Optical sectioning methods in three-dimensional bioimaging

Abstract

In recent advancements in life sciences, optical microscopy has played a crucial role in acquiring high-quality three-dimensional structural and functional information. However, the quality of 3D images is often compromised due to the intense scattering effect in biological tissues, compounded by several issues such as limited spatiotemporal resolution, low signal-to-noise ratio, inadequate depth of penetration, and high phototoxicity. Although various optical sectioning techniques have been developed to address these challenges, each method adheres to distinct imaging principles for specific applications. As a result, the effective selection of suitable optical sectioning techniques across diverse imaging scenarios has become crucial yet challenging. This paper comprehensively overviews existing optical sectioning techniques and selection guidance under different imaging scenarios. Specifically, we categorize the microscope design based on the spatial relationship between the illumination and detection axis, i.e., on-axis and off-axis. This classification provides a unique perspective to compare the implementation and performances of various optical sectioning approaches. Lastly, we integrate selected optical sectioning methods on a custom-built off-axis imaging system and present a unique perspective for the future development of optical sectioning techniques.

Fig. 1. Schematic diagram of coaxial and off-axis imaging.

a Schematic diagram of coaxial imaging. b Schematic diagram of mixed detection in off-axis imaging. c Schematic diagram of separated detection in off-axis imaging. Images beside each system diagram represent the point spread function along the xz direction

Fig. 2. Schematic diagram of optical sectioning methods.

a Focal plane conjugation in coaxial imaging. b Modulated illumination in coaxial imaging. c Mixed detection in off-axis imaging. d Separated detection in off-axis imaging. e Multiple scanning of separated detection

Fig. 3. Optical sectioning strength curves for various sectioning methods.

a Optical sectioning strength curves for focal plane conjugation. b Optical sectioning strength curves for intra-focal excitation. c Optical sectioning strength curves for modulated illumination, where s_max is set to 2, and ‘bw’ represents the filter bandwidth. d Optical sectioning strength curves for mixed detection. e Background suppressions in thick samples for various sectioning methods. f Optical sectioning strength for various reconstruction methods in separated detection

Fig. 4. Spatiotemporal multiplexing scanning of separated detection in conjunction with various reconstruction methods.

a Schematic diagram of spatiotemporal multiplexing scanning. b Intensity variation along the off-axis direction and the defocused direction. c Schematic diagram of acquiring raw data. d Schematic diagram of different reconstruction methods based on raw data

Fig. 5. Simulation results of different optical sectioning methods by off-axis spatiotemporal multiplexed detection on the synthetic sample.

a–f Raw data at different off-axis positions. g The 512th layer of the synthetic sample. h–j Reconstructions of g via DSIM, DHiLo, and LiMo, respectively. The red arrows indicate the difference between ground truth and reconstructions. Scale bar: 3 μm in a–j

Fig. 6. The schemes of recent developments.

a Spinning disk microscopy. b Deep-learning assisted throughput enhancement. c PSF engineering-based throughput enhancement. d Stimulated emission depletion microscopy. e Image scanning microscopy. f Fluorescence differential microscopy. g Random-access multi-photon microscopy. h Temporally focused multi-photon microscopy. i Total internal reflection fluorescence microscopy. j DMD-based structured illumination microscopy. k 3D structured illumination microscopy. l Hybrid illumination microscopy. m Digitally scanned light sheet microscopy. n Open-top light sheet microscopy. o Single-objective light sheet microscopy. p Line-illumination modulation microscopy. OL, objective lens; TL, tube lens; DM, dichroic mirror; BFP, back focal plane; CL, Cylindrical lens

Fig. 7. Curves of performance metric.

a Curves of axial responds via different methods, indicating optical sectioning strength. b Curves of lateral PSF, indicating resolution. c, d Imaging time varies with sample size via different scanning modes when the sample size is larger than the field of view and smaller than the field of view, correspondingly. “TP” and “WF” represent two-photon and wide-field microscopy, respectively

Fig. 8. Performance comparison for different metrics.

a–d Comprehensive maps of line confocal, two-photon, HiLo, and LiMo microscopy, respectively. To highlight differences, the dimensionality charts are drawn based on theoretical conditions, from 1 (worst) to 5 (best)

Fig. 9. Recommended solutions for different applications.

Based on the analyses above, we establish a selection process for different applications based on the sample geometry and specific needs in Fig. 9. In the category of cell samples or even thin slices, most optical sectioning methods are suitable due to the low background signals. The experimental plan depends on specific needs, like speed for live imaging, resolution for fixed samples, and system accessibility. In the category of localized observation in thick samples or blocks, for transparent live animals like zebrafish, where speed and motion resistance are more critical than sectioning strength and thus part of the coaxial and off-axis methods are suitable; for transparent fixed samples like in situ observation of 3D gene expression, due to the density of gene expression and high demands for optical sectioning and resolution, off-axis imaging systems have more advantages; for opaque samples, two-photon microscopy offers a distinct advantage in penetration depths. In whole-mount or organ imaging, field of view (FOV) stitching is employed to cover large areas, making imaging speed equally crucial to other requirements. For imaging single neuronal morphology with axonal resolution through the whole brain, with millimeter dimensions and sub-micrometer resolution, off-axis separated detection with line scanning is preferred for its high optical sectioning and throughput, surpassing light sheet microscopy with NA limitations in resolution. For mapping cell distribution or vessel networks with large volumes and microscale structures, throughput is prioritized over optical sectioning and resolution, making both on-axis and off-axis strip scanning imaging methods suitable. In summary, the coaxial system is comparable to the off-axis system in imaging thin samples with low background signals or thick samples with details to be resolved far bigger than the resolution limit, where the coaxial system is preferred due to its popularity and accessibility. However, the off-axis system is indispensable in imaging thick samples with detailed features close to the resolution limit

Fig. 10. Images of a 50μm thick Thy-1 -GFP M line transgenic mouse brain slice in both line scanning off-axis separated detection and wide-field coaxial systems to compare SIM and HiLo in different detection approaches.

a, b Maximum intensity projections (MIPs) of the mouse brain slice by SIM reconstruction via wide-field (WF) and line scanning imaging approaches, respectively. c, d WF-SIM- and WF-HiLo-reconstructed enlarged views from the red square in a, respectively. e, f DSIM- and DHiLo-reconstructed enlarged views of the red square in (b), respectively. g, h Enlarged images from the yellow squares in (c) and (e), respectively. i, j Enlarged images from the blue square in (c) and (e), respectively. k, l Enlarged images from the yellow square in (d) and (f), respectively. m, n Enlarged images from the blue square in (d) and (f), respectively. Scale bar: 1 mm in a and b, 100 μm in c–f, and 10 μm in g–n

Fig. 11. Flexible reconstruction implementations of off-axis separated detection in different applications.

Reconstructions of a pollen, b gene distribution in 3T3 cell, c gene distribution in mouse brain slice, d mouse brain slice, and e mouse liver via different methods. Reconstructions of the f mouse brain slice when the linear illumination is titled and g mouse liver in vivo via different methods. h Coronal plane of the mouse brain and the reconstruction time via different methods. Scale bar: 10 μm in (a–c), (e), and (g), 5 μm in (d), 20 μm in (f), and 1 mm in (h)