poisson–Gaussian noise in fluorescence microscopy images using ImageJ," Optica 9, 335-345 (2022); DOI: 10.1364/OPTICA.448287
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:111887
Collection:CaltechAUTHORS
Deposited By: Tony Diaz
Deposited On:16 Nov 2021 17:17
Last Modified:08 Apr 2022 17:55

Repository Staff Only: item control page