Real-time in vivo 3D fluorescence sectioning microscopy
Wide-field fluorescence microscopy is an imaging technique commonly used by researchers and clinicians for investigating rapid biological events occurring in live samples. However, standard wide-field microscopes lack the optical sectioning capabilities that would enable the capture of only in-focus information. Out-of-focus light and optical system aberrations lead to blurred images, limiting the use of conventional wide-field fluorescence microscopy in resolving volumetric biological samples. Although image postprocessing techniques such as deconvolution can improve image quality1, fluorescent light from other focal planes and aberrations still impair image resolution and contrast, a problem characteristic of wide-field fluorescence microscopy.
Optically sectioned images are the basis for creating three-dimensional images of various biological tissues. In short, a 3D image is a stack of 2D images. Optical sectioning techniques reject out-of-focus light from other sections by suppressing out-of-focus (background) signals while enhancing the desired in-focus (low spatial frequency and high spatial frequency) information. The most commonly used optical sectioning modality in biomedical imaging is based on confocal microscopy, which achieves excellent rejection of background light by including a physical pinhole in front of the detector to block any out-of-focus signal1, 2. However, the challenge for 3D confocal microscopy is its dependence on point-by-point scanning, which not only increases acquisition time in proportion to the number of desired voxels (i.e., the product of 3D space and bandwidth) for a high-definition image but also increases the chance of photobleaching during scanning. Alternative approaches to confocal microscopy, such as structured illumination techniques that project a pattern onto a sample, can reduce acquisition time, but they often require complicated mechanical scanning in the axial direction and still take considerable time to complete a 3D scan.
In this article, we present an in vivo, mechanical-scan-free, high-resolution, wide-field optical sectioning microscope (Fig. 1) for fast 3D imaging of biological specimens, whereby image contrast and axial scanning are achieved using a commercially available digital micromirror device (DMD, 1920×1080 mirrors, Texas Instruments) and a focus tunable lens (FTL), respectively. The pattern displayed on the DMD, which is placed in the illumination path, structures the illumination, and the variability of the FTL (a liquid lens consisting of a membrane and optical grade fluid) changes the focal length of the microscopy system. The proposed imaging modality dispenses with mechanical scanning and acquires images at different depths with constant magnification and contrast when the FTL is placed in the detection path and arranged in an object-space telecentric configuration. The periodic grid pattern displayed on the DMD and projected onto the sample enables the extraction of in-focus information and suppresses the out-of-focus information coming from other sections. Therefore, our approach does not disturb live samples because moving parts have been eliminated, and the system can provide in vivo 2D optically sectioned images of a biological sample at a speed close to that of video.
The design and implementation of the system are experimentally demonstrated here through in vivo 3D imaging of a mixture of pollen grains (Fig. 2) on a slide and an in vivo sample of Caenorhabditis elegans (Fig. 3). As shown in Fig. 2, the proposed imaging modality enhances in-focus information, as demonstrated by imaging an autofluorescent pollen sample. Figures 2(a) and 2(b) show images of the mixed pollen grains (Carolina, USA) at two different depths under standard epi-illumination. The out-of-focus light results in a haze that is evident in the images at both depths. Figures 2(c) and 2(d) are zoomed-in pollen images under structured illumination at depths corresponding to 2(a) and 2(b). Figures 2(e) and 2(f) show the final images after the use of our image postprocessing technique. Figure 2(g) shows the difference in cross-sectional profiles at Δz=3 μm between the signal-to-background ratios under wide-field uniform illumination and those under structured illumination. The haze (i.e., background) coming from out-of-focus light has been suppressed significantly, so that fine features previously shown in low contrast become visible due to the improved signal-to-background ratio.
To demonstrate the in vivo imaging capability of the microscopy system, we performed an experiment to observe the growth cones of live autofluorescent transgenic C. elegans worms. Worms paralyzed by levamisole in a microwell plate were used for in vivo imaging3. Figure 3 shows in vivo images of the growth cones of the worm taken with our proposed modality. The images of in vivo C. elegans in Figs. 3(a1)-3(a3) were taken at different depths by applying different input currents to the FTL and acquired with uniform illumination, i.e., while a uniform pattern was displayed on the DMD. However, out-of-focus background is significantly visible in all these sections because the fluorescent light emitted from other sections of the thick sample causes blurring. Figures 3(c) and 3(d) are zoomed-in images of the areas highlighted with boxes in Figs. 3(a2) and 3(b2) under wide-field uniform illumination and structured illumination at the corresponding depths. As the hazy background has been suppressed significantly by the structured illumination, the low-contrast features of the anvil-shaped growth cones are clearly visible in Fig. 3(d). A video showing a postprocessed axial scan obtained using the FTL for an entire live worm can be viewed in Ref. .
In summary, an in vivo multiplane optical sectioning modality based on an epi-illumination configuration without a mechanical stage has been developed and demonstrated in this article. Constant image contrast and high acquisition speed are achieved using a programmable pattern displayed on a DMD to modulate illumination and an FTL placed in the imaging path to change the focal length of the system for axial scanning. Our proposed microscope can also review a sample before acquisition by simply adjusting the input current to the FTL. Optically sectioned images can be acquired at close to video rate while effectively suppressing undesirable background. The amount of spherical aberration may be exacerbated if the axial scanning range is increased3 and if the system focal plane is shifted away from the nominal focal plane of the objective lens. However, this depth-induced aberration can be reduced by using an additional FTL, a wavefront correction device or an image postprocessing technique such as deconvolution to further improve the optically sectioned image quality.
Figure 1. A schematic setup for structured illumination microscopy. The pattern on the digital mirror device (DMD) reflects light from the laser into the biological sample, passing through a dichroic beam splitter (square with diagonal line), focus tunable lens (FTL) and relay lens. The fluorescent signal being collected by the Objective lens is recorded by a Charge-coupled Device (CCD) camera.
Figure 2. (a, b) Standard wide-field images of mixed autofluorescent pollen grains at Δz=3 μm and 18 μm, respectively. (c1, c2) Zoomed-in structured illumination images of the upper pollen grain at Δz=3 μm and 18 μm, respectively. (d1, d2) Zoomed-in structured illumination images of the mixed pollen grain at Δz=3 μm and 18 μm, respectively. (e, f) Corresponding images processed using a hybrid pairwise imaging algorithm. (g) Intensity cross-section at Δz=3 μm along the solid and dashed lines of the uniformly illuminated and processed images (a and e), respectively.
Figure 3. (a1-a3) Standard wide-field images of a transgenic C. elegans worm at three different depths. (b1-b3) Images processed using the hybrid imaging algorithm at the corresponding three depths. (c, d) Zoomed-in views of the dashed-box (blue) and solid-box (green) regions of (a2) and (b2), respectively. Scale bars are 10 μm in length.
1. Yuan Luo, Vijay Raj Singh, Dipanjan Bhattacharya, Elijah Y. S. Yew, Jui-Chang Tsai, Sung-Liang Yu, Hsi-Hsun Chen, Jau-Min Wong, Paul Matsudaira, Peter T. C. So and George Barbastathis. (2014). Talbot holographic illumination non-scanning (THIN) fluorescence microscopy, Laser and Photonics Reviews, 8(5), L71- L75. DOI: 10.1002/lpor.201400053
2. Chen Yen Lin, Wei-Tang Lin, Hsi-Hsun Chen, Jau-Min Wong, Vijay Raj Singh and Yuan Luo. (2016). Talbot multi-focal holographic fluorescence endoscopy for optically sectioned imaging, Optics Letters. 41(2), 344-347. DOI: https://doi.org/10.1364/OL.41.000344.
3. Chen-Yen Lin, Wei-Hsin Lin, Ju-Hsuan Chien, Jui-Chang Tsai, and Yuan Luo*. (2016). In Vivo volumetric fluorescence sectioning microscopy with mechanical-scan-free hybrid illumination imaging, Biomedical Optics Express, 7(10), 3968-3978. DOI: https://doi.org/10.1364/BOE.7.003968
Associate Professor Yuan Luo
Institute of Medical Device and Imaging & Molecular Imaging Center
Chen Yen Lin
Institute of Medical Device and Imaging & Molecular Imaging Center
Graduate Institute of Photonics and Optoelectronics
Figure 1. A schematic setup for structured illumination microscopy.