CHAPTER 1: Fast Volumetric Imaging Using Light-sheet Microscopy. Principles and Applications

The parallel development of genetic engineering and microscopy techniques has propelled biology into a new era. It is now possible to genetically encode fluorescent proteins in targeted cell populations and then to monitor optically their dynamics and activity in vivo with exquisite precision. Such approaches have become a central tool to probe the structure, development and function of biological tissues.

In recent decades, great efforts have been made to improve the performances of fluorescence imaging systems, the challenge being how to record the fluorescence signal of a specific section/volume within a living specimen at high spatial and temporal resolution while minimally damaging the tissue. As we will show, light-sheet fluorescence microscopy (LSFM) surpasses classical optical methods for each of these criteria. Combining massive parallelization of data acquisition and efficient optical sectioning, LSFM allows long-term and fast volumetric recordings at high spatiotemporal resolution with minimal photodamage.

Since the publication of the seminal article by Stelzer's group in 2004,¹  the impact of LSFM has been significant in numerous fields of biology such as neuroscience, anatomy, physiology, developmental biology and cell biology (Figure 1.1). This technique has been used to image biological tissues at scales ranging from subcellular compartments²  to entire organs;³  it has enabled the long-term monitoring of entire developing embryos over days,^4,5  and has also provided video-rate four-dimensional (4D) movies of rapid physiological processes such as heart beating in zebrafish.⁵⁹  The power of LSFM lies in its unique amenability to diverse biological questions. However, each of these questions generally requires specifically adapting the microscope to the sample and to the type of information that one wishes to obtain. Although a standard LSFM instrument is relatively inexpensive to build and is now commercially available, it can also be combined with more advanced imaging techniques, such as super-resolution,⁶  two-photon excitation⁷  and structured illumination.⁸ 

In this chapter, we show how LSFM has emerged as a novel and powerful alternative to confocal and two-photon microscopy. After having established the principles underlying this new imaging method, we will review some of the most significant improvements that have been developed in recent years to address specific biological questions. Finally, we present in detail one particular application of LSFM, namely the functional imaging of the entire zebrafish larva brain.

Probing the physiology of living tissues in a minimally invasive way is arguably the central goal of optical microscopy in biology. Starting with the early microscopes designed in the seventeenth century, microscopy techniques have always been an important source of new discoveries for biologists. In the twentieth century, the fluorescent labeling of specific cellular structures was developed, giving rise to fluorescence microscopy. However, one important issue remained – how to record the fluorescence signal of a specific section/volume within a volumetric biological tissue with large contrast. In standard epifluorescence microscopy, the entire specimen is illuminated, hence the collected fluorescent photons originate from both the in-focus and out-of-focus regions. These out-of-focus photons inevitably blur the imaged plane such that, for thick samples, the contrast vanishes. During the second half of the twentieth century, optical techniques were therefore developed in order to selectively image a thin section within a specimen, a process called optical sectioning. Two different approaches were successively proposed: confocal imaging and two-photon microscopy.

In 1955, Marvin Minsky, who later worked on artificial intelligence, laid the foundation of the confocal microscope when he was a young professor: One day it occurred to me that the way to avoid all that scattered light was to never allow any unnecessary light to enter in the first place. […] There is no way to eliminate every possible such ray, because of multiple scattering, but it is easy to remove all rays not initially aimed at the focal point; just use a second microscope (instead of a condenser lens) to image a pinhole aperture on a single point of the specimen.⁹  The confocal microscope thus uses a pinhole conjugated to the focal point of the objective to block all photons originating from outside the focal volume. This simple method allows for the recording of the fluorescence signal from only one small focal volume of the sample. To build a complete image, one sequentially scans this elementary volume throughout the entire sample by moving the observation (pinhole) and the excitation arm together. Confocal microcopy has several limitations: (i) the need to scan throughout the sample severely limits the acquisition frame rate; (ii) the photoefficiency is intrinsically low since the entire specimen is continuously exposed to the excitation beam even if only one voxel is imaged at a given time; and (iii) confocal microscopy generally uses visible light and therefore has a limited penetration depth in heterogeneous (light-scattering) samples.

Fluorescence can be induced by the absorption of one photon of a given energy or by the simultaneous absorption of two photons of half the energy (and thus twice the wavelength). In the latter mode, the fluorescence signal varies quadratically (rather than linearly) with the intensity of the excitation light, and it is extremely low under standard illumination conditions. Hence a molecule of the natural chromophore rhodamine B excited by sunlight will experience a two-photon absorption transition every 10⁷ years.¹⁰  Observing this phenomenon therefore requires the use of spatially and temporally focused light. In two-photon microscopes, a near-infrared laser beam with 100 fs long high-energy pulses at a repetition rate of 80 MHz is focused through a high numerical aperture objective.

The quadratic dependence of the fluorescence emission with the intensity yields a natural optical sectioning mechanism: only at the focal point of the objective does the spatial and temporal focusing yield a sufficient peak intensity to evoke significant fluorescence. The use of infrared light further minimizes scattering in the tissue, which allows for the imaging of thicker samples, and also limits photodamage in comparison with confocal microscopy.

Since the invention of the two-photon microscope in 1990,¹¹  numerous applications have been developed in biology, especially in neuroscience,^12–14  in immunology¹⁵  and in developmental biology.^16,17 

Although they use different approaches for optical sectioning, both confocal and the two-photon microscopes share the need for faster scanning across the specimen in order to acquire a complete image. The laser dwell time τ_dwell at each voxel needs to be large enough to allow for a significant number of fluorescent photons to be emitted (typically τ_dwell = 0.1–1 µs). This in turn imposes a compromise between the size of the imaged region (number of pixels N) and the acquisition frame rate f_acq, such that Nf_acq = 1/τ_dwell.

This drastic limitation in terms of the data throughput of point-scanning imaging motivated the development of light-sheet fluorescence microscopy (LSFM). In LSFM, the illumination is provided by a micrometer-thick sheet of light projected across the sample while the fluorescent photons are collected at 90° with a camera. This approach allows for a massive parallelization of the data collection, as all pixels from a single plane are now imaged simultaneously (see Figure 1.2). The maximum frame rate becomes f_acq = 1/τ_dwell and is in practice limited by the camera dynamics (typically 100 Hz for 10⁶ pixels).

Beyond this approximately 100-fold increase in data throughput, LSFM also offers unique performance in terms of photodamage. Indeed, in this particular geometry, only the imaged plane is illuminated, which maximizes the photonic efficiency. This contrasts with confocal microscopy, in which the sample is evenly illuminated while only photons originating from the focal plane are used. As photodamage is mostly associated with non-linear processes, spreading the light dose across time and space, thus reducing the peak intensity, provides a further advantage compared with raster scanning techniques (see Figure 1.3).

In the first modern LSFM instruments, described by Huisken et al. in 2004,¹  a cylindrical lens was used to focus a laser beam and to form a static light sheet. The sample was then sequentially moved along the vertical axis to create a volumetric image of the tissue. It was then proposed to form the light sheet by rapidly scanning a laser beam across the sample.¹⁸  This so-called digitally scanned light-sheet fluorescence microscope (DSLM) offers several advantages: each line in the specimen is illuminated with the same intensity, which can be beneficial for quantitative imaging of large specimens; the microscope does not rely on apertures to shape the beam profile, which reduces optical aberrations; and the DSLM mitigates striping artifacts induced by scattering/absorption objects in the tissue and thus provides a better image quality. However, the use of a DSLM results in a higher laser peak power for a similar average illumination and thus greater photodamage.

Before going into the technical details of the performance of light-sheet microscopy, let us take a brief historical look at its origins, because the foundation of this new imaging technique in fact goes back a long time.

In 1902, Siedentopf and Zsigmondy published an article in Annalen der Physik¹⁹  describing a new method for the optical measurement of the size of gold particles. They projected sunlight through a slit aperture in the focal plane of an observation objective. The light in the sample was thus projected orthogonally to the observation objective (Figure 1.4), and they collected the scattered photons. They called this technique ultramicroscopy owing to its ability to image objects smaller than the diffraction limit.

Although in 1925 Zsigmondy received the Nobel Prize for Chemistry in part for this invention, it remained confined to the domain of colloidal physics for a long time. In 1993, Voie, Burns and Spelman²⁰  reintroduced this optical method, albeit in a completely different context, renaming it orthogonal-plane fluorescence optical sectioning (OPFOS). Spelman's group developed the OPFOS device and used it to optically section, for the first time, whole fluorophore-stained and cleared cochleas,^20–22  stating: Another approach to optical sectioning involves the use of a planar illumination beam. Planar illumination has been used in other imaging modalities such as flow cytometry and flow visualization. The illumination beam is focused into a plane with a cylindrical lens, and aligned to be co-planar with the depth of field of the imaging detector. The beam is thus orthogonal to the imaging axis. The region where both the imaging system and laser illumination focus coincide is the system focal zone.

In 1994, Stelzer's laboratory was trying to improve the axial resolution of confocal microscopy. They developed an oblique illuminating confocal microscope called a confocal theta microscope.^23,24  Their 1995 paper²⁴  cited Voie et al.’s work on OPFOS, and confocal theta microscopy appeared to lay the foundation for their subsequent version of a light-sheet microscope device called a selective- or single-plane illumination microscope (SPIM).

In 2004, a paper by Stelzer's group¹  demonstrated the usefulness of light-sheet fluorescence microscopy for investigating embryonic development. This paper also showed images of developing embryos of medaka (the small fish Oryzias latipes) and Drosophila melanogaster embryos and ganglion cells monitored for 17 h. The spatial resolution was 6 µm with a field of view of 1.5 × 0.9 mm. Although Voie and colleagues published several articles^20–22  claiming that light-sheet microscopy was very efficient, it was Huisken et al.’s paper¹  that triggered the rapid development of light-sheet microscopy among biologists.

In this section, we provide a simple evaluation of the spatial resolution in LSFM and the way in which it depends on the optical characteristics of the illumination and detection optics.

In light-sheet microscopy, a cylindrical or scanned spherical Gaussian beam is used to illuminate the sample (see Figure 1.5). In the focal zone, the laser spot exhibits a Gaussian intensity distribution. In the x–z plane, the intensity profile can be expressed as

where the beam width w(x) follows the equation of a hyperbole:

At the waist x = 0, the intensity is maximum and the beam radius is w₀ = λ/πNA, where λ is the excitation wavelength and NA the numerical aperture of the illumination arm. Far from the waist, the beam diverges with an angle θ such that NA = n sin θ, where n is the refractive index. The focal region of the beam, called the Rayleigh range, is the region around the waist within which the beam radius is less than . This region is bounded by ±x₀, where x₀ = πnw₀²/λ. Hence large NA values yield small waist radii but also a small Rayleigh range.

In standard bright-field microscopy, the spatial resolution reflects the shape of the PSF (point-spread function) of the objective and the detector, i.e. the response of the optical system to a point-like object. The lateral resolution (in the focus plane), is given by r_lat ∼ λ/NA, where λ is the emission wavelength and NA is the numerical aperture of the objective. Its axial resolution (along the optical axis) is r_axial ∼ λ/NA², and is thus always larger than the lateral resolution.

In DSLM, the lateral resolution is set by the detection objective and is therefore similar to that of a classical bright-field microscope: r_lat ∼ λ/NA_det. Note, however, that this expression is true only as long as the pixel size of the detection camera remains smaller than r_lat/2. In contrast to two-photon and confocal microscopy, in light sheet microscopy the lateral and axial resolutions are decoupled: the latter is essentially controlled by the thickness of the light sheet. Using a high-NA illumination objective, it is possible to produce a very thin light sheet, whose axial resolution r_axial = λ/NA_ill is comparable to the lateral resolution. DSLM thus offers the unique possibility of producing isotropic resolution. However, such configurations have very limited fields of view. Indeed, as shown in the preceding section, a very thin Gaussian beam has a small Rayleigh length, i.e. the region over which the beam is highly focused is extremely narrow.

In light-sheet microscopy, any configuration therefore reflects a trade-off between the axial resolution and the accessible field of view. A small calculation can be done in order to evaluate the best accessible axial resolution within a certain field of view, i.e. up to a given distance l from the waist. This axial resolution, denoted W, is the width of the Gaussian beam used to form the light sheet at a distance l from the waist, and it reads [see eqn (1.2) and Figure 1.5].

where θ is the beam divergence angle, such that NA = n sin θ. In the small angle limit, we have

Therefore, θ is minimal for and its value is then . More precisely, the exact minimization of W leads to . This trade-off is illustrated in Figure 1.6, which shows, for each choice of NA, the associated axial resolution and accessible field of view.

In this section, we present the most significant improvements that have been implemented to increase the spatial or temporal resolution, enhance the penetration depth, limit photobleaching or adapt the geometry to accommodate a specific specimen.

Increasing the resolution is fundamental in order to probe biological structures at the subcellular scale. Many studies have demonstrated that LSFM permits the imaging of small objects such as mitochondria, microtubules,²⁵  single fluorescently labeled DNA binding proteins,²⁶  single molecules²⁷  and chromosomes.² 

As mentioned in Section 1.3, by combining high-NA objectives in both the illumination and detection arms, it is in principle possible to obtain a very high and isotropic spatial resolution, albeit within a rather limited field of view. In practice, however, such configurations are prohibited by the steric constraints associated with high-NA (and thus massive) objectives. Two different strategies have been proposed to circumvent this issue. Galland et al.²⁷  used a single high-NA objective with a 45° micromirror (Figure 1.7a). Illumination and detection are thus performed through a unique objective. In a study by Gebhardt et al.,²⁶  two vertically opposed objectives (NA 1.4 and 1.35) faced each other and a thin light sheet was obtained by a small 45° mirror (Figure 1.7b).

Because of the divergence of Gaussian beams, it is impossible to maintain a sub-micrometer axial resolution over a field of view of tens of micrometers (see the preceding section). This limitation motivated the use of non-diffracting beams. Bessel beams display an invariant profile over the direction of propagation. Their cross-section is well described by a Bessel function, i.e. it contains a central spot and a series of annular rings of decreasing intensity. These extra rings carry a large fraction of the illumination power and are therefore highly detrimental in one-photon imaging, making the use of Bessel beams impracticable. However, in the two-photon regime, owing to the quadratic dependence of the fluorescence emission on the illumination intensity, their contribution to the fluorescence signal becomes negligible, as demonstrated by Betzig's group.²  This approach allowed the authors to image mitotic chromosomes with a 0.3 µm axial resolution over a 40 µm field of view. However, certainly the most impressive use of Bessel beams was reported in 2014, when the same group described the development of lattice light-sheet microscopy.²⁸  The trick here was to produce a linear array of visible Bessel beams, the period of which was adjusted to induce destructive interferences between the rings. This method yielded an ultrathin 2D optical lattice, which, after dithering, formed a uniformly thin light sheet and allowed diffraction-limited axial resolution over a field of view of 80 µm (see Figure 1.7c). They also demonstrated that with the lattice light-sheet microscope, photobleaching and phototoxicity were reduced by one to two orders of magnitude compared to those seen with a 1D scanned Bessel beam.

Rather than using non-diffracting beams, other groups proposed the use of electroacoustic lenses in the illumination arm in order to sweep the location of the waist rapidly along the propagation axis. This approach allowed for the production of uniform, divergence-free light sheets with 465 nm resolution over 50 µm.^29,30 

LSFM largely outperforms other fluorescence techniques in terms of the imaging speed for a given specimen section. To perform fast 3D imaging, one further needs to displace rapidly both the light sheet and the focal plane of the objective along the z-axis. This is generally done by simultaneously moving the objective with a piezo device and the light sheet with a scanning mirror.³¹  The motion of the objective is in practice limited in terms of speed and range by the piezoelectric devices. To circumvent this issue, Fahrbach et al.³²  used a remote focusing approach in which an electrically tunable lens, placed along the detection path, allowed them to displace the focal plane optically without any mechanical motion, allowing them to record up 30 volumetric images per second (Figure 1.8).

The parallelization of data collection provided by LSFM comes at a cost: non-ballistic fluorescent photons, i.e. those scattered by the tissue and collected by the camera, induce a diffuse background signal, which tends to degrade the image contrast. This issue may become detrimental for thick and strongly scattering specimens. Two strategies have therefore been developed with the aim of rejecting these photons: line confocal detection and structured illumination.

Line confocal LSFM relies on the same principle as confocal imaging. The idea is to use an optical slit (the 2D equivalent of a pinhole) conjugated with the digitally scanned laser beam to block any non-ballistic photons from impinging on the sensor. Silvestri et al.³  demonstrated how this method allows for the recovery of high-contrast subcellular resolution in thick tissue, such as an entire clarified mouse cerebellum. One very clever implementation of the line confocal method has recently been introduced that exploits the rolling shutter mode of SCMOS cameras to produce a virtual slit. In this mode, a few lines of the camera are exposed at each instant, and this region “rolls” across the sensor at constant speed to eventually a full frame. By continuously conjugating the beam and rolling shutter, the confocal effect can be obtained without the need for extra optical components.

Another approach to reject background signals is to implement structured illumination, a standard technique of contrast enhancement in fluorescence microscopy. This technique consists of illuminating the sample with a sinusoidal grating and recording several images for different spatial phase shifts of the projected pattern. One can then computationally recover the in-focus image using these different recorded frames, while eliminating out-of-focus signals. Keller et al.¹⁸  produced striped illumination patterns by rapid temporal modulation of the laser intensity during scanning. This approach allowed for a 2–3-fold contrast enhancement in developing Drosophila embryos. However, this technique requires at least three frames to be recorded for each time step, and thus reduce the accessible frame rate by the same factor.

The ability to image deep down in tissue is of paramount importance to a number of bioimaging applications. As discussed in the previous section, the scattering of fluorescent photons degrades the image contrast by adding a diffuse background to the in-focus image. Another issue lies in the blurring of the light sheet itself, which effectively reduces the axial resolution. This problem motivated the recent development of two-photon light-sheet imaging, which uses a near-infrared pulsed laser. This alternative offers two advantages compared to standard one-photon DLSM. First, light scattering is significantly reduced in the near-infrared compared with the visible range, thus enhancing the penetration depth of the light sheet. Second, in this non-linear regime of excitation, the relative contribution of the scattered photons to the overall fluorescence signal is highly reduced, so that the axial resolution is preserved. This approach, pioneered by Fraser's group in 2011,³³  was shown to provide a significant gain in axial resolution when imaging, e.g. Drosophila embryos (Figure 1.9). In the specific context of functional imaging, the use of a near-infrared source can also be beneficial as this wavelength range is outside most animals' visible spectrum, and thus does not interfere with the visual system.⁷ 

One important limitation of this technique, however, needs to be mentioned: inducing two-photon fluorescence in the LSFM geometry requires the use of high laser power, which results in higher photodamage compared with one-photon LSFM for similar signal-to-noise ratios.³⁴ 

One of the major practical limitations of LSFM is that it necessitates optical access of the specimen along two orthogonal axes. This requirement restricts its use to relatively small systems: for instance, it is incompatible with functional brain imaging in the mouse cortex. To circumvent this limitation, single-objective LSFM has been proposed in the last few years. Building on the highly inclined laminated optical sheet microscope invented by Tokunaga et al.,³⁵  Dunsby³⁶  developed a light-sheet fluorescence method, called oblique plane microscopy, that used a single high-NA objective both to produce an oblique light sheet in the sample and to collect the fluorescent photons. Because the light sheet does not align with the focal plane of the imaging system, another objective is used to rotate the imaging plane. This complex optical path results in a low efficiency in photon collection, and a subsequent limited accessible frame rate. Recently, Bouchard et al.³⁷  proposed a novel single-objective light-sheet design called the swept, confocally aligned planar excitation microscope, in which the second objective is replaced by a scanning mirror. This new LSFM instrument offers a better acquisition rate, and was shown to permit the recording of neuronal activity in behaving mice.

This section focuses on the application of LSFM in zebrafish neuroscience. It aims to illustrate how the unique performance of this new imaging method may shed a new light on brain-scale neuronal processing in the vertebrate brain.

In the last two decades, the zebrafish (Danio rerio) has emerged as an important model in circuit neuroscience. Native to Asia, this small vertebrate displays important physiological and genetic homologies with mammals. Its genome is fully sequenced and a large panoply of genetic tools is now available. The zebrafish was originally used in developmental biology because its main organs develop within the first day post-fertilization and this process can be easily monitored. At this early stage, the larva is indeed transluscent—only a few pigments are present on the skin—which permits exceptional optical access to the entire specimen. The identification of a mutation controlling pigment cell formation³⁸  led to the design of the so-called nacre zebrafish line that lacks skin pigments, thus improving the advantages of zebrafish larvae for in vivo imaging.

With the development of genetically encoded calcium indicators, this asset was further exploited to perform in vivo functional imaging.^39–41  The zebrafish larval brain being quasi-transparent, small and compact [the 10⁵ neurons of a 6 dpf (days post-fertilization) larva are contained within a volume of 2 × 0.5 × 0.3 mm], the entire brain can be optically monitored in a minimally invasive way.^39–41  To increase the number of neurons that could be simultaneously recorded, several strategies were developed, including high-speed random-access imaging using acousto-optical deflectors^42–44  and simultaneous multi-point excitation.⁴⁵  These methods still proved to be too slow to record the entire brain volume of zebrafish larvae with sufficient dynamics. A whole-brain map of neural activity therefore required the sequential recording of the various brain areas from different individuals and then the patching together of these different regions.^14,46  Such a sequential approach was tedious and only yielded mean activity patterns: it did not allow for the probing of large-scale concerted neural activities spanning distant brain regions. This imaging limitation precluded the study of complex neural processing that involved extended neural circuits.

In 2013, the first LSFM-based functional recordings were reported on zebrafish larvae expressing the genetically encoded calcium reporter GCaMP pan-neurally, generating activity time traces of almost 80 000 neurons at 0.7 Hz or 25 000 neurons at 4 Hz.^47,48  This new technique offered the first whole-brain recording at cellular resolution in a vertebrate, using activity correlation analysis. With these new sets of data, rich structures in space and time were now accessible, which could be revealed in a straightforward way using correlation analysis (Figure 1.10). It also paved the way to the study of brain-scale processing of complex sensorimotor tasks, as illustrated in the following section.

Although it possesses ∼10 times fewer neurons that a human retina, a 6-day-old zebrafish larva exhibits a rich behavioral repertoire, including goal-directed navigation (spontaneous swimming along illumination or thermal gradients), optomotor response (swimming in response to a global visual flux in order to maintain a constant position) and hunting (paramecia capture). These behaviors involve the integration and processing of various sensory stimuli to trigger adequate motor responses.

LSFM allows the monitoring of the entire brain, from visual to integrative and motor centers, and thus offers a unique window into the neural substrates of such complex sensorimotor behaviors. Although the animal needs to be tethered in agarose in order to perform brain imaging, it is possible to record motor outputs either by video monitoring tail or eye motion (after partially removing the agarose around the tail or eyes) or by extracellular recording of fictive swim bouts in paralyzed preparations. These recordings can then be used to produce, using a real-time feedback loop, a sensory environment with which the animal can virtually interact.⁴⁹ 

These approaches have recently been used, in two independent studies, to reveal specialized hindbrain circuits that control spontaneous and phototactic (towards a light source) navigation. Dunn et al.⁵⁰  used whole-brain LSFM to identify two bilaterally distributed neural assemblies whose activity drives the (left/right) orientation of successive swim bouts, and thus orchestrates the spatiotemporal pattern of spatial exploration (see Figure 1.10). Wolf et al.⁵¹  used two-photon LSFM to record brain activity when the zebrafish was exposed to visual stimulation. They were able to show that the same circuit is under partial control of visual stimuli, in such a way that the animal's trajectory is biased towards a light source.

One of the strong assets of LSFM is to provide a straightforward way to map the entire circuit involved in a particular task. This can be achieved by simply correlating the activity of the entire brain with the motor or sensory signals. Such functional mapping approaches do not rely on pre-existing hypotheses regarding the involved brain regions, as the entire brain can be analyzed at once.

Light-sheet fluorescent microscopy is a powerful technique that allows fast volumetric imaging at high spatial resolution with minimal phototoxicity. Since its “re-invention” about 10 years ago, LSFM has rapidly spread through various fields of biology, reflecting its extreme versatility. Various imaging methods, such as confocal filtering, two-photon excitation, structured illumination and super-resolution, have been combined with LSFM to increase its performance and expand the scope of its applications.

This evolution is likely to continue in the coming years. Hence adaptive optics technology, which is becoming standard in epifluorescence microscopy,^52–54  will likely be used to improve the quality of the light sheet by compensating for tissue-induced optical aberration and scattering. Similarly, dynamic control of the light intensity during recordings, using computational methods developed for confocal microscopes,⁵⁵  might further limit photobleaching in LSFM. Efforts to miniaturize LSFM,^56,58  using optical fibers,⁵⁷  may eventually allow the design of light-sheet endoscopes. Finally, the impact of LSFM may be reinforced when combined with other techniques, such as optogenetics or laser ablation, in order to probe the response of a biological system to a controlled perturbation.

Beyond the technical issues associated with the optical design, LSFM also raises important computational and theoretical challenges: how to process, analyze and navigate through the massive datasets produced by such a technique. This problem already constitutes a practical bottleneck in many applications, such as whole-brain calcium recording or high-resolution imaging of intact clarified brains. Quantifying, standardizing and automating these processing methods will likely become crucial requirements in the near future.