Light-sheet fluorescent microscopy: fundamentals, developments and applications

Light-sheet illumination was first introduced over a century ago in 1903 as ultramicroscopy by Siedentopf and Zsigmondy [12] when they were working for Carl Zeiss company. In their Ultramicroscope, sunlight was projected through a slit aperture to observe gold particles. The idea did not stimulate further developments until 1993, when it was reintroduced by Voie and colleagues as an imaging tool for orthogonal plane fluorescence sectioning (OPFOS) to visualize the morphology of guinea-pig cochlea [13]. OPFOS included all elements present in basic LSFM devices including laser source, beam expander, light-sheet generator via a cylindrical lens, specimen chamber, orthogonally illuminating sample with regard to detection arm, z-stack generation through specimen movement, and fluorescent labeling. The basic theory and design of LSFM were discussed in detail by Stelzer and Lindek in 1995, while they were developing an oblique illuminating confocal microscope to improve the axial resolution of confocal microscopy [14]. In their study, detection objective was mounted at 102° with respect to illumination objective. This system was known as confocal theta microscopy [14] that later created a foundation for their subsequent version of a LSFM device called selective or single-plane illumination microscopy (SPIM) in 2004 [10]. Moreover, Fuchs et al (2002) described an OPFOS-like device, which they called it a thin laser light sheet microscope (TLSM) to examine microorganisms in seawater [15]. Finally, the first use of LSFM for real-time in vivo imaging was reported in 2004 by Stelzer's group to investigate the embryonic development of Medaka and Drosophila [10]. Named as selective plane illumination microscopy (SPIM) by its developers, this technology created a turning point for the adoption of LSFM in high-speed imaging of live biological events.

To produce more uniform light-sheet excitation for improved imaging quality of large specimens Keller et al developed digital scanned laser light-sheet fluorescence microscopy (DLSM) [16] in 2008. Unlike OPFOS and SPIM that employ a cylindrical lens to form a light -sheet, DLSM generates a virtual light-sheet by rapidly scanning a single beam across FOV with a scanning mirror. A camera collects all signals obtained during one operation period of the scanning mirror, figure 2. The beam needs to have higher power than a static light sheet system since each fraction of the final image is captured from a fraction of the camera's exposure time to achieve the same fluorescence yield in the same imaging period. While this may result in the saturation of fluorophore with high photo bleaching, it is compatible with applications that require high laser power such as two-photon excitation and beam shaping applications.

Zoom In Zoom Out Reset image size

There are a large variety of configurations for LSFM; however, their functions revolve around the same principles. In this section, design of a basic LSFM relying on a static unidirectional illumination is discussed. In addition, required optical considerations to design a LSFM based on sample size are presented. For the sake of simplicity, relation between the thickness of light-sheet in illumination arm and the axial resolution in detection arm is neglected. Calculations with more details could be found in [17–19].

Figure 3 shows a simplified version of LSFM system. The system is comprised of two perpendicular independent arms, for illuminating and detecting purposes, and a scanning mechanism. In the illumination arm, a cylindrical lens is utilized to produce a light-sheet. A beam expander, which constitutes of two lenses (L₁, and L₂), is used to widen laser beam thickness at the entrance aperture of a cylindrical lens. In this way, the effective numerical aperture is increased that allows for producing thinner light-sheets. The clear entrance aperture of the cylindrical lens could be determined using an adjustable slit to regulate the thickness of the light-sheet.

Zoom In Zoom Out Reset image size

3D images are formed through stacking of 2D images captured from different layers of sample using the detection arm. Different layers of sample are illuminated via relative displacement of sample and light-sheet. In the simplest case, this could be achieved with placing sample on a stage to move it along the detection axis, perpendicular to the illumination arm.

The detection arm is formed from a regular microscope that includes a detection objective, a tube lens, a filter and a camera.

The lateral resolution of captured images for both fluorescence microscopy and LSFM is affected by the wavelength of the fluorescence emission, λ, the focal length of the detection lens, ${\rm{f}},$ and the entrance aperture of the detection lens, $D,$ as follows [20]:

$$\begin{eqnarray}&&R=1.22\displaystyle \frac{\lambda f}{D}=0.61\displaystyle \frac{\lambda }{N{A}_{det}\,}\end{eqnarray} \tag{ 1 }$$

where $N{A}_{det}$ is the numerical aperture of detection lens.

Considering the perpendicular configuration of illumination and detection arms, the imaging plane is formed at the confocal region of illumination and detection arms, figure 4. The field of view (FOV) is determined as:

$$\begin{eqnarray}&&FOV=\displaystyle \frac{L}{M}=\displaystyle \frac{{n}_{pixel}\times {d}_{pixel}}{M}\end{eqnarray} \tag{ 2 }$$

where, $L$ is sensing length of camera, ${n}_{pixel}$ shows the number of pixels in the direction of illumination, ${d}_{pixel}$ is pixel pitch, and M refers to magnification factor of the imaging system.

Zoom In Zoom Out Reset image size

Magnification of imaging system equals to the ratio of focal length of tube lens (${f}_{TL}$) to focal length of detection objective (${f}_{DO}$):

$$\begin{eqnarray}&&M=\displaystyle \frac{{f}_{TL}}{{f}_{DO}}\end{eqnarray} \tag{ 3 }$$

The main advantage of LSFM is that the need for illumination is only confined to a thin layer of sample at each step of imaging. However, the thickness of illumination beam is wider at both sides of focal plane of cylindrical lens forming a parabolic–like light-sheet. As a result, axial resolution and contrast of recorded images are decreased at regions far from Rayleigh length, figure 4. Rayleigh length is the distance through which the expansion of laser beam could be neglected. The width of a beam expands with a factor of $\sqrt{2}$ upon propagation within a Rayleigh length. The Rayleigh length is introduced with the following equation [21]:

$$\begin{eqnarray}&&{Z}_{R}=\displaystyle \frac{\pi {{W}_{0}}^{2}}{\lambda }\end{eqnarray} \tag{ 4 }$$

where, ${W}_{0}$ and ${Z}_{R}$ are the beam waist in the focus of illuminating and the Rayleigh length, respectively. A light-sheet with a thickness of ${W}_{0}$ could be produced using an arrangement of a Gaussian incident beam and a cylindrical lens having a known focal length $\left({f}_{CL}\right),$ entrance aperture of an adjustable slit (${D}_{AS}$), and the obtained numerical aperture of illumination ($N{A}_{ill}$), as follows:

$$\begin{eqnarray}&&{W}_{0}=\displaystyle \frac{2\lambda {f}_{CL}}{\pi {D}_{AS}}=\displaystyle \frac{\lambda }{\pi N{A}_{ill}}.\end{eqnarray} \tag{ 5 }$$

Thus, through regulating aperture of cylindrical lens using an adjustable slit, an illumination plane with a desired thickness could be produced. The thickness of the light-sheet should be chosen in which $2{Z}_{R}$ is greater than or equal to either the FOV of the microscope or sample size.

An example for determining optical features of a LSFM for imaging of samples with different sizes is presented in table 2. The calculations were carried out for a system illuminated by a laser with wavelength of 488 nm via one of two cylindrical lenses having focal length of 5 cm or 10 cm, combined with a fluorescence emission at 520 nm.

Displacing the sample along the detection axis is the most straightforward approach to acquire a z-stack, figure 5(a). Since only the sample is moved and the rest of the optics remains fixed, this method is robust and ensures that the light-sheet stays in the detection's focal plane. Although this solution is technically simple, mechanical scanning creates vibrations that might induce imaging artifacts. Furthermore, such translation may be too slow to track some of the fast dynamic processes over the whole sample volume.

Zoom In Zoom Out Reset image size

In an alternative approach, which seems to be more suitable for fragile samples or samples that need to be kept in a water-like medium, sample is stationary to be scanned by LS. This is possible by either using a motorized stage to move the illumination objective axially, or keep the illumination objective fixed and move the LS using a scanning mirror. Obviously, in these methods, LS should remain in focus plane of the detection objective. In order to satisfy this prerequisite, different solutions have been proposed as listed below:

• A straightforward approach is to use mechanical refocusing, mounting the detection objective to an axially movable stage motor and keeping its relative distance to the illumination objective fixed [22, 23], figure 5(b).
• Without moving a specimen or detection objective, it is possible to achieve flexible volume imaging at much higher speed by utilizing remote focusing. It performs with an electrically tunable lens (ETL) in the detection path between the detection objective and the camera [24], figure 5(c).
• Finally, one could also expand the detection objective's depth of focus by a Wavefront coding technique to cover the entire scan range. The LS can be freely scanned, thus illuminating different planes that will eventually form a volume. This system incorporates a cubic phase mask in the output pupil of the detection objective and results in a slight degradation of lateral resolution [25], figure 5(d).

Optical resolution describes the ability of an imaging system to determine details of an object. Resolution of LSFM could be defined as lateral resolution and axial resolution. Lateral resolution is defined in imaging plane of detecting arm, while axial resolution is described in the perpendicular direction to this plane.

In conventional microscopes, 2D images, as illustrated in the y-z plane of figure 1, are captured by a camera, in which the lateral resolution (${R}_{L}$) is defined for this plane. On the other hand, the axial resolution (${R}_{A}$) is defined for 3D images along the optical axis of detection arm, as shown along x direction of figure 1.

In an imaging system using a 2D array detector, specification of CCD plays an important role in defining the final imaging properties including FOV and resolution [19]. Ignoring the role of detector, the imaging quality of conventional widefield microscopy is mainly determined by imaging optics. $PS{F}_{det}$ as lateral and axial intensity distribution of point object in the focal plane of imaging lens is determined by calculating the Fresnel-Kirchhoff integral. Consequently, the axial and lateral resolutions are calculated as below [26]:

$$\begin{eqnarray}&&{R}_{L}=0.61\displaystyle \frac{{\lambda }_{Em}}{N{A}_{Det}}\end{eqnarray} \tag{ 6 }$$

$$\begin{eqnarray}&&{R}_{A}=1.78\displaystyle \frac{n{\lambda }_{Em}}{N{A}_{Det}^{2}}\end{eqnarray} \tag{ 7 }$$

where ${\lambda }_{Em}$ is the wavelength of the emitted fluorescence, $N{A}_{Det}$ is the numerical aperture of detection, 1.78 is a coefficient that is related to the FWHM of the central lobe of the $sin{c}^{2}$ function [19], ${R}_{L}$ and ${R}_{A}$ are the lateral and axial resolution in LSFM, respectively. Comparison of equation (6) with equation (7) reveals that the axial resolution $\left({R}_{A}\right)$ is poorer than lateral resolution $\left({R}_{L}\right),$ since the lateral resolution is inversely proportional to the numerical aperture of the system ($NA$), however, the axial resolution is inversely proportional to the $N{A}^{2}$ [27].

The image formation in LSFM is described as below [28, 29]:

$$\begin{eqnarray}&&{I}_{image}\left(r\right)=\left({I}_{ill}\left(r\right).c\left(r\right)\right)\ast PS{F}_{Det}\left(r\right)\end{eqnarray} \tag{ 8 }$$

where ${I}_{image}$ is the captured image, ${I}_{ill}$ is the distribution of illumination intensity, c specifies fluorophores distribution, and $PS{F}_{Det}$ refers to the point spread function of detection system. ∗ refers to convolution operator.

Image of ideal point object in basic LSFM configuration is defined as $PS{F}_{sys}.$ Considering the ideal point object placed at origin as $\delta \left(x,\,y,z\right)$ and uniform illumination across FOV in y-z plane as ${I}_{ill}(x)$ in a simplified approach, the reconstructed image is described as follows:

$$\begin{eqnarray}&&PS{F}_{sys}\left(x,\,y,\,z\right)={I}_{ill}\left(x\right)\ast PS{F}_{Det}\left(x,\,y,\,z\right)\end{eqnarray} \tag{ 9 }$$

Thus, the axial resolution in a LSFM differs from conventional widefield microscopes and related to both detection specifications [30] and light-sheet width. Hence, decreasing the width of light-sheet results in improving the detection's axial resolution.

Isotropic resolution means that the resolution in the transversal plane (y–z plane) and that in the longitudinal direction (x direction) are equivalent. Isotropic resolution is an important feature in LSFM to improve image quality. There are several methods to perform isotropic resolution such as rotation of sample [31], replacement of illumination and objective arms [32], multi-directional illumination and detection [33], and use of structured beams [28, 34–36].

2.3.1. LSFM super resolution

Optical resolution of any microscopy method is restricted to diffraction criteria. As LSFM create 3D images, both lateral and axial resolutions are important. The affecting parameters on the optical resolution of LSFM have been discussed in the previous section (equations (6)–(9)).

Many biological and medical fields require subcellular resolution [10] that cannot be achieved using the basic configurations of LSFM. To enhance resolutions higher than Rayleigh criterion for super resolution microscopy, various techniques such as structured illumination microscopy (SIM) [37], stimulated emission depletion (STED) [38–40], and photo-activated localization microscopies (PALM) [41] could be integrated with LSFM.

An approach to enhance lateral resolution of LSFM could be obtained using high NA objectives as detection objectives (DO) (equation (6)). However, this method is not very compatible with conventional configurations of LSFM due to the need for bulky objectives with short working distance. To overcome this issue, use of a single objective with high NA has been proposed [42–44].

Highly inclined and laminated optical sheet microscopy (HILO) was introduced by Tokunaga et al in 2008 [43]. As shown in figure 6(a), in this technique, an objective-type total internal reflection fluorescence microscopy (TIRF) setup was used to direct the incident laser light at angles slightly below critical angle to pass through the sample. The TIRF illumination method is only for the surface area of the sample. By this method, researchers could observe individual molecules on the cell surface. However, the application of TIRF is limited to surfaces because it is using reflection method. By using HILO this issue is eliminated. In this method, the incident beam was refracted at a large angle and obliquely sectioned cell sample near glass surface. The beam waist is affected inversely with the incident angle at the specimen, Then, the beam waist could be well-tuned by changing the incident angle. Once the incident angle moved toward the critical angle, the refracted light intensity increased as its thickness decreased at the critical angle, HILO converged to TIRF with illumination intensity four times higher than the incident light [43]. The advantage of this method is mainly due to use of the illumination beam passing through the center of sample plane that produces a powerful platform for 3D imaging.

Zoom In Zoom Out Reset image size

In 2011, Cella Zanacchi et al proposed individual molecule localization-selective plane illumination microscopy (IML-SPIM). They demonstrated three-dimensional (3D) super-resolution live-cell imaging for specimens of 50–150 μm sizes. In their method, SPIM has been coupled with far-field individual molecule localization. The improved signal-to-noise ratio (SNR) of selective plane illumination that allowed for nanometric localization of single molecules in thick scattering specimens without activating or exciting molecules outside the focal plane. They reported 3D super-resolution imaging of cellular spheroids [44]. They were able to continuously image whole sample to achieve 3D imaging with radial and axial resolutions of 63 and 140 nm, respectively.

In 2013, Gebhardt et al developed reflected light-sheet microscopy (RLSM) to reduce out-of-focus fluorescence. In RLSM, a thin light-sheet parallel to the imaging plane and close to the sample surface is generated by reflecting an elliptical laser beam with a small mirror (figure 6(b)). The thin light-sheet allows for an increased signal-to-background noise ratio superior to that in previous illumination schemes and enables imaging of single fluorescent proteins [42].

In 2013, Hu et al suggested light-sheet Bayesian super resolution microscopy (LSBM). They used light-sheet illumination along with a Pellin–Broca prism for super-resolution imaging (figure 6(c)). The prism illuminated a thin slice of the nucleus with high SNR. Their microscopy imaging method was coupled with a Bayesian algorithm to resolve overlapping fluorophores from high-density areas. They showed nanoscopic features of the heterochromatin structure for the first time in both fixed and live human embryonic stem cells. Due to enhanced temporal resolution, dynamic imaging of heterochromatin with a lateral resolution of 50–60 nm on a time scale of 2.3 s was illustrated experimentally [46].

In 2014, Chen et al proposed Lattice LSFM, which utilized a 2D optical lattice to enable higher acquisition speed for studying single molecule dynamics, along with a high resolution for large specimens [45]. Figure 6(d) depicts the schematic of Lattice LSFM. They used a pair of cylindrical lenses in one direction to stretch the input beam, and another pair of lenses in the perpendicular direction for light compression to produce the lattice. A spatial light modulator (SLM) was mounted at the sample plane to create the desired lattice pattern. Lattice LSFM had two-operation regime: the super-resolution regime, which generated 3D images beyond the Rayleigh criterion and the dithered regime for a rapid acquisition of large specimens. Some advantages of the method are high-speed acquisition, capturing larger and more densely fluorescent specimens, single-molecule sensitivity and low photobleaching and phototoxicity.

Gustavsson et al presented tilted light-sheet microscopy with 3D point spread functions (TILT3D) in 2018 (figure 6(e)). TILT3D was designed to image thick cells with features like low-background light, 3D super-localization of single molecules and 3D super-resolution. TILT3D combined a tilted light-sheet illumination strategy with long axial range point spread functions (PSFs). The axial positions of single emitters were encoded in the shape of each single-molecule image rather than in the position or thickness of the light-sheet. Therefore, very thin light-sheet could not be achieved. TILT3D was built upon a standard inverted microscope and had minimal custom parts. A simple and flexible 3D super resolution imaging with tens of nm localization precision throughout thick mammalian cells was reached with TILT3D [47].

Furthermore, the combination of LSFM with STED [38, 39] has been reported to improve the resolution of LSFM. In STED microscopy method, a fluorophore must be employed as an active element during imaging process. The STED method uses stimulated emission to achieve super resolution through depleting unwanted excited fluorophores very locally around the focus center. Friedrich et al combined STED method with SPIM to produce super resolution imaging of model organisms and tissues. They used a modified optical pathway of an original STED method by overlapping the excitation beam with a depleting laser beam with TEM₀₁ mode (figure 6(f)). They significantly enhanced the original advantages of this technique and enabled fast imaging of biological tissues with axial resolution beyond diffraction limits [38].

Recently deep-learning has been employed with LSFM. In this way, deep-learning super-resolution light-sheet add-on microscopy (Deep-SLAM) was developed by Zhao et al in 2020 [48]. They used an add-on that converted a conventional inverted microscope to a LSFM as an efficient and cost-effective approach for super resolution imaging.

Yang et al presented DaXi, that is a single-objective light-sheet microscope using oblique plane illumination. This device benefits from several features including (i) a wide field of view whit high-resolution imaging, (ii) fast volumetric imaging over large volumes, and (iii) high throughput multi-well imaging [49]. Their design could produce a resolution of 450 nm laterally and 2 μm axially over an imaging volume of 3,000 × 800 × 300 μm.

Chen et al integrated structured illumination microscopy (SIM) and light-sheet fluorescence microscopy (LSFM) to enable three-dimensional (3D) imaging using multidirectional structured illumination [50]. The multidirectional structured illumination could be successfully implemented in oblique plane microscopy, a LSFM technique that employs a single objective for excitation and detection in a straightforward manner. They demonstrated isotropic lateral resolution below 150 nm. Compared to traditional SIM systems, lower phototoxicity was achieved with volumetric acquisition speed exceeding 1 Hz.

The incredible versatility of LSFM allows for its deployment in various biological case studies. LSFM offers several benefits over conventional optical imaging methods. The main advantage of using LSFM in biology is mainly associated with high reduction of photo-toxicity and undesired illumination area owing to the use of selective illumination. Hence, LSFM allows for long-term imaging that is required for studies related to developmental biology. LSFM offers a gentle 3D imaging approach for a wide range of sample sizes from a micron-sized single cells [34, 51, 52] to a centimeter-sized whole organ [11]. It provides an appropriate method for studying many biological processes that occur in a 3D environment such as developing embryos and interactions of neurons. Unlike point scanning procedure in confocal and multiphoton microscopies, the entire plane is illuminated and imaged simultaneously in LSFM. This allows for capturing fast biological processes like brain functional imaging [52]. LSFM offers an opportunity for 3D imaging with isotropic resolution with the same lateral and axial resolutions due to multi-view imaging [11]. Decoupled illumination and detection optical path in LSFM results in a comfortable combination of this method with other microscopy techniques to enhance image quality.

Sample mounting is the most noticeable challenge for user of LSFM. In fact, a standard mounting approach still does not exist for LSFM. A researcher should consider a proper mounting tactic prior to imaging considering the LSFM configuration. Depending on sample size and required conditions for sample mounting, different optical arrangement for LSFM have been developed that will be discussed in section 4. These configurations include a wide variety of designs to hold samples in different sizes including few micrometer-sized specimen in Petri dishes Like light-sheet for Bayesian microscopy (LSBM) or on a conventional glass slide like inverted SPIM (iSPIM [53]), to centimeter-sized sample like a whole mouse brain with ultramicroscopy [11].

The large dimensions of a sample might be considered as another limiting factor. Large samples are quite opaque and scatter light that decrease imaging quality. Large samples should be threated through tissue clearing process prior to imaging [54] for deeper microscopy of opaque structures to obtain more details.

Although, some reports showed the implementation of this method for 3D-imaging based on elastic [55] and inelastic (Raman scattering) scattering [56, 57], LSFM configurations are mainly designed to image fluorescently stained specimens.

Unlike the typical horizontal configuration of two orthogonally mounted objectives in basic SPIM, several illumination arrangements have been reported for LSFM. Depending on application and requirement for sample mounting, these arrangements are different in how and how many objectives are employed. Some of the most well-known methods are briefly reviewed in this section.

In conventional orthogonal arrangement of illumination and detection objectives of LSFM, placement of commercially available high NA objectives is difficult due to their short working distance and bulky design. Moreover, mounting a sample through standard biological method is not possible. Hence, many authors have altered the settlement geometry of objectives. Gebhardt et al [42] suggested a novel configuration of reflected light-sheet microscopy (RLSM) using two high NA objectives in opposite, figure 6(b) Thin light-sheet is brought to imaging plane and close to sample surface by a polished AFM cantilever used as a reflector. As shown in figure 6(c), in light-sheet Bayesian microscopy (LSBM), a prism is placed after illumination objective to redirect light-sheet horizontally onto a sample at the imaging plane of the detection system [58]. Additionally, the Bayesian algorithm effectively resolves sub-diffraction cellular structures. This configuration is more suitable for situations in which the sample needs to be held in Petri dishes.

Objectives arranged in V-shape configuration are well suited for microscopies that require standard sample mounting methods. Inverted or dual-view inverted SPIM (iSPIM\diSPIM) was introduced from Hari Shroff group [32, 53] for single-cell imaging or any samples that need to be mounted on a conventional glass coverslip. Figure 7(a) depicts the schematics of objectives in diSPIM. In this method, two similar objectives are placed above a sample in a V-shape arrangement. Excitation and imaging alternate between these two objectives. Hence, isotropic resolution with equal lateral and axial resolution is achieved in this method.

Zoom In Zoom Out Reset image size

Wu et al developed the dual-view light-sheet microscopy and computational method to a simultaneous triple-view acquisition microscopy configuration (figure 7(b)) [59]. Similar to iSPIM, open-top light-sheet (OTLS) microscope takes advantage of V-shape arrangement of illumination and detection arms (in an upside-down direction in comparison to iSPIM) (figure 7(c)) [60–62]. In a developed open-top light-sheet microscope by McGorty et al, a water prism compensates aberrations introduced as a result of 45° orientation of coverglass to illumination and detection path [63]. This configuration is suitable for sample mounting over glass slide, multiwell plates, and microfluidic chips. Alternatively, as depicted in figure 7(d), in a diagonally scanned light-sheet microscopy, V-shape configuration performs by inclining the sample instead of optics [64].

In single-objective LSFM configurations, a thin sheet of illumination light is generated using the same high NA objective that simultaneously collects fluorescence light. The most significant feature of this approach is the possibility of implementation of widely accessible inverted microscopes. Besides, the possibility of using high NA objectives, they are compatible with standard biological sample mounting like slide glass, petri dish, and multiwell plate. Dunsby et al corrected the different axial positions of the image plane by adding another microscope to tilt the imaging plane [65, 66]. Therefore, in-focus imaging of a whole oblique illuminated plane of sample is acquired. They illustrated the application of their adapted oblique plane microscopy for sample mounting in commercially available glass-bottomed 96-well plates [67]. The scanning procedure in this method is effortless. Due to the oblique direction of illumination, the sample horizontally scans through the light-sheet. Hence the optical path is constant during the scanning procedure, and the illuminating plane is always in focus. There is no requirement for using a water chamber or moving the detection objective. Turning the light-sheet to the horizontal plane, i.e. perpendicular to the detection arm, is the other solution to get rid of the difficulty of the oblique imaging plane, figure 7(e). In single-objective cantilever selective plane illumination microscopy (socSPIM), light-sheet is 90° rotated by an AFM tip. The authors demonstrated the application of the setup in the case of mounting the sample on a multiwell plate [68]. Similarly, Light-sheet could also be turned to the horizontal plane in a single objective configuration through the use of a micromirror [69, 70]. Alternatively, the imaging plane is rotated by a small mirror rather than rotating the illumination plane to image a single vertically illuminated plane in a PRISM method [71].

Due to light absorption and scattering, and degradation of the quality of images obtained from the depths of a sample, single view imaging LSFM does not allow for imaging of whole large multicellular organisms. Multiview LSFM (figure 7(f)) [72], in which the sample is rotated, and 3D-stacks are acquired from multiple view angles, could partially enhance the capability of imaging dense large specimens. However, it has some disadvantages including low temporal resolution of imaging of live specimens, induced vibration in the specimen due to mechanical rotation, complex and time-consuming algorithm that needs to merge several stacks from different views to reconstruct a final 3D image, and the extra illumination dose that increase photodamages. In multidirectional selective plane illumination microscopy (mSPIM) [70], absorption and scattering of illuminated light in one direction is compensated by illuminating the sample from opposing direction (figure 7(g)). This solution provides uniform illumination with fewer stripes artifacts across whole FOV. Unlike the ultramicroscopy [11], dual-sided illumination is performed separately - i.e. first from one side and then from the other side - in this method. The scattering of each beam in simultaneous illumination of a sample by two counterpropagating beams results in widening of an overlapped light-sheet in comparison to the width of the light-sheet in single side illumination. However, still the image quality of planes that are closer to the detection objective is better than the deeper planes. A more efficient solution would be to acquire multiple stacks from different views simultaneously and then reconstructing the final 3D image by merging these stacks. As depicted in figure 7(h), in simultaneous multi-view (SiMView) light-sheet fluorescence microscopy implementation, image stacks are acquired by sequentially alternating light-sheet excitation in two opposing scanned light-sheets illumination arms [73]. Two opposite detection objectives gather the fluorescence light. As a result, four complementary views of each depth plane of the specimen are captured in two sequentially illuminations and two simultaneous acquisitions for each one. In state of the art isotropic multiview (isoView) LSFM configurations, four objective lenses are utilized both for illumination and detection to capture stacks of sample from four different views [33]. The arrangement of objectives is depicted in figure 7(i). This method consists of four identical objectives in two orthogonally arranged optical arms, which are simultaneously utilized to illuminate the specimen with a scanned light-sheet and gather the emitted fluorescence of the specimen. To perform volumetric imaging, light-sheets are swept across a specimen, and illuminated planes are kept in focus by translating the detection objectives with piezoelectric stages.

It is worth mentioning that one microscope lens together with a conventional (cylindrical) lens is often sufficient for LSFM, when a low aperture is preferred for illumination. Indeed, in case of making a light sheet thinner (with high NA), an objective lens is employed.

Objective configurations with higher number of illumination arms could enhance the penetration depth of excitation light into dense and larger samples and reduce stripes in a captured image. However, such configurations are usually costly and need precise alignment, which makes these methods not easy to implement especially for whom, are not an expert in optics. These are banning issues, especially for those looking for building their own system based on an open platform of SPIM (OpenSPIM) [74]. Hence, instead of increasing the number of illumination arms, many researchers have proposed various forms of illumination light-sheets to enhance the quality of captured images that are discussed in the following section.

An immersed coupling strategy for middle and end objectives of an oblique light-sheet microscope was achieved by Gong et al to enhance alignment and efficiency of coupling [75]. Their design showed higher performance than the conventional designs that used only air objectives in resolution and light-collection power. They further demonstrated the capacity of this configuration to capture large fields-of-view images in combination with a camera with built-in electronic binning.

Xiong et al proposed a mirror-enhanced scanning LFM (MiSLFM) to achieve high-speed 3D imaging and high axial resolution with a single objective through exploiting the extended depth of field for LFM [76]. They used a tilted mirror that was placed below samples. Superior axial resolution was obtained that enabled acquisition more robust blood cell tracking in zebrafish larvae at high speed.

Some quality features of LSFM images like penetration depth; contrast and present of stripes could be strongly modulated by the method of sample illumination. The penetration of excitation light is affected by intrinsic spreading nature of light, along with absorption and scattering of light once propagating through a sample.

As for the contrast and axial resolution, it is ideal to produce a highly focused light-sheet with minimum thickness; however, the FOV is simultaneously minimized and limited to few micrometers. A conventional LSFM illumination beam has a Gaussian intensity profile with a minimum thickness at its focal point, while the beam diverges at both sides of the focal point. The Rayleigh length is introduced as a distance where beam thickness remains approximately constant. If the specimen is larger than the Rayleigh length, images would be sharply appeared in focus only within the Rayleigh length. Hence, to provide a larger FOV, the width of the illumination beam has to be increased. Consequently, the axial resolution is degraded. Buytaert and Dirckx [77] resolved this problem by stitching columns of single images together, which each image is extended over the Rayleigh length of the illuminating beam. However, this method is time-consuming, and it also does not seem to be suitable for imaging thick samples since the quality of captured images is strongly degraded due to scattering through thick sample. Moreover, interference of various parts of the excitation beam after passing through inhomogeneities or absorbing structures in a sample creates stripes patterns along light propagation path. These artifacts create serious issues for imaging of dense and large specimens. Such stripes are more pronounced in configurations with side illumination. Multiview fusion and multidirectional illumination methods enables to decrease stripe artifacts, however, the adoption of these methods are limited by some disadvantages. The Multiview fusion could not be employed for all samples and sample holding conditions due to the need for rotating sample for imaging from different views. In this way, sophisticated reconstruction algorithms with powerful processing system are required to merge several stacks obtained from different views. In addition, illumination of a sample from different views induces high level of photodamage. The multidirectional illumination method is usually costly and need precise alignment. Hence, these methods could not be employed for all samples. Thus, in the follow section, we will focus only on different approaches of single side illumination with single view imaging for LSFM that are applicable for a wide range of samples.

In the first step to manipulate illumination, static light-sheet with a cylindrical lens could be replaced with scanning light-sheet that forms through scanning a beam in a transverse plane [16]. Thus, the area behind the obstacles is illuminated from different directions and the uniformity of illumination is enhanced. Moreover, the presence of artifacts due to stripes produced by inhomogeneities in a sample is more profound when illumination is spatially coherent, i.e. static light-sheet, in comparison to decrease coherency in scanning light-sheet illumination [78].

Fahrbach et al proposed illumination with Bessel [79, 80] and sectioned Bessel beams [81] as non-diffracting beams. The unique non-diffractive propagation and self-healing properties of these beams are used to overcome scattering and produce larger FOV. Bessel beam's thin core is accompanied by a set of rings producing a large amount of unwanted out-of-focus excitation that results in creation of undesired emissions with subsequent image contrast reduction. It has been attempted to eliminate such unwanted emissions using confocal-line detection [82] and two-photon excitation [26]. In the first method, synchronization between beam position and active line on the camera is crucial. In the second method, the need for using expensive Titanium-Sapphire laser and the possibility of photodamages induced by high power illumination are potential prohibiting issues.

Lattice LSFM, suggested by Betzig Lab [45], is another advanced approach for suppressing the defects of side lobes while benefiting from non-diffractive propagation and self-healing properties of Bessel beam's core. In Lattice LSFM, Bessel's rings could be eliminated through generating a linear array of interfering Bessel beams to make ultra-thin cores that are particularly suitable for imaging single cells. A substantial limitation of Lattice LSFM is owing to the use of only a few quasi-discrete wave-vectors of the ring at a generalized pupil, which considerably reduces beam self-healing property. Therefore, lattice light-sheet could be affected by perturbations produced by a sample that prevent the application of this technology for imaging thick samples.

Wang et al published an article investigating the effect of near-infrared structured illumination on light sheet microscopy [83]. They developed near-infrared II (NIR-II) (1,000 to 1,700 nm) structured-illumination light-sheet microscopy (NIR-II SIM) with ultralong excitation and emission wavelengths of up to ∼1,540 and ∼1,700 nm, respectively. This method allowed for imaging of large volumetric three-dimensional (3D) tissues with deep-axial penetration depths. Furthermore, it diminished background and improved spatial resolution by approximately two folds.

Hung et al combined sensorless adaptive optics (AO), in-situ 3D-PSF calibration, and a single-objective lens inclined light sheet microscope (SOLEIL), termed (AO-SOLEIL), to mitigate deep tissue-induced deteriorations [84]. They applied AO-SOLEIL on several dSTORM samples including brains of adult Drosophila. They observed a 2x improvement in the estimated axial localization precision with respect to widefield without aberration correction while they used synergistic solution.

Dohlakia and his research group proposed the utilization of an Airy beam in LSFM configuration in several publications [85–90]. In comparison to the Bessel beam, Airy beam has less side lobes owing to its asymmetric intensity distribution. Hence, the effect of side lobes on contrast decline is decreased [91].

Akhte et al reported an open-source toolbox for beam shaping by a spatial light modulator (SLM) to be applied within OpenSPIM platform [92]. The toolbox is able to generate several patterns for a wide range of illumination beams, including static and scanning Gaussian beams, Bessel, lattice, and Airy beams.

Producing a robust thin light-sheet over a large FOV is still challenging in LSFM. Recently, some researchers have investigated the role of different illumination beam on image quality

Kafian et al investigated the effect of static, Gaussian, Bessel, and 1D Airy beams and scanning Gaussian, Bessel, and 2D Airy beams on obtained images in terms of contrast, penetration depth, and the presence of stripes artifacts in LSFM, using simulation and experimental tools [91]. Results revealed that due to decreased spatial coherency and self-healing feature, scanning 2D Airy light-sheet is less affected by inhomogeneities of sample and provides deeper penetration length through a sample with higher contrasts and uniform resolution over a wide field-of-view. In addition, it was found that scanning 2D Airy light-sheet is more appropriate choice for illuminating large, i.e. hundreds of microns, and dense samples like tumors.

In a parametric study, Remacha et al suggested helpful guidelines to choose the best suitable illuminating beam and its operating parameters for various applications [93]. They also numerically studied the role of different conventional illumination light-sheets on axial resolution of LSFM. Based on their findings, Gaussian light-sheet is the most appropriate approach for imaging of large, light-sensitive, or thickly labeled samples.

Despite all benefits of using structured beams in illumination, the reduction of contrast in captured images is still remaining as an issue. Besides, the requirement of using costly beam shaping devices and with complex configuration, e.g. extra optical components needed in beam shaping unit, are considered as limiting factors. An alternative for beam manipulation and structuring is pivoting ligh-sheet, especially when combined with multidirectional illumination [94]. This method allows for suppressing stripe artifacts to enhance image quality. Light-sheet direction should be quickly altered using a scanning device, e.g. scanning mirror, within a limited range. Therefore, excitation light is capable of illuminating dark regions behind an obstacle. Pivoting light-sheet approach is not only used for static illumination, but also it could be combined with a scanning illumination method like digital scanned light-sheet microscopy (DSLM) [95]. Sancataldo et al reported some advantages provided by the use of Acousto Optic Deflectors (AODs) instead of a typical scanning mirror for the generation of MHz pivoting light-sheet. They depicted that seven pivoted static light-sheets is indistinguishable from continuously pivoted light-sheet [96]. Nowadays, MEMS devices with affordable prices are available at the market. The integration of LSFM configuration with MEMS-based scanning devices provides a great opportunity for miniaturization of LSFM at lower prices [97]. Itoh et al demonstrated the combination of structured and pivoting illuminations for high quality rapid image acquisition [98] using 1-D spatial light modulator (SLM). In this way, acquired images have less stripe artifacts due to pivoting illumination. Image contrast is increased in comparison with structured illumination.

In overall, designing an easy-to-perform illumination unit at an affordable price to produce high-quality images is still highly demanded. In our opinion, among the above-mentioned LSFM configurations, pivoting static illumination provides the best features in terms of cost, implementation difficulty, and performance. However, for stripes suppression, this solution is not as capable as multidirectional illumination.

Also, the Raleigh length of pivoting static illumination could be enhanced just through applying little changes in the optical setup without increasing the beam waist to preserving the axial resolution [99]. Use of pivoting static illumination to make illumination unit as an add-on for conventional fluorescent microscopes, which are widely available in biological labs, can notably decrease the investment expenses for lab equipment. Usage of MEMS-based scanning devices to build a more compact setup for pivoting static illumination is an enabling approach to decrease the size of illumination unit. An alternative approach for pivoting static illumination is the use of Multiview imaging to reduce stripes artifacts. Additionally, the rotation of sample in Multiview imaging improves the image quality of depth planes that are far away from detection objective. Usually, images from depth planes are degraded by scattering of emission light propagating through the sample to reach the detection objective.

In general, any absorptive or inhomogeneous along illumination path induces stripes artifacts. Such artifacts are serious prohibiting issues to use LSFM for imaging of dense and large specimens.

Stripes are especially pronounced due to sample's unidirectional illumination from side view. Stripes artifacts can be mitigated or completely removed via extra illumination from other sides of a sample or through image processing methods. The first approach includes multiview imaging [72] or multidirectional illumination solutions [11, 33, 94], as discussed in section 4: Configurations of LSFM . Regarding the second approach, several destriping methods that some of them were initially developed for other microscopy methods, could be also employed for LSFM, as explained below.

Wu et al 's proposed a method to mitigate both identical and different stripes at different layers of a sample 3D stack [162] for Optical Coherence Tomography Angiography (OCTA) images.

Pollatou et al proposed an automatic stripe removing method for stitched biological images based on finding the location of stripes and modeling their background [163]. An algorithm, called Variational Stationary Noise Remover (VSNR) was developed as a free Fiji plugin [164]. The authors depicted the application of their method with images acquired using LSFM.

Pavone et al presented the optical and post-processing methods for removing strips in a review article [165].

In addition to microscopic images, stripe artifact is a severe limiting issue in remote sensing systems. Several studies provided solutions for this problem [166, 167], which could be also utilized for LSFM images with minor modifications. As an example, Chang et al estimated complex stripes distribution by wavelet-based deep convolutional neural network method for destriping remote sensing systems' images [168]. They improve the performance of their algorithm by simultaneously modeling the stripe and image components.