Selective Plane Illumination Microscopy (SPIM) is attractive for its ability to acquire 3D images with high 3D spatial resolution, good optical sectioning capability and high imaging speed. However, tradeoffs have to be made when a large field of view (FOV) is required, results in lower axial resolution or worse optical sectioning capability. Here, we present a novel method for 3D imaging by SPIM that is capable to maintain its high 3D spatial resolution and good optical sectioning capability within a large FOV. Instead of trying to generate a large and uniformly thick excitation light sheet, the method tiles a relative small light sheet quickly to multiple positions within the image plane by defocusing the excitation beam used to create the light sheet, and takes one additional image at each position, so that a large FOV can be imaged by repeating this process and stitching all images together. By implementing this method, light sheets with thin thickness and good excitation light confinement can be used for SPIM imaging with slightly compromised imaging speed. The method was investigated through both numerical simulation and experiments, and the imaging performance was demonstrated by imaging fluorescent particles embedded in agarose gel and live C. elegans embryos.
© 2015 Optical Society of America
Selective plane illumination microscopy (SPIM) is becoming increasingly attractive for 3D live fluorescence imaging in biological research [1–8]. In comparison with conventional fluorescence imaging techniques based on the epi-illumination configuration, such as confocal microscopy, SPIM gains its advantages by separating the sample excitation from the fluorescence detection using two objectives with optical axis orthogonal to each other, so that the excitation light is confined near the detection focal plane rather than passing through the entire specimen. By doing so, SPIM improves both the axial resolution and the optical sectioning capability, reduces the photobleaching and phototoxic effect, while retains the high imaging speed as that of the wide field microscopy at the same time.
Volumetric imaging by SPIM is usually performed by scanning the excitation light sheet through the specimen, and taking a series of 2D images at different axial positions sequentially. The imaging performance of SPIM highly depends on the geometry and the intensity profile of the excitation light sheet. Generally, the thinner the excitation light sheet, the higher the theoretical axial resolution, and the closer the excitation light is confined near the detection focal plane, the better the optical sectioning capability and the easier to achieve the theoretical resolution in practice. Meanwhile, the FOV of SPIM is determined by the size of the excitation light sheet, over which the thickness of the light sheet is reasonably uniform. In SPIM, the excitation light sheet is usually a virtual light sheet created by sweeping an excitation beam or dithering an excitation beam array through the detection focal plane using a scanning device such as a galvanometer scanner [2, 4, 6]. The width of the light sheet is usually controlled by the scanning distance of the excitation beam, and it is only limited by the dimension of the optical elements in the excitation optical path, while the length of the excitation light sheet is limited by the Rayleigh length of the excitation beam or the beam array used to create the light sheet. Therefore, the FOV of SPIM is mainly limited by the length of the excitation beam used to create the excitation light sheet.
In order to increase the FOV of SPIM, the length of the excitation beam used to create the light sheet needs to be extended and it is usually done by limiting the wave vector difference of the excitation light wavefront along the beam propagation direction. For instance, the length of a Gaussian beam can be increased by reducing the excitation numerical aperture (NAexc), and a so called nondiffracting beam, such as a Bessel beam [3,9], can be obtained by limiting excitation light wavefront wave vector difference using an annular mask. However, tradeoffs exist between the excitation beam length, thickness and the excitation light confinement. The excitation beam has to be thicker, or the excitation light is less confined as the beam length increases, result in lower axial resolution or worse optical sectioning capability. Therefore, how to create a thin and uniformly thick excitation light sheet, which is also big enough to cover the FOV becomes the key problem of SPIM. Different methods have been developed to create such a light sheet in SPIM, such as the Gaussian light sheet , the Bessel light sheet [3, 4], and the recently developed lattice light sheet . Nevertheless, the tradeoffs remain valid for all kinds of light sheets no matter what method is used, and the problem is especially challenging when a submicron thin light sheet is required to image a FOV of dozens of microns or larger.
Recently, Dean and Fiolka presented another method to create a large and uniform light sheet , by which the excitation light focus is scanned along the light propagation direction using a focus tunable lens, so that a long and uniform virtual excitation beam can be created, and a large virtual excitation light sheet can be further generated by scanning the virtual excitation beam. However, there are two limitations with this approach. First, the light sheet created by this method has a poor confinement on the excitation light due to the integration of the excitation beam tails during scanning. Therefore, the optical sectioning capability is poor, and it has to be used together with confocal detection to reject the fluorescence background [10–12]. Second, since the excitation light sheet produced is a virtual light sheet, the focus has to be scanned through the whole FOV during one camera exposure cycle. Thus, the actual exposure time at any specific position within the FOV is extremely short, and a relative high excitation power has to be used to generate enough fluorescence signal, which leads to stronger phototoxic effect.
In this research, we propose an alternative solution to increase the FOV of SPIM without increasing the light sheet thickness or loosing the excitation light confinement. As shown in Fig. 1, instead of using a larger excitation light sheet, a large FOV can be imaged by tiling the light sheet within the detection focal plane along the light sheet length direction (Y direction), and taking an image at each position. After that, the image of the whole FOV can be reconstructed by combining all images together. Despite the reduced imaging speed by this operation, the benefit is significant. A thinner light sheet with better excitation light confinement can be used to image the same specimen, so that higher axial resolution and better optical sectioning capability can be obtained.
In order to implement this method in practice, either the specimen or the excitation beam has to be shifted quickly along the beam direction during imaging. We chose to move the excitation beam by defocusing the beam using a phase only spatial light modulate (SLM), because it is too slow to move either the specimen or the excitation objective. Moreover, due to the physical constrain between the excitation objective and the detection objective in SPIM, the excitation NA in SPIM is usually smaller than 0.6, hence the excitation beam can be defocused for a few hundred microns with very little influence on its lateral intensity profile. In this research, we investigated the method via both numerical simulation and experiments, and the imaging performance was demonstrated by imaging fluorescent particles embedded in agarose gel, and fluorescently labeled C. elegans embryos.
The method was first investigated via numerical simulation. The excitation light wavefront modulation was performed at the pupil plane of the excitation objective. In order to defocus the excitation beam, a spherical phase map was added to the excitation light wavefront. It was first applied to a Gaussian beam and a Bessel beam as shown in Fig. 2(a), 2(b) and Fig. 2(c), 2(d). For the Gaussian beam, a shorter but thinner beam (NAexc = 0.35) can be used to image the same FOV as that of a longer beam (NAexc = 0.2) by shifting the shorter beam to three positions. Although the imaging speed of using the shorter beam is three times slower, the thickness of the light sheet created by the shorter Gaussian beam is almost twice thinner than the light sheet created by the longer Gaussian beam, as seen from the cross section intensity profile of the beams and the light sheets created by scanning the corresponding beam (Fig. 2(b)). For the Bessel beam, a shorter beam (NAOD = 0.4, NAID = 0.22) but with significantly less side lobes can be used to image the same FOV as that of a longer Bessel beam (NAOD = 0.4, NAID = 0.35) with the same central peak thickness. In comparison, the central peak of the shorter Bessel beam carries ~40% of the total excitation energy, while the central peak of the longer Bessel beam only carries ~10% of the total energy. Therefore, the light sheet created by the shorter Bessel beam offers much better optical sectioning capability despite the slower imaging speed.
The lattice light sheet created by dithering a coherent Bessel beam array is another kind of light sheets developed recently. The lattice light sheet is capable to maintain a uniform thickness over a long distance as the Bessel light sheet does, while offers better excitation light confinement due to the destructive interference of the Bessel beam side lobes. The coherent Bessel beam array can be generated by shaping the intensity profile of excitation light wavefront intensity profile, and it can also be defocused by adding a spherical phase map to the wavefront as show in Fig. 2(e). Again, due to the relaxed length requirement, a shorter coherent Bessel beam array with better excitation light confinement can be used to image the same FOV (Fig. 2(f)).
There are two ways to implement the proposed method for 3D imaging by SPIM in practice. One way is to take a complete 3D image stack at each beam position, and then shift the excitation beam to the next position and repeat the process until the whole specimen is imaged. The other way is to shift the excitation beam within the image plane and take one image at each position until the whole plane is imaged before imaging the next image plane. The same process is repeated at all image planes until the whole specimen is imaged. In comparison, the second way is more suit for imaging fast biological activities due to the shorter time delay between different areas on the same image plane, and faster wavefront modulation speed is required to operate in this way. Unfortunately, a high resolution continuous phase SLM is usually expensive and the modulation speed is limited to sub-hundred Hz, so that it takes at least 10-20 milliseconds to apply a new phase map and shift the excitation beam to a new position. On the contrary, a binary phase SLM is cheaper and the modulation speed is much faster, which allows a new phase map to be applied in less than a millisecond .
Therefore, we further investigated the possibility to shift the excitation beam using a binary phase SLM. The binary phase map shown in Fig. 3a was generated by resetting the pixel values of the continuous spherical phase map shown in Fig. 2(a). Values between 0 and π were set to 0, and values between π and 2 π were set to π. Instead of moving the excitation beam to a new position, multiple orders of the excitation beam were generated by applying the binary phase map, and these orders were distributed on both sides of the original excitation beam position. To avoid the interference between different orders, the applied binary phase map was shifted off-axis to the excitation light as shown in Fig. 3(b)–3(d). As a result, different diffraction orders of the excitation beam were separated not only in the axial direction but also in the lateral direction as shown Fig. 3(b)–3(d). Thereafter, undesired diffraction orders can be blocked by putting different diffraction orders far enough from each other in the lateral direction and placing an optical iris or a slit at the corresponding imaging plane after the SLM.
Next, the method was verified experimentally. The configuration of the experimental systems is shown in Fig. 4. A 488 nm CW laser source (Coherent, Sapphire, 200 mW) was used for linear excitation. The laser beam was collimated and expanded to 1.5 mm diameter before sent to an acousto-optical tunable filter (AOTF, AA Opto-Electronic, nAOTFnC-400.650-TN) used to modulate the laser beam intensity. The laser beam was expanded to 15 mm diameter and sent to the SLM assembly after passing the AOTF (focal length L1 = 25 mm, L2 = 250 mm). The SLM assembly consisted of a polarizing beam splitter cube, a half-wave plate and a 1280 × 1024 pixel binary SLM (Forth Dimension, SXGA-3DM). The three components worked together as a binary phase modulator to modulate the phase of the excitation light wavefront. After the phase modulation, the laser beam was send to the beam scanning assembly consisted of a pair of galvonometer mirrors (Cambridge Technology, 6215HP-1HB) for beam scanning, while both galvo mirrors were conjugated to the SLM through relay lenses L3 = 250 mm, L4 = 100 mm and L5 = 125 mm, L6 = 125 mm. An optical iris was placed at the image plane of L3 to block the undesired diffraction orders of the excitation light generated by the binary SLM. The laser beam was further related to a photomask conjugated to both of the galvo mirrors through relay lenses L7 = 50 mm and L8 = 100 mm to shape the laser beam intensity profile, and finally conjugated to the rear pupil of the excitation objective (Nikon, CFI Apo 40XW NIR) through relay lenses L9 = 175 mm and L10 = 175 mm. The detection objective (Nikon, CFI Apo 40XW NIR) was placed orthogonal to excitation objective and the emitted fluorescence was collected through the detection objective and imaged onto a sCMOS detection camera (Hamamatsu, Orca Flash 4.0) with tube lens L11 = 225 mm. The imaging chamber was filled with imaging buffer during imaging, and both objectives and the specimen were emerged in the imaging buffer during imaging.
The experimental system was first characterized using Alexa 488 dye solution as the imaging buffer, so that the excitation beam can be visualized. By choosing the excitation NAOD = 0.4, NAID = 0.16 with the photomask and applying three phase maps shown in Fig. 5(a)–5(c) on the SLM, Bessel excitation beams shown in Fig. 5(d)–5(f) were observed. By adjusting the position and the size of the optical iris after the lens L3, only the beam on the top is remained as shown in Fig. 5(g)–5(i) and the rest orders were blocked. In comparison, in order to image the same FOV as that of shifting the beam to the three positions, either a twice thicker Gaussian beam (NAexc = 0.2), or Bessel beams (NAOD = 0.4, NAID = 0.32 and NAOD = 0.4, NAID = 0.36) with significantly more side lobes have to be used. (Fig. 5(l)–5(n)).
The imaging performance of the proposed method was first characterized using 100 nm yellow-green fluorescent particles embedded in 1.5% agarose gel. The same Bessel beam (NAOD = 0.4, NAID = 0.16) was used to create a Bessel excitation light sheet and image the specimen. The light sheet created by scanning the beam confines ~70% of total energy within the central light sheet due to the low side lobe energy of the selected beam. Figure 6(a)–6(c) show the axial maximum projection of three 3D image stacks collected for the same FOV with three different excitation beam positions (left, middle and right). Figure 6(d)–6(f) show the corresponding deconvolved image results. Figure 6(h)–6(m) show the same view of the selected fluorescent particles with higher magnification before and after deconvolution, and also the axial intensity plot of each selected particle.
As seen in Fig. 6, at each beam position, a higher axial resolution was obtained in the corresponding sub area compared to that of the rest areas due to the shorter excitation beam length compared to the FOV. Nevertheless, a uniform axial resolution of ~500 nm can be obtained within the whole FOV by tiling the excitation light sheet at three positions. Meanwhile, due to the good excitation light confinement, the experimental axial resolution is almost identical to the theoretical axial resolution of ~470 nm using the selected excitation beam and detection NA by sheet scan. It is unusual to obtain the theoretically axial resolution in Bessel SPIM by sheet scan in practice due to the strong side lobe energy carried of most long Bessel beams.
Following this idea, even higher axial resolution can be obtained in a larger FOV just by tiling a thinner light sheets with equal or better excitation light confinement at more positions, but of course with the compromise of the imaging speed. Furthermore, due to the limited experimental resources, a 0.8 NA detection objective was used in the system, while the spatial resolution can be easily improved to ~240 nm laterally and ~360 nm axially with the same excitation beam by using a 1.1 NA detection objective (Nikon CFI Apo LWD 25XW) as reported previously.
The imaging performance was further demonstrated by imaging a live C. elegans embryo (OD95) labeling cell membrane at three fold stage. There are roughly 550 cells within an embryo at this stage, and the cellular structure of such an embryo is very dense. It also causes strong optical aberration and scattering. Good optical sectioning capability is necessary to obtain 3D images with submicron axial resolution and high signal to noise ratio (SNR) for this kind of specimens. The same excitation beam was shifted between three positions to image the whole embryo, and the lateral maximum projection of embryo is shown in Fig. 7(a). To better reveal the internal cellular structure of the embryo, the cross section views of the embryo along the horizontal dash line, extracted from the images taken at different beam positions, are shown in Fig. 7(b)–7(d), and the cross section views of the embryo along the three vertical dash lines are shown in Fig. 7(e)–7(g). As expected, by shifting the excitation beam and taking extra images, the whole embryo was able to be imaged with ~500 nm axial resolution and high SNR due to the use of a thin excitation beam with good excitation light confinement, in spite of the complex structure and the optical aberration of the embryo.
We proposed and demonstrated a novel method for 3D imaging by SPIM to achieve high 3D spatial resolution and good optical sectioning capability together with a large FOV in this research. In comparison with the previous solutions towards this goal, instead of trying to extend the size of the light sheet to make it large enough to cover the entire FOV, it tiles a relative small excitation light sheet to multiple positions by defocusing the excitation beam and taking multiple images of the same image plane. By doing so, light sheets with thinner thickness and better excitation light confinement can be used to image a large specimen with a compromised imaging speed.
The proposed methods is valuable for two reasons. First, it is very difficult to create a light sheet with thin thickness, good excitation light confinement and large size at the same time. Although different kinds of light sheets have been introduced to provide a solution to this problem, it is still general that either the light sheet becomes thicker or the excitation light confinement, i.e. optical sectioning capability, becomes worse as the light sheet size increases, no matter what method is used to create the excitation light sheet. There is unlikely to be a perfect solution either due to the diffraction of light. Therefore, the proposed method provide an alternate strategy to solve the problem when none of the axial resolution, optical sectioning capability and the FOV can be compromised. Second, the proposed method remains effective for any kind of SPIM light sheet, which means higher axial resolution and better optical sectioning capability can always be obtained from the same specimen by using the same kind of light sheet of smaller size, but with thinner thickness or better excitation light confinement.
There are also several problems need to be taken into consideration during the implementation of the method. First, the imaging speed is decreased because multiple images have to be taken from each image plane. Fortunately, the imaging speed of the latest sCMOS camera is as fast as 100 frames/sec, and it can be increased to a few hundred frames/sec if only a part of the image sensor is used. Therefore, the actual imaging speed of the method can still reach dozens of planes per second if necessary, which is fast enough for most applications. Second, photobleaching and phototoxicity of the proposed method are expected to be higher than regular SPIM imaging for the same reason. Nevertheless, in comparison with other approaches to obtain higher axial resolution and better optical sectioning capability, such as using a virtual confocal slit in image detection [10, 12], or combining the structured illumination microscopy (SIM) with SPIM , which also requires multiple raw images, the proposed method is more efficient in spending the limited photon budget because it uses a thinner light sheet with better excitation light confinement. And also, it improves the SNR by generating less background fluorescence rather than trying to remove the background after generating it. Therefore, the photobelaching and phototoxicity of the proposed method should be moderate and a more detailed study is to be done in future research. Furthermore, the proposed method can also be combined with the previous methods, such as confocal detection and SIM to further improve the imaging performance. Finally, because of the time delay between different areas of the same image plane, more accurate image registration is preferred compared to the direct stitching of the acquired images.
Besides the discussed benefits, the ability to move the excitation beam freely within the image plane could also be useful for many other applications. For instance, in some applications, only a few separated sub areas within a large FOV are of interest. It is extremely inefficient to create a large light sheet to cover the whole FOV while only a few sub areas need to be imaged with high 3D resolution. The ability to send a thin excitation light sheet to only the areas of interest within a large FOV could bring substantial benefit in such applications.
The control software of the experimental system was licensed from Howard Hughes Medical Institute, Janelia Farm Research Campus. The author thanks Dr. David Matus for supplying the C. elegans strain used in the research.
References and links
1. J. Huisken, J. Swoger, F. Del Bene, J. Wittbrodt, and E. H. K. Stelzer, “Optical sectioning deep inside live embryos by selective plane illumination microscopy,” Science 305(5686), 1007–1009 (2004). [CrossRef] [PubMed]
2. P. J. Keller, A. D. Schmidt, J. Wittbrodt, and E. H. K. Stelzer, “Reconstruction of zebrafish early embryonic development by scanned light sheet microscopy,” Science 322(5904), 1065–1069 (2008). [CrossRef] [PubMed]
3. T. A. Planchon, L. Gao, D. E. Milkie, M. W. Davidson, J. A. Galbraith, C. G. Galbraith, and E. Betzig, “Rapid three-dimensional isotropic imaging of living cells using Bessel beam plane illumination,” Nat. Methods 8(5), 417–423 (2011). [CrossRef] [PubMed]
4. L. Gao, L. Shao, C. D. Higgins, J. S. Poulton, M. Peifer, M. W. Davidson, X. Wu, B. Goldstein, and E. Betzig, “Noninvasive imaging beyond the diffraction limit of 3D dynamics in thickly fluorescent specimens,” Cell 151(6), 1370–1385 (2012). [CrossRef] [PubMed]
5. Y. Wu, A. Ghitani, R. Christensen, A. Santella, Z. Du, G. Rondeau, Z. Bao, D. Colón-Ramos, and H. Shroff, “Inverted selective plane illumination microscopy (iSPIM) enables coupled cell identity lineaging and neurodevelopmental imaging in Caenorhabditis elegans,” Proc. Natl. Acad. Sci. U.S.A. 108(43), 17708–17713 (2011). [CrossRef] [PubMed]
6. B.-C. Chen, W. R. Legant, K. Wang, L. Shao, D. E. Milkie, M. W. Davidson, C. Janetopoulos, X. S. Wu, J. A. Hammer 3rd, Z. Liu, B. P. English, Y. Mimori-Kiyosue, D. P. Romero, A. T. Ritter, J. Lippincott-Schwartz, L. Fritz-Laylin, R. D. Mullins, D. M. Mitchell, J. N. Bembenek, A. C. Reymann, R. Böhme, S. W. Grill, J. T. Wang, G. Seydoux, U. S. Tulu, D. P. Kiehart, and E. Betzig, “Lattice light-sheet microscopy: Imaging molecules to embryos at high spatiotemporal resolution,” Science 346(6208), 1257998 (2014). [CrossRef] [PubMed]
8. R. Tomer, K. Khairy, F. Amat, and P. J. Keller, “Quantitative high-speed imaging of entire developing embryos with simultaneous multiview light-sheet microscopy,” Nat. Methods 9(7), 755–763 (2012). [CrossRef] [PubMed]
9. J. Durnin, “Exact solutions for nondiffracting beams. I. The scalar theory,” J. Opt. Soc. Am. 4(4), 651–654 (1987). [CrossRef]
13. L. Shao, P. Kner, E. H. Rego, and M. G. L. Gustafsson, “Super-resolution 3D microscopy of live whole cells using structured illumination,” Nat. Methods 8(12), 1044–1046 (2011). [CrossRef] [PubMed]