What light sheet microscopy is, in one paragraph
Light sheet microscopy, also called light sheet fluorescence microscopy (LSFM) or selective plane illumination microscopy (SPIM), illuminates a fluorescent sample with a thin sheet of laser light from the side and records the emitted fluorescence with a camera placed at a right angle to that sheet. Only the plane that is being imaged receives light, so there is no out-of-focus fluorescence to remove, the whole plane is captured in a single camera exposure, and the rest of the sample is left in the dark until its turn comes.
Moving the sample (or the sheet) step by step through the illumination plane produces a stack of optical sections and, from them, a 3D volume. Because the sheet is typically a few hundred nanometres to a few micrometres thick, the sectioning quality is comparable to confocal microscopy, but the acquisition is orders of magnitude faster and far gentler on living specimens. That combination is why light sheet systems have become the standard tool for imaging developing embryos, whole zebrafish larvae, organoids and cleared mouse brains.
Light sheet microscopy principle: decoupling illumination from detection
Every other fluorescence method sends the excitation light through the same objective that collects the emission. A widefield microscope floods the whole sample, and a confocal microscope focuses a point and rejects out-of-focus light afterwards with a pinhole. Light sheet microscopy takes a different route: the illumination and detection paths are physically separate and orthogonal.
How the sheet is generated
- Static sheet with a cylindrical lens. A cylindrical lens focuses a collimated laser beam in one axis only, producing a thin, wide sheet. Simple, cheap and fast, but the sheet thickness varies across the field of view and the Gaussian intensity profile is not perfectly uniform.
- Scanned sheet (digital scanned laser light sheet, DSLM). A galvanometer mirror sweeps a focused Gaussian beam up and down thousands of times per second; the time average forms a virtual sheet. The intensity is more uniform and the laser power is used more efficiently.
- Bessel, Airy and lattice beams. Shaped beams keep a thin core over a longer distance than a Gaussian beam, which extends the usable field of view. Bessel beams carry side lobes that add background; lattice light sheet microscopy (introduced in 2014) interferes many Bessel beams so that the side lobes cancel, giving a thin sheet suitable for live sub-cellular imaging.
The thickness versus field-of-view trade-off
A Gaussian beam stays thin only over a limited length (its Rayleigh range). A very thin sheet therefore covers only a small field of view before it spreads out; a sheet that stays thin across a 1 mm field of view is necessarily several micrometres thick. Every light sheet design is a compromise between axial resolution and field size, and remedies such as tiling several sheets, axially sweeping the beam waist across the field during the exposure, or using shaped beams exist precisely to soften that compromise.
Detection and resolution
The detection arm is a normal widefield fluorescence path: an objective, an emission filter and a fast scientific CMOS camera. Lateral resolution is set by the detection objective and is in the range of roughly 250 to 500 nm for green emission with high-numerical-aperture lenses. Axial resolution is determined by whichever is thinner, the light sheet or the detection depth of field, and is typically several times worse than lateral resolution unless the sheet is very thin or the sample is imaged from several angles and the views are fused.
Light sheet microscopy vs confocal: where each one wins
The question "light sheet or confocal" has a clear answer once the sample is defined. Confocal excels at small, thin, fixed samples where sub-micrometre detail in a single plane matters. Light sheet excels at large or living samples where speed, low photodamage and volume coverage matter more than the last fraction of resolution.
| Criterion | Point-scanning confocal | Light sheet (SPIM/LSFM) |
|---|---|---|
| Illumination | Whole cone of light through the sample for every scanned point | Only the imaged plane is illuminated |
| Image formation | One pixel at a time, point detector (PMT) | Whole plane at once, camera |
| Speed | Seconds per optical section at full resolution | Tens to hundreds of planes per second; 100 to 1,000 times faster than point scanning |
| Photobleaching and phototoxicity | High: every plane is exposed while any plane is imaged | Low: each plane is exposed only once per volume |
| Lateral resolution | Diffraction limited, slightly better than widefield | Diffraction limited, set by detection objective |
| Axial resolution | Set by pinhole and objective NA, about 500 to 800 nm with high NA | Set by sheet thickness; a few hundred nm (lattice) to several µm (large fields) |
| Sample size | Cells, thin tissue sections, small embryos | Embryos, larvae, organoids, cleared organs up to whole mouse brains |
| Live imaging over days | Limited by photodamage | Routine: time-lapse volumes every few seconds to minutes for days |
| Sample mounting | Standard slides and dishes | Agarose cylinders, holders, or clearing-medium chambers; needs optical access from two sides |
| Data per experiment | Megabytes to gigabytes | Gigabytes to terabytes |
Two points in the table are decisive in practice. First, photon budget: in a confocal stack of 100 planes, each plane is illuminated 100 times (once for every plane in the stack) but contributes signal only once. In a light sheet stack, each plane is illuminated once. For a live embryo imaged every two minutes for 24 hours, that difference decides whether the embryo survives the experiment. Second, speed: a camera records a megapixel plane in a few milliseconds, which is what makes whole-brain calcium imaging in larval zebrafish possible at several volumes per second.
If you want to revisit how the confocal pinhole creates optical sections before comparing the two, see our guide to confocal microscopy and the pinhole.
Light sheet microscopy of the brain and other large samples: tissue clearing
Light cannot travel through a millimetre of opaque tissue, so light sheet imaging of whole organs depends on tissue clearing. Clearing removes lipids and matches the refractive index of the remaining tissue to the surrounding medium, so that scattering drops and the sample becomes transparent. The field has several families of protocols.
| Clearing family | Examples | Principle | Typical use with light sheet |
|---|---|---|---|
| Solvent based | 3DISCO, iDISCO, uDISCO | Dehydration in organic solvents, lipid extraction, index matching in a high-index solvent; tissue shrinks | Whole mouse brain and other organs with immunolabelling; fast; quenches some fluorescent proteins |
| Aqueous, hyperhydrating | CUBIC, Scale | Urea or amino-alcohol solutions remove lipids and hydrate tissue | Preserves fluorescent proteins well; slower; tissue may expand |
| Hydrogel embedding | CLARITY and derivatives | Tissue is locked in a hydrogel, lipids removed by detergent or electrophoresis | Compatible with repeated antibody staining of the same brain |
Cleared samples impose their own optical demands. Immersion media have refractive indices between about 1.45 and 1.56, so the detection objective must be designed or corrected for that medium, usually with a long working distance of many millimetres to reach the centre of an intact organ. Dedicated mesoscale light sheet systems, including the open-source mesoSPIM platform, were built for exactly this: centimetre-sized cleared samples imaged at a few micrometres resolution within tens of minutes to a few hours per channel.
Typical brain applications include mapping the distribution of labelled neurons across an entire mouse brain, tracing long axonal projections, counting activated cells after a behavioural task, and visualising vasculature or amyloid plaques in disease models. The output is a complete 3D data set that can be registered to a reference brain atlas and quantified region by region, which sectioning-based histology can only approximate.
Live imaging: embryos, zebrafish and organoids
The historical motivation for SPIM, published in 2004, was to watch embryos develop without harming them, and that remains the core application. Because each plane is illuminated only once per volume, a zebrafish, fruit fly or mouse embryo can be imaged every minute or two for a day or longer, and the resulting four-dimensional data allow every cell to be tracked from the first divisions onwards.
- Zebrafish. The transparent larva fits a light sheet almost perfectly. Whole-brain neuronal activity can be recorded with genetically encoded calcium indicators at volume rates of a few hertz; the beating heart has been imaged in real time.
- Developmental biology. Multi-day time-lapse recordings of embryogenesis with automated cell tracking and lineage reconstruction are a standard light sheet experiment.
- Organoids and spheroids. Three-dimensional cultures several hundred micrometres across are too thick for confocal time-lapse at useful speed but ideal for light sheet, often in simple inverted or open-top geometries.
- Plants. Roots growing in gel can be followed for days with minimal light stress.
Sample mounting
Because the illumination enters from the side, the specimen cannot lie on a slide. Live organisms are usually embedded in a cylinder of low-concentration agarose and hung in a water-filled chamber in front of the detection objective; cells may be grown on a glass strip or cultured in index-matched foil bags. Inverted and open-top designs, where both objectives point at the sample from below at 45 degrees, allow conventional dishes and multiwell plates and have made light sheet accessible for routine cell biology.
Artefacts and limitations you should know before buying time on a system
Light sheet microscopy is not a universal upgrade. Several artefacts are specific to the geometry and must be managed in the design of the experiment.
- Stripes and shadows. Anything absorbing or scattering in the illumination path, such as pigment cells or yolk, casts a shadow along the sheet, producing dark stripes across the image. Remedies include illuminating from two opposite sides, pivoting the sheet by about 10 degrees during exposure, and post-processing stripe filters.
- Degradation deep in the sample. Even in a cleared sample the sheet broadens and the detection path scatters with depth. Multi-view imaging (rotating the sample and fusing several angles) recovers isotropic resolution at the cost of acquisition time and data volume.
- Field size versus sheet thickness. As discussed above, a thin sheet means a small field. Decide which matters more for the question at hand.
- Refractive index mismatch. Mounting medium, clearing medium and objective correction must match; otherwise the sheet and the focal plane drift apart with depth and images become blurry on one side.
- Sample size limits. The two objectives must have physical access to the sample at right angles, which constrains chamber design and sample thickness.
- Not a super-resolution method on its own. Standard light sheet resolution is diffraction limited. Lattice light sheet improves axial resolution and gentleness, and can be combined with structured illumination or single-molecule methods, but a plain SPIM does not resolve below about 250 nm.
The data problem: handling terabytes of images
Speed produces data. A single camera at full frame rate can write hundreds of megabytes per second, and a multi-day, multi-view time-lapse of an embryo or a multi-channel whole-brain scan routinely reaches terabytes. Many first-time users discover that the microscope was the easy part.
Plan the analysis before the acquisition. Decide what will be quantified, downsample or crop at acquisition time where possible, and make sure the storage, transfer and processing pipeline can keep pace with the camera. A perfect data set that nobody can open is not a result.
- Storage and formats. Chunked, multi-resolution formats (such as HDF5-based or Zarr-based containers used by the main open-source viewers) allow a terabyte volume to be browsed on a laptop without loading it completely.
- Registration and fusion. Multi-view data must be registered, often using fluorescent beads in the agarose as landmarks, before views are fused and deconvolved.
- Segmentation and tracking. Cell nuclei detection, tracking over time and atlas registration are typically done with open-source toolkits; deep learning methods now carry much of this load.
- Compute. GPU workstations or cluster access are a practical necessity for whole-brain or long time-lapse data sets; budget for them alongside the instrument.
A short history and the current landscape
The idea is older than fluorescence microscopy itself. The ultramicroscope of the early twentieth century used a sheet of light and side-on observation to see colloidal particles too small to resolve, work for which Zsigmondy received the Nobel Prize in Chemistry in 1925. In 1993 the geometry was first combined with fluorescence to image the cochlea in a method called orthogonal-plane fluorescence optical sectioning. The modern era began in 2004 with SPIM applied to live embryos, followed by scanned sheets, multi-view fusion, two-sided illumination, Bessel and lattice beams, open-source designs such as OpenSPIM (2012) and mesoSPIM, and the first commercial systems around 2010 to 2012.
Today the choice ranges from compact commercial instruments for cleared tissue and organoids, through lattice systems for sub-cellular live imaging, to home-built or open-hardware set-ups adapted to a single question. Which one is right depends on three questions: how big is the sample, is it alive, and how much data can you handle.
Fluorophore choice, filter sets and photobleaching behave in a light sheet much as they do elsewhere; our overview of fluorescence microscopy, filters and common pitfalls covers the basics that apply before any light sheet experiment.
Conclusion
Light sheet microscopy replaces the confocal strategy of "illuminate everything, reject most of it" with "illuminate only what you record". The result is optical sectioning with a fraction of the light dose, acquisition speeds that make whole-brain activity and multi-day development observable, and, together with tissue clearing, complete 3D images of intact organs. The price is a more demanding sample mounting, geometry-specific artefacts, and data volumes that require real infrastructure. For large or living samples, that trade is almost always worth making.