Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2024:3:0095.
doi: 10.34133/icomputing.0095. Epub 2024 Jul 4.

Intelligent Beam Optimization for Light-Sheet Fluorescence Microscopy through Deep Learning

Chen Li  1   2 , Mani Ratnam Rai  1   2 , Yuheng Cai  1   2 , H Troy Ghashghaei  2   3 , Alon Greenbaum  1   2   4
Affiliations

Intelligent Beam Optimization for Light-Sheet Fluorescence Microscopy through Deep Learning

Chen Li et al. Intell Comput. 2024.

Abstract

Light-sheet fluorescence microscopy (LSFM) provides the benefit of optical sectioning coupled with rapid acquisition times, enabling high-resolution 3-dimensional imaging of large tissue-cleared samples. Inherent to LSFM, the quality of the imaging heavily relies on the characteristics of the illumination beam, which only illuminates a thin section of the sample. Therefore, substantial efforts are dedicated to identifying slender, nondiffracting beam profiles that yield uniform and high-contrast images. An ongoing debate concerns the identification of optimal illumination beams for different samples: Gaussian, Bessel, Airy patterns, and/or others. However, comparisons among different beam profiles are challenging as their optimization objectives are often different. Given that our large imaging datasets (approximately 0.5 TB of images per sample) are already analyzed using deep learning models, we envisioned a different approach to the problem by designing an illumination beam tailored to boost the performance of the deep learning model. We hypothesized that integrating the physical LSFM illumination model (after passing it through a variable phase mask) into the training of a cell detection network would achieve this goal. Here, we report that joint optimization continuously updates the phase mask and results in improved image quality for better cell detection. The efficacy of our method is demonstrated through both simulations and experiments that reveal substantial enhancements in imaging quality compared to the traditional Gaussian light sheet. We discuss how designing microscopy systems through a computational approach provides novel insights for advancing optical design that relies on deep learning models for the analysis of imaging datasets.

PubMed Disclaimer

Conflict of interest statement

Competing interests: The authors declare that they have no competing interests.

Figures

Fig. 1.
Fig. 1.
Deep design approach to improve optical sectioning and image quality in LSFM. (A) Simplified optical setup with spatial light modulator (SLM) for controlling the illumination beam. The close-up highlights a challenge encountered when employing a Gaussian beam in light-sheet fluorescence microscopy: diffraction causes the illumination beam to widen at the edges. Consequently, out-of-focus beads are unintentionally illuminated, introducing noise into the image. BS, beam splitter; BE, beam expander. (B) Joint optimization scheme. The locations of randomly distributed beads are fed into a physical optical layer to generate simulated images. The prediction network outputs a 2D downsized image to predict the positions of beads in the focal plane. The deep learning network and input phase mask are simultaneously updated based on the loss function.
Fig. 2.
Fig. 2.
Simulation of DD-mediated optimization. (A) Optimized phase mask to improve bead detection accuracy. Compared with a Gaussian beam (flat phase mask), the butterfly beam exhibits better image contrast on the edges (see inside the blue rectangular areas), rendering a better prediction. (B) Beam profile comparison. FWHM for Gaussian and butterfly beams plotted against the direction of propagation. The graph illustrates that the butterfly beam exhibits reduced FWHM at the edges, while the Gaussian beam is narrower at the center of the field of view. (C and D) Network performance. A comparison of classification metrics (confusion matrix and prediction measurement) between frozen and butterfly phase scenarios. The optimized phase mask demonstrates improved bead detection capability and achieves higher scores across the evaluated metrics. (E) Comparison between middle and edge areas. The field of view is divided into middle and edge areas. The metrics are calculated separately, and the butterfly beam provides better performance on the areas near the edges.
Fig. 3.
Fig. 3.
Experimental results—characterization of butterfly beam. (A) Static profiles of Gaussian and butterfly beams along beam propagation. The propagation distance of 0 μm corresponds to the focal point of the excitation objective, where the Gaussian beam reaches its minimal waist. As the light sheet is generated by dithering the static beam up and down, light-sheet thickness was measured by summing the values along the columns. Experimental measurements of the beam were conducted using a single lens after the spatial light modulator, outside the immersion chamber. The scale bar is 10 μm. (B) Schematic depicting formation of light sheet through up-and-down dithering of static beam. The line profile along the scanning direction compares the Gaussian beam (in green) with the optimized/butterfly beam (in red). To maintain a narrow profile at the edges of the field of view, the deep design optimization elongated the profile in the direction of the scan.
Fig. 4.
Fig. 4.
Experimental results for LSFM-based imaging. (A) Comparison of beam profiles. The beam profile and the FWHM curve demonstrate that the deep design “butterfly” phase mask provides narrower illumination at the edges of the FOV. Note that the butterfly beam has a narrower profile at the edges, even when simulating a Gaussian beam with a similar waist. (B) Maximum intensity projection image of z-stack acquired from tissue-cleared mouse brain. The zoomed-in images demonstrate that the butterfly beam (red) exhibits a superior axial point spread function (XZ profile) compared to the Gaussian beam, particularly at the edges. The scale bar is 100 μm. (C) Improvement in axial resolution. The butterfly beam has a narrower axial profile in the edges of the FOV, whereas in the middle, the axial profile is comparable to that of the Gaussian beam.
See this image and copyright information in PMC

Similar articles

See all similar articles

References

    1. So PTC, Dong CY, Masters BR, Berland KM. Two-photon excitation fluorescence microscopy. Annu Rev Biomed Eng. 2000;2(1):399–429. - PubMed
    1. Olarte OE, Andilla J, Gualda EJ, Loza-Alvarez P. Light-sheet microscopy: A tutorial. Adv Opt Photon. 2018;10:111–179.
    1. Shechtman Y. Recent advances in point spread function engineering and related computational microscopy approaches: From one viewpoint. Biophys Rev. 2020;12(6):1303–1309. - PMC - PubMed
    1. Saxena M, Eluru G, Gorthi SS. Structured illumination microscopy. Adv Opt Photon. 2015;7(2):241–275.
    1. Vicidomini G, Bianchini P, Diaspro A. STED super-resolved microscopy. Nat Methods. 2018;15(3):173–182. - PubMed

LinkOut - more resources

  NODES
Experiments 1
HOME 2
innovation 1
iOS 1
Javascript 1
Note 1
os 31
text 12
Training 1
twitter 2
web 4