Open Access
Add to BibSonomy
Abstract
Line confocal (LC) microscopy is a fast 3D imaging technique, but its asymmetric detection slit limits resolution and optical sectioning. To address this, we propose the differential synthetic illumination (DSI) method based on multi-line detection to enhance the spatial resolution and optical sectioning capability of the LC system. The DSI method allows the imaging process to simultaneously accomplish on a single camera, which ensures the rapidity and stability of the imaging process. DSI-LC improves X- and Z-axis resolution by 1.28 and 1.26 times, respectively, and optical sectioning by 2.6 times compared to LC. Furthermore, the spatially resolved power and contrast are also demonstrated by imaging pollen, microtubule, and the fiber of the GFP fluorescence-labeled mouse brain. Finally, Video-rate imaging of zebrafish larval heart beating in a 665.6 × 332.8 µm2 field-of-view is achieved. DSI-LC provides a promising approach for 3D large-scale and functional imaging in vivo with improved resolution, contrast, and robustness.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Advances in optical imaging technology are promoting the development of biomedical research. Optical imaging of biological tissues, such as brain imaging in vivo, requires the recording of submicron fine structures and time-varying functional information in a strong scattering background [1,2]. To achieve high-quality biological tissue imaging, optical microscopy needs to have high spatiotemporal resolution and strong background suppression ability. Confocal laser scanning microscopy has been widely used for biomedical imaging due to its high spatial resolution and excellent optical sectioning ability [3,4]. However, single point scanning mode is time-consuming and not suitable for three-dimensional (3D) structural imaging of large scale [5] and fast time-varying functional imaging [6]. Spinning disk scanning improves the imaging speed of confocal microscopy significantly through multi-point parallel imaging [7]. However, signal crosstalk by parallel detection weakens its optical sectioning ability, leading to poor image quality [8]. Line confocal (LC) microscopy is another effective approach to speed up the confocal imaging process [9]. However, the unfocused dimension, equivalent to wide-field (WF) imaging, reduces spatial resolution and optical sectioning ability [10].
Structured illumination microscopy (SIM) with WF imaging mode offers fast high-contrast imaging [11,12]. It benefits from that the modulated signal loses its modulation characteristics during the defocusing process. However, strong scattering in thick-tissue imaging reduces the contrast of SIM imaging [13,14]. Recently, line-scanning SIM has been reported to address the issue of strong scattering [15,16]. However, the construction of modulating illumination by multiple scans is a time-consuming process. Given this problem, our group proposes a series of single-scan SIM approach to avoid repetitive data acquisition [17–19]. Although the methods mentioned above enhance the optical sectioning ability and imaging speed, SIM algorithms, such as square law detection [11] and high and low-frequency filtering [12], don't improve spatial resolution.
Light sheet microscopy [20], as an alternative optical sectioning technology, has been widely adopted by biomedical imaging. The light sheet only illuminates the sample within the focal depth of detection, and the fluorescence signal is recorded by WF imaging [21]. This method allows for imaging in the frame-rate limit of the camera. However, the heterogeneity of axial resolution reduces image quality due to the divergence of the illumination beam. Axial tile of scanning light sheet has been reported to solve this problem but slows down imaging speed [22]. Additionally, the configuration of horizontal illumination and vertical detection limits high-resolution imaging of large-scale samples due to the short work distance of high numerical aperture. Although several methods [23–25] have been used to improve the resolution of the light sheet, all of them slow down the imaging speed.
As mentioned above, previous methods engender trade-offs between spatial resolution, imaging speed, and optical sectioning ability. To address this issue, we introduce a difference method that has been employed in various imaging modes to improve spatial resolution and imaging contrast due to its simplicity [26–29]. Recently, the difference methods had been used to improve LC performance, but these methods used multiple scans or multiple detectors to obtain different point spread functions (PFS) [30–32]. The multi-scanning acquisition would reduce imaging speed. The usage of multiple detectors increases the system complexity, and the strict registration between multiple channels isn’t conducive to long-term stable imaging. More importantly, the previous usage only improved optical sectioning ability without improving spatial resolution.
In this work, we report a differential synthetic illumination (DSI) method based on multi-line detection to enhance the spatial resolution and optical sectioning capability of the LC system. Unlike the construction of different PSFs in detection, we synthesize different illuminations by using redundant information on a single camera in a single scan. The DSI method could enhance spatial resolution and optical sectioning ability at video rate simultaneously. The 3D imaging of fluorescent beads and the fluorescent plane show our improvements in spatial resolution and optical sectioning ability. We imaged pollen grain, cell microtubule, and Thy1-GFP transgenic mouse brain tissue to demonstrate the enhancement of spatial resolution and contrast. In addition, imaging the heart beating of larval zebrafish in 665.6 × 332.8 µm2 field-of-view at 24 Hz also demonstrated that our system has capability at video rate.
2. Materials and methods
2.1 DSI line-scanning imaging system
Figure 1 shows the DSI line confocal (DSI-LC) system configuration. We use a continuous laser with tunable power (488 nm, 100 mW, 0.7 mm diameter, Cobolt, Sweden) as a fluorescence excitation source. Then, a Kepler telescope, consisting of a pair of positive lenses (L1, f = 7.5 mm; L2, f = 200 mm, Thorlabs), is used to expand the laser beam. The collimated beam is one-dimensional focused into a linear beam by a cylindrical lens (CL, f = 100 mm, Thorlabs) and then reflected by a dichroic mirror (DM; ZT405/488/561rpc, Chroma, USA) to a uniaxial galvanometer mirror (GM; 6240 H, CTI). Finally, the linear beam passes through a scan lens (L3, f = 100 mm, Thorlabs), tube lens (L4, f = 100 mm, Thorlabs), mirror and objective lens (OL; NA 1.0, XLUMPLFLN 20XW, Olympus, Japan), and the GM conjugate with the pupil plane of the objective lens. For detection, the fluorescence signal from the sample was relayed to a scientific CMOS camera (sCMOS, ORCA-Flash 4.0, Hamamatsu, Japan) via an intermediate optical system, which was composed of OL, mirror, L3, L4, GM, DM, emission filter (EM; ZET405/488/561 m, Chroma, USA), and tube lens (TL; f = 180 mm, Olympus). The sCMOS possessed 2,048 × 2,048 pixels, and each pixel size was 6.5 × 6.5 µm2. The central lines of the camera are operated in sub-array mode as a multi-line array detector, and the region of interest was 2,048 × 8 pixels. The center lines of the camera should be aligned with the linear beam. The camera is fixed to a manual triaxial translation stage (X-axis, Y-axis: SPYD65R; Z-axis: OMHB90B, Wuhan Red Star Yang Science and Technology Co., Ltd, China) for precise adjustment. The optical power density on the sample is 464.6 W/cm2.
Fig. 1. Imaging system and principle. (a) Schematic of DSI-LC system. L1–L4, lens; CL, cylindrical lens; DM, dichroic mirror; GM, galvanometer mirror; M, mirror; OL, objective lens; EM, emission filter; TL, tube lens. (b) Schematic of synthetic illumination based on multi-line detection. Blue lines represent the signals detected by the corresponding lines of the camera subarray. (c) Synchronization sequence among the exposure and clock signals (CS) of the camera and the moving of GM in the beam-scanning mode. (d) Synchronization sequence between the stage moving and camera exposure in the stage-scanning mode.
2.2 System synchronization
To facilitate different usage scenarios, our system includes two scanning modes of beam scanning and stage scanning. For fast imaging in a single field-of-view (FOV), the GM is employed to scan the beam. The specimen was motionless, and the linear beam was scanned by the GM in beam scanning mode. Due to the de-scan effect of the GM, the collected fluorescent signals arrive at the same position of the detector each time. The camera works in trigger output mode to provide a clock signal for the GM to generate the synchronization sequence between the continuous exposure of the camera and the continuous sweep of the GM as shown in Figs. 1(c). The motion trail of the GM is controlled by a homemade voltage curve. The maximum frame rate of the camera is 25,655 fps in the sub-array mode at 2,048 × 8 pixels. The GM scanned 1,024 pixels per cycle for video-rate imaging, and the step size is the pixel size. The imaging speed of beam scanning is 24 Hz in 665.6 × 332.8 µm2 FOV. For large-scale 3D imaging, a high-accuracy 3D stage (X: ABL20020, Y: ANT130, Z: AVL125, Aerotech, USA) was used to translate the sample in place of the beam scanning of the GM. The synchronization sequence of the camera and stage is shown in Figs. 1(d). The camera works in start trigger mode. A trigger signal is given to start the camera when the stage moves to the appointed position. To stay in sync, the exposure speed of the camera should be equal to the moving speed of the stage.
2.3 Principle of DSI
The previous difference method [26] is an effective means to improve the confocal resolution by subtracting different PSFs. It results in the high-frequency component of the system is enhanced and the low-frequency component is suppressed. Their different PSFs were constructed by adjusting the size of the pinhole or slit in the detection terminal, which is completed at different times or on multiple detectors. Here, we construct different PSFs by the illumination synthesis method in a single camera simultaneously. The schematic of synthetic illumination based on multi-line detection is shown in Fig. 1(b). The equation of subtraction imaging Isub can be expressed as [26]:
where ILC is LC imaging, ITDI is the time delay integral (TDI) imaging, $\alpha $ is the subtraction coefficient. The value of $\alpha $ is 0.8–1 in our system to avoid signal loss in the DSI process. The intensity values of the image, ILC, and ITDI, are also equal to the convolution of the corresponding PSF with the specimen. Equation (1) would be rewritten as: where o is 3D object distribution, $\otimes $ is convolution operation, ${h_{LC}}$ is the PSF of LC, hTDI is the PSF of TDI. In our system, the PSF of LC hLC can be written as [10]:3. Results
3.1 Measurement of spatial resolution and optical sectioning
To demonstrate spatial resolution enhancement by the DSI method, the 200 nm fluorescent beads (FluoSpheres Carboxylate-Modified Microspheres, ThermoFisher, USA) were imaged. The diluted solution of fluorescent beads was uniformly smeared on the slide. The mounted fluorescent beads were randomly distributed in three dimensions. We imaged 50 layers with a 0.4 µm axial scanning step to reconstruct the 3D distribution of fluorescent beads. To calculate the resolution, the intensity profiles of the beads were fitted by the Gaussian function. Figure 2(a) shows the X-axis resolution data, and the reconstrued images of the X-Y plane about LC and DSI-LC were in the upper left corner of Fig. 2(a). The full width at half maximum (FWHM) of LC and DSI-LC are 0.41 ± 0.05 µm and 0.32 ± 0.03 µm (mean ± s.d., beads number, $n$=5), respectively. The X-axis resolution of the DSI-LC method is improved by 1.28 times that LC method. The Y-axis resolutions of LC and DSI-LC were both 0.48 ± 0.05 µm in Fig. 2 (b). The 0.325 µm pixel size in the focal plane leads to undersampling under the sampling process. Therefore, the lateral FWHM was slightly larger than the diffraction limit. The X-axis resolution was better than the Y-axis resolution because the linear beam was only focused on the X-axis. In other words, the X-axis data had a confocal effect, and the Y-axis corresponded to wide-field imaging. Hence, the X-axis resolution was improved based on the DSI method, while the Y-axis resolution remained unchanged. The reconstrued images of the Y-Z plane about LC and DSI-LC were in the upper left corner of Fig. 2(c), and the Z-axis resolutions of LC and DSI-LC were 2.34± 0.1 µm and 1.86 ± 0.08 µm, respectively. The Z-axis resolution of the DSI-LC method is improved by 1.26 times that LC method. Since the objective is without a correction collar, the aberration introduced by the coverslip leads to the Z-axis resolution being worse than the diffraction limit. To further verify the optical sectioning enhancement of the DSI method, we also imaged a fluorescent plane and axially scanned 100 layers with a 0.2 µm step. The undiluted solution of fluorescent beads was uniformly smeared on the slide to produce the fluorescent plane. The defocus curve of the fluorescent plane quantitatively reveals the optical sectioning ability of the imaging system as shown in Fig. 2(d). The average intensity from the 20× 20-pixel2 region of interest was used to plot the defocus curve. The FWHMs of LC and DSI-LC are 5.2 and 2 µm, respectively. The optical sectioning ability of the DSI-LC method is 2.6 times better than that of the LC method. The experimental results demonstrate that the DSI method improved both the spatial resolution and optical sectioning ability of the LC system.
Fig. 2. System performance measurements of LC and DSI-LC systems. Resolutions in the X- (a), Y- (b), and Z-directions (c) using 200 nm fluorescent beads, as well as optical sectioning performance (d) using fluorescence plane. Color values are the FWHMs of the corresponding color curves.
3.2 Pollen grain and cell microtubule imaging
To further show the resolution and optical sectioning enhancement of our method, we imaged a pollen grain and cellular microtubule. The pollen grain labeled by hematoxylin (Mixed Pollen Grains, Life Technologies, USA) was axially scanned 20 layers with a 2 µm step. Figures 3(a)–(c) provides a comparison of imaging quality among the WF, LC, and DSI-LC images. The three imaging modes share the common objective lens, tube lens, and camera. The WF image was acquired in full-frame mode, and the images of LC and DSI-LC were acquired in sub-array mode. The central line data was reconstructed LC image. DSI-LC image was reconstructed from the redundant information on the camera based on the DSI method. Figures 3(a)–(c) show the single-layer images of WF, LC, and DSI-LC at Z = 20 µm, respectively. The images of the WF and LC are blurry, while the DSI-LC image inhibits the background signal better than other methods. The contrast of Figs. 3(a)–(c) are 0.7, 4.3 and 16.2. Figure 3(d) shows the normalized intensity profiles along the corresponding colored solid lines on the small spikes in Figs. 3 (a)–(c). It also demonstrates that the optical sectioning ability of the DSI-LC system is superior to the one of the LC system, and the optical sectioning ability of the LC system is superior to the one of the WF system. Figures 3(a1)–(c1) are 40 µm-thick maximum intensity projections (MIPs) of the WF, LC, and DSI-LC images, respectively. DSI-LC image has a lower background and higher resolution than LC image.
Fig. 3. Pollen grain and cell microtubule imaging. (a)–(c) Single-layer images of pollen grain at Z = 20 µm obtained by WF, LC, and DSI-LC, respectively. (a1)–(c1) The 40-µm thick MIPs of the same sample by WF, LC, and DSI-LC, respectively. (d) Normalized intensity profile along colored solid lines corresponding with (a1)–(c1). (e) and (i) Cellular microtubule obtained by LC and DSI-LC, respectively. (f)–(g) Expanded views of the areas by the red dashed lines in (e). (h) Normalized intensity profile along colored solid lines corresponding with (f)–(g). (j)–(k) Enlarged views of yellow dashed boxes in (i). (l) Normalized intensity profile along corresponding colored solid lines in (j)–(k). Scale bars: 5 µm in (a)–(c) and (a1)–(c1), 15 µm in (e) and (i), and 3 µm in (f)–(g) and (j)–(k).
Figures 3(e) and (i) show the LC and DSI-LC images of the same cellular microtubule labeled by phalloidin (BPAE cells, Thermo Fisher Scientific, USA), also demonstrating less background of DSI-LC than the one of LC. To show imaging details clearly, we further enlarge the areas indicated by the corresponding color-dotted boxes in Figs. 3(e) and (i). The cellular microtubule along the X direction is shown in Figs. 3(f) and (g). Adjacent lines are separated clearly in Fig. 3(f), but close to each other in Fig. 3(g). Figure 3(h) shows the normalized intensity profiles along the corresponding colored solid lines in Figs. 3 (f)–(g). The profile of the DSI-LC image has a lower background than the one of the LC image, and the FWHM of the DSI-LC line has narrower than the one of the DSI-LC line. The contrast of Figs. 3(f)–(g) are 5.3 and 58.1, respectively. In contrast, another cellular microtubule along the Y direction is shown in Figs. 3(j)–(k). Figure 3(l) shows the normalized intensity profiles along the corresponding colored solid lines in Figs. 3 (j)–(k). The DSI profile has a lower background than the LC one but without resolution improvement of the Y axis. The DSI method preliminarily demonstrates the improvements of the spatial resolution and optical sectioning abilities in imaging both pollen grain and cell microtubule, which has the potential to distinguish fine and complex details and inhibit strong scattering in tissues.
3.3 High-resolution, high-contrast 3D imaging of mouse brain ex vivo
The super-complex fine neuronal fibers and strong tissue scattering hinder high-quality data acquisition in the mouse brain. To demonstrate our advantages of high-quality imaging, we imaged the Thy1-GFP transgenic mouse (Jackson Laboratory, Bar Harbor, ME, USA) brain. The mouse brain was embedded with resin, and the tissue thickness was more than 1 cm. All animal experiments followed procedures approved by the Institutional Animal Ethics Committee of Huazhong University of Science and Technology. Since the size of the mouse brain is much larger than the entire FOV of the objective lens, the GM remains stationary during imaging, and the sample is scanned by the 3D stage. The surface of the sample was cut by a diamond knife at 2-µm Z-step and then imaged by the DSI-LC system until a total of 64 layers were imaged. Figure 4(a) shows the result of a single layer. The left half of Fig. 4(a) [LC imaging] has a stronger background than the right half [DSI-LC imaging]. Figure 4(b) is the 1137.5 × 665.5 × 128 µm3 3D reconstruction of Fig. 4(a) using the DSI method. The neuronal fiber and the soma are clear in the 3D reconstruction. To further show details, we enlarge the yellow dotted boxes in Fig. 4(a) as shown in Figs. 4(c) [LC image] and 4(d) [DSI-LC image]. The strong background surrounding the neuronal fibers was effectively rejected by using the DSI method. Figures 4(c1) and 4(d1) are the 3D reconstruction with the 128 µm depth of Figs. 4(c) and 4(d), respectively. The corresponding X-Y and Y-Z MIPs of Figs. 4(c1) and (d1) are shown in Fig. S1 (Supplement 1). The 3D reconstruction of DSI-LC imaging has better image quality than the one of LC imaging. To show the resolution enhancement, we further enlarge the areas indicated by the dotted boxes of Figs. 4(c)–(d) as shown in Figs. 4(e)–(f). After the resolution enhancement and background suppression using the DSI method, the dendritic spines and neuronal fibers are more easily distinguished in the bottom left corner of Fig. 4(e) and (f). Figures 4(e1)–(f1) are Fourier transform corresponding to Figs. 4(e)–(f), respectively. Figure 4(f1) has a more high-frequency component in the x direction than Fig. 4(e1), which also indicates the LC resolution has been improved by the DSI method. The normalized intensity profiles along the corresponding yellow dotted lines in Figs. 4 (e)–(f) were plotted in Fig. 4 (g). The contrast of Figs. 4(e)–(f) are 1.2 and 13.1, respectively. The normalized intensity profiles and Figs. 4(e)–(f) corroborate each other in terms of performance enhancements. Figures 4(e2) and 4(f2) are the 10 µm-deep 3D reconstruction corresponding to Figs. 4(e) and 4(f), respectively. The neuronal fiber labeled by the red arrow is displayed in Fig. 4(f2). These results indicate that our method facilitates the high-quality data acquisition of whole-brain imaging.
Fig. 4. 3D imaging of mouse brain tissue. (a) Single-layer image of the sample. (b) 3D reconstruction of (a) based on the DSI method. (c) and (d) Enlarged views of the yellow dotted box of (a). (e) and (f) Enlarged views of the yellow dotted box corresponding with (c) and (d), respectively. (c1) and (d1) 3D reconstructions of (c) and (d), respectively. (e1) and (f1) Fourier transform of (e) and (f), respectively. (g) Normalized intensity profile along yellow dashed lines in (e) and (f). (e2) and (f2) 3D reconstructions of (e) and (f), respectively. Scale bars: 150 µm in (a), 50 µm in (c) and (d), and 10 µm in (e) and (f).
3.4 High-quality 3D imaging of optical clearing mouse brain
Optical clearing technology could effectively address the tissue scattering problem in optical imaging and had been widely used in 3D tissue imaging. However, insufficient transparency of deep tissue affects the quality of 3D tissue imaging. This issue could be effectively solved by our method with excellent optical sectioning ability. A Thy1-GFP transgenic mouse brain was sliced into 250 µm pieces by a vibratome, and the brain slice was processed by CUBIC optical clearing technology [33]. The cleared brain slice was axially scanned 125 layers with a 2 µm step. Figure 5(a) shows the MIP of the brain slice image stack. The left side of Fig. 5(a) [LC imaging] has a higher background than the right side [DSI-LC imaging]. After processing by the DSI method, the neuronal fibers were more easily distinguished and still allowed continuous 3D reconstruction. To further show the image details, we enlarged the areas indicated by the yellow dotted box in Fig. 5(a). Figures 5(b)–(b2) and (c)–(c2) are single-layer images of LC and DSI-LC at different depths, respectively. The background of LC images is gradually increased from shallow to deep, and the background of DSI-LC images is suppressed at different depths. To show partial details, the areas of the yellow dotted box correspond with Figs. 5(b)–(b2) and (c)–(c2) were enlarged at the upper right corner. The normalized intensity profiles along the dashed lines corresponding to Figs. 5(b)–(b2) and 5(c)–(c2) are plotted in Figs. 5(h)–(h2). The contrasts of LC image in Figs. 5(h)–(h2) are 8.7, 1.1, and 0.8, respectively. The contrasts of the DSI-LC image in Figs. 5(h)–(h2) are 28.1, 5.8, and 5.4, respectively. The intensity profiles of DSI-LC images had lower backgrounds than the intensity profiles of LC images, and FWHMs of DSI-LC intensity profiles were smaller than FWHMs of DSI-LC intensity profiles. Figures 5(d)–(d2) and (e)–(e2) were 50 µm-thick 3D reconstructed image stacks corresponding to Figs. 5(b)–(b2) and 5(c)–(c2), respectively. Figures 5(f) and (g) were MIP images of LC and DSI-LC corresponding to the yellow dotted box in Fig. 5(a). Figures 5(f1) and (g1) were 3D reconstruction images of LC and DSI-LC corresponding to the yellow dotted box in Fig. 5(a). The corresponding X-Y and Y-Z MIPs of Figs. 5(f1) and (g1) are shown in Fig. S2 (Supplement 1). The neurons and neuronal fibers can be clearly and continuously displayed by using the DSI method. The experimental results suggest that the DSI method with its excellent optical sectioning ability could effectively improve the image quality of insufficient transparency tissue.
Fig. 5. Imaging of cleared mouse brain tissue. (a) 250 µm MIP of the brain slice. (b)–(b2) Enlarged views of LC image at different depths corresponding to the yellow dotted box of (a). (c)–(c2) Enlarged views of DSI-LC image at different depths corresponding to the yellow dotted box of (a). (d)–(d2), (e)–(e2) 50 µm 3D reconstruction corresponding with (b-b2) and (c)–(c2), respectively. (f) 250 µm MIP of LC image corresponding to the area indicated by the yellow dotted box in (a). (g) 250 µm MIP of DSI-LC image corresponding to the area indicated by the yellow dotted box in (a). (f1), (g1) 250 µm 3D reconstruction of the area indicated by the yellow dotted box in (a). (h)–(h2) Normalized intensity profile along colored dashed lines corresponding with (b)–(b2) and (c)–(c2), respectively. Scale bars: 100 µm in (a), and 20 µm in (b)–(b2), (c)–(c2), and (f), (g).
3.5 High contrast 3D imaging of larval zebrafish neural network in vivo
Imaging of neural networks in larval zebrafish is conducive to neuroscience research. To demonstrate the advantage of high-contrast imaging, we image a neural network of larvalzebrafish labeled by the elval3 gene in vivo. The elval3 gene specifically expresses green fluorescent protein in systemic neurons. A zebrafish at 5 days post fertilization (dpf) is anesthetized and then embed by 1% low-melting-point agarose in a confocal culture dish. The zebrafish had been axially imaged in 200 layers with a 2 µm step. Since the camera has a 16- bit dynamic range, while the displayer has only an 8-bit dynamic range, too high and too low gray values of the raw data can’t be shown in the displayer at the same time. To solve this problem, we did a gamma transform in Figs. 6(a)–(g). Figure 6(a) is 400 µm MIP based on the DSI method. The Figs. 6(b)–(d) [LC image], and Figs. 6(e)–(g) [DSI-LC image] are the enlarged figures of the red dotted boxes in Fig. 6(a), respectively. The neuronal fibers indicated by the red arrows in Fig. 6(g) are shown full and clear, while in Fig. 6(d) invisible due to strong background covering. Besides, Figs. 6(e)–(g) have lower background than Figs. 6(b)–(d), which is conducive to the visual reconstruction of neural networks. Figures 6(b1)–(d1) and Figs. 6(e1)–(g1) are 3D reconstruction of Figs. 6(b)–(d) and Figs. 6(e)–(f), respectively. The 3D reconstruction images of DSI-LC have a better background suppression than the one of LC. It can be seen from the above experiments that zebrafish are transparent during the early stages of incubation, which helps observe the development process of zebrafish by optical microscopy. However, “transparency” isn’t a single refractive index in the physical. LC image is still affected by tissue scattering in deep imaging, which causes signal submersion by the background. In contrast, our method enables us to effectively suppress deep scattering and easily separate neuronal fibers from the background.
Fig. 6. Imaging of larval zebrafish neural network. (a) The 400 µm MIP of the whole body image based on the DSI method. (b)–(d) Enlarged views of the LC image corresponding to the red dashed box of (a). (e)–(f) Enlarged views of DSI-LC image corresponding to the red dotted box of (a). (b1)–(d1), (e1)–(g1) 3D reconstruction corresponding with (b)–(d) and (e)–(g), respectively. Scale bars: 200 µm in (a), 150 µm in (c), (d), (f) and (g), and 50 µm in (b) and (e).
3.6 Video-rate imaging of larval zebrafish heartbeat in vivo
Fast imaging is very important for recording time-varying functional information in vivo. To examine our fast imaging capability, we image the heart beating of larval zebrafish in vivo. The runx1: GFP zebrafish (labeling the hematopoietic stem cells) at 6 dpf is anesthetized and then embed in 1% low-melting-point agarose in the confocal culture dish. The whole zebrafish is 3D imaged by stage scanning and the MIP of the result is shown in Fig. 7(a). Then, the zebrafish heart is translated into the middle of the microscope’s FOV by the stage. The dynamic process of zebrafish heartbeat is recorded by beam scanning (Visualization 1) at 24 Hz in a 665.6 × 332.8 µm2 FOV. Figure 7(b) shows the statistical results of stress rate in one complete heartbeat cycle. The strain rate increases during the diastole stage of the beating ventricle, while decreasing during the systole stage of the beating ventricle. Figures 7(c) and (d) show the two images 41.7 ms apart in one heartbeat cycle. During the diastole, cells flow to the ventricle, which gradually grows larger. We recorded the corresponding positions of the cell indicated by the yellow circle in Figs. 7(c)–(d) and calculated that the blood flow velocity was 1.4 mm/s [34]. Figure 7(e) shows the 3D reconstruction of the zebrafish heart. Figure 7(f) shows its cross-sectional image, and the atrium and ventricle are visible. The experimental results suggest that the DSI-LC method could realize high-definition imaging of larval zebrafish heartbeat at video rate in vivo and potentially facilitates cardiac hemodynamics research.
Fig. 7. Video-rate imaging of larval zebrafish heartbeat. (a) MIP of the whole zebrafish based on the DSI method. (b) The images of the strain rate of the myocardium during one heartbeat cycle. (c), (d) Images of two randomly-selected periods in one heartbeat cycle. Yellow circles indicate the positions of the same cell at different moments. Yellow dashed circles indicate the ventricle. (e) 3D reconstruction of zebrafish heart of yellow dashed rectangle in (a). (f) Cross-sectional image of zebrafish heart. The Red dashed circle indicates the heart. A, atrium; V, ventricle. Scale bars: 200 µm in (a), 30 µm in (b), and 50 µm in (c), (d), and (f). The bold dashed lines are the reference lines of spatial coordinates to identify the cell movement.
4. Discussion and conclusions
All the above experimental results demonstrate that the spatial resolution and optical sectioning ability of the LC system are improved by the DSI method based on multi-line detection. Generally, previous approaches [8,15,25] have tended to improve single system parameters, such as spatial resolution, optical sectioning ability, and imaging speed, at the expense of other system performances. These improvements often limit their application scenarios and prevent their widespread application in biomedical research. Unlike previous difference methods [30–32] using multi-scanning or multi-detector, our method constructs the different PSFs integrating TDI technology and redundant information on multi-line detection in a single scan. On one hand, the DSI method improves both spatial resolution and optical sectioning ability without sacrificing the imaging speed of the LC system. On the other hand, the usage of a single camera avoids the registration of multi-channel data and reduces the amount of multiple data acquisition. The characteristic of natural registration improves the robustness of the system which ensures long-term high-quality imaging of the large-scale sample.
The imaging speed in a single FOV is 24 Hz, limited by the maximal frame rate and minimal sub-array line number of the camera. In the future, the imaging speed would be further improved by high-speed linear detectors. For example, the linear camera of DALSA (ML-HM-16k30H-00-R, Teledyne DALSA, Waterloo, ON, Canada) can reach as fast as 300 kHz at 16,384 × 4 pixels. In theory, only a two-line detector enables to achieve the DSI method. It indicates that the decrease of the minimal line number of the camera potentially further improves the imaging speed in the future. The axial scanning of our system is realized by a 3D stage. The inertia of the stage limits the scanning speed of volumetric imaging. To address this issue, the tunable acoustic gradient lens should be integrated into our system to improve the axial speed of volume imaging. Furthermore, the DSI method achieves the resolution improvement of the LC system but hasn’t broken its diffraction limit. In the next step, the super-resolution method, such as PSF engineering, can be employed to further improve the resolution beyond the diffraction limit. Imaging depth is also an important index for imaging in vivo. In our method, the imaging depth of the single photon is limited to about 100 µm, which limits the detection of deep tissue. Multi-photon excitation is an effective approach to achieving deeper imaging by using longer wavelengths and nonlinear effects.
In summary, we have demonstrated for the first time, to the best of our knowledge, that the DSI-LC system has successfully improved both the spatial resolution and optical sectioning ability without sacrificing the imaging speed. We propose the DSI method to employ a single camera in a single scan to achieve the construction of different PSFs. Our method reduces the system cost, simplifies the system configuration, and improves the system stability on the premise of guaranteeing imaging performance. The related experimental results show that this method is beneficial for high-throughput and high-quality data acquisition in different imaging scenes. Given the characteristic of high spatiotemporal resolution and strong background suppression ability, the DL system is a promising tool for biomedical imaging, especially large-scale and in vivo imaging.
Funding
STI2030-Major Projects (2021ZD0201001); National Natural Science Foundation of China (81827901).
Acknowledgments
The authors thank Prof. Jingxia Liu and Dr. Zhipeng Tai of Huazhong Agricultural University for providing help with zebrafish experiments.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
Supplemental document
See Supplement 1 for supporting content.
References
1. P. Osten and T. W. Margrie, “Mapping brain circuitry with a light microscope,” Nat. Methods 10(6), 515–523 (2013). [CrossRef]
2. X. Zhu, Y. Xia, X. Wang, K. Si, and W. Gong, “Optical brain imaging: a powerful tool for neuroscience,” Neurosci. Bull. 33(1), 95–102 (2017). [CrossRef]
3. J. Pawley, Handbook of Biological Confocal Microscopy (Springer Science & Business Media, 2006), Vol. 236.
4. M. Muller, Introduction to Confocal Fluorescence Microscopy (SPIE Press, 2006), Vol. 69.
5. X. Qi, H. Xiong, X. Lv, J. Chen, H. Gong, Q. Luo, and S. Zeng, “Improved detectability of neuronal connectivity on mechanical sectioning setup by using confocal detection,” J. Biomed. Opt. 18(5), 050506 (2013). [CrossRef]
6. J. Jonkman, C. M. Brown, G. D. Wright, K. I. Anderson, and A. J. North, “Tutorial: guidance for quantitative confocal microscopy,” Nat. Protoc. 15(5), 1585–1611 (2020). [CrossRef]
7. A. Nakano, “Spinning-disk confocal microscopy—a cutting-edge tool for imaging of membrane traffic,” Cell Struct. Funct. 27(5), 349–355 (2002). [CrossRef]
8. A. Egner, V. Andresen, and S. Hell, “Comparison of the axial resolution of practical Nipkow-disk confocal fluorescence microscopy with that of multifocal multiphoton microscopy: Theory and experiment,” J. Microsc. 206(1), 24–32 (2002). [CrossRef]
9. K. D. Mac, M. M. Qureshi, M. Na, S. Chang, T. J. Eom, H. S. Je, Y. R. Kim, H.-S. Kwon, and E. Chung, “Fast volumetric imaging with line-scan confocal microscopy by electrically tunable lens at resonant frequency,” Opt. Express 30(11), 19152–19164 (2022). [CrossRef]
10. T. Wilson and S. Hewlett, “Imaging in scanning microscopes with slit-shaped detectors,” J. Microsc. 160(2), 115–139 (1990). [CrossRef]
11. M. A. Neil, R. Juškaitis, and T. Wilson, “Method of obtaining optical sectioning by using structured light in a conventional microscope,” Opt. Lett. 22(24), 1905–1907 (1997). [CrossRef]
12. D. Lim, K. K. Chu, and J. Mertz, “Wide-field fluorescence sectioning with hybrid speckle and uniform-illumination microscopy,” Opt. Lett. 33(16), 1819–1821 (2008). [CrossRef]
13. N. Hagen, L. Gao, and T. S. Tkaczyk, “Quantitative sectioning and noise analysis for structured illumination microscopy,” Opt. Express 20(1), 403–413 (2012). [CrossRef]
14. J. Mazzaferri, D. Kunik, J. Belisle, K. Singh, S. Lefrançois, and S. Costantino, “Analyzing speckle contrast for HiLo microscopy optimization,” Opt. Express 19(15), 14508–14517 (2011). [CrossRef]
15. O. Mandula, M. Kielhorn, K. Wicker, G. Krampert, I. Kleppe, and R. Heintzmann, “Line scan-structured illumination microscopy super-resolution imaging in thick fluorescent samples,” Opt. Express 20(22), 24167–24174 (2012). [CrossRef]
16. H. Zhang, K. Vyas, and G.-Z. Yang, “Line scanning, fiber bundle fluorescence HiLo endomicroscopy with confocal slit detection,” J. Biomed. Opt. 24(11), 116501 (2019). [CrossRef]
17. W. Qiao, R. Jin, T. Luo, Y. Li, G. Fan, Q. Luo, and J. Yuan, “Single-scan HiLo with line-illumination strategy for optical section imaging of thick tissues,” Biomed. Opt. Express 12(4), 2373–2383 (2021). [CrossRef]
18. Q. Zhong, A. Li, R. Jin, D. Zhang, X. Li, X. Jia, Z. Ding, P. Luo, C. Zhou, and C. Jiang, “High-definition imaging using line-illumination modulation microscopy,” Nat. Methods 18(3), 309–315 (2021). [CrossRef]
19. Q. Zhong, C. Jiang, D. Zhang, S. Chen, R. Jin, H. Gong, and J. Yuan, “High-throughput optical sectioning via line-scanning imaging with digital structured modulation,” Opt. Lett. 46(3), 504–507 (2021). [CrossRef]
20. J. Huisken, J. Swoger, F. Del Bene, J. Wittbrodt, and E. H. Stelzer, “Optical sectioning deep inside live embryos by selective plane illumination microscopy,” Science 305(5686), 1007–1009 (2004). [CrossRef]
21. E. G. Reynaud, J. Peychl, J. Huisken, and P. Tomancak, “Guide to light-sheet microscopy for adventurous biologists,” Nat. Methods 12(1), 30–34 (2015). [CrossRef]
22. Y. Chen, X. Li, D. Zhang, C. Wang, R. Feng, X. Li, Y. Wen, H. Xu, X. S. Zhang, and X. Yang, “A versatile tiling light sheet microscope for imaging of cleared tissues,” Cell Rep. 33(5), 108349 (2020). [CrossRef]
23. Y. S. Hu, M. Zimmerley, Y. Li, R. Watters, and H. Cang, “Single-molecule super-resolution light-sheet microscopy,” ChemPhysChem 15(4), 577–586 (2014). [CrossRef]
24. B.-C. Chen, W. R. Legant, K. Wang, et al., “Lattice light-sheet microscopy: Imaging molecules to embryos at high spatiotemporal resolution,” Science 346(6208), 1257998 (2014). [CrossRef]
25. Y. Wu, P. Chandris, P. W. Winter, E. Y. Kim, V. Jaumouillé, A. Kumar, M. Guo, J. M. Leung, C. Smith, I. Rey-Suarez, H. Liu, C. M. Waterman, K. S. Ramamurthi, P. J. La Riviere, and H. Shroff, “Simultaneous multiview capture and fusion improves spatial resolution in wide-field and light-sheet microscopy,” Optica 3(8), 897–910 (2016). [CrossRef]
26. S. Hewlett and T. Wilson, “Resolution enhancement in three-dimensional confocal microscopy,” Machine Vis. Apps. 4(4), 233–242 (1991). [CrossRef]
27. C. Kuang, S. Li, W. Liu, X. Hao, Z. Gu, Y. Wang, J. Ge, H. Li, and X. Liu, “Breaking the diffraction barrier using fluorescence emission difference microscopy,” Sci. Rep. 3(1), 1441 (2013). [CrossRef]
28. H. Dehez, M. Piché, and Y. De Koninck, “Resolution and contrast enhancement in laser scanning microscopy using dark beam imaging,” Opt. Express 21(13), 15912–15925 (2013). [CrossRef]
29. C. Liu, X. Yu, C. Bai, X. Li, Y. Zhou, S. Yan, J. Min, D. Dan, R. Li, and S. Gu, “Axial resolution enhancement for planar Airy beam light-sheet microscopy via the complementary beam subtraction method,” Appl. Opt. 60(32), 10239–10245 (2021). [CrossRef]
30. V. Poher, G. T. Kennedy, H. B. Manning, D. M. Owen, H. X. Zhang, E. Gu, M. D. Dawson, P. M. French, and M. A. Neil, “Improved sectioning in a slit scanning confocal microscope,” Opt. Lett. 33(16), 1813–1815 (2008). [CrossRef]
31. M. Hughes and G.-Z. Yang, “Line-scanning fiber bundle endomicroscopy with a virtual detector slit,” Biomed. Opt. Express 7(6), 2257–2268 (2016). [CrossRef]
32. Y. Tang, A. Kortum, I. Vohra, J. Carns, S. Anandasabapathy, and R. Richards-Kortum, “Simple differential digital confocal aperture to improve axial response of line-scanning confocal microendoscopes,” Opt. Lett. 44(18), 4519–4522 (2019). [CrossRef]
33. E. A. Susaki, K. Tainaka, D. Perrin, H. Yukinaga, A. Kuno, and H. R. Ueda, “Advanced CUBIC protocols for whole-brain and whole-body clearing and imaging,” Nat. Protoc. 10(11), 1709–1727 (2015). [CrossRef]
34. H. C. Yalcin, A. Amindari, J. T. Butcher, A. Althani, and M. Yacoub, “Heart function and hemodynamic analysis for zebrafish embryos,” Dev. Dyn. 246(11), 868–880 (2017). [CrossRef]
