---
title: "Single-Shot 3D SIM: How a Spinning Disk Reconstructs an Image"
date: 2026-09-06
updated: 2026-09-06
category: optics
excerpt: "How spinning-disk 3D structured illumination microscopy produces one super-resolved optical section per exposure, with the principle, measured results, and practical limits explained."
author: "Youchang Zhang"
tags: [single-shot 3D SIM, structured illumination microscopy, spinning disk, optical reconstruction, super-resolution]
---

**Single-shot 3D structured illumination microscopy (3D SIM)** can produce a super-resolved image of one optical section during a single camera exposure. In the spinning-disk method described here, a rotating patterned disk creates the illumination and optically demodulates the returning fluorescence. A complete 3D volume still requires a stack of sections at different depths.

**By Youchang Zhang (Joshua Zhang)** · A plain-language companion to our 2023 paper, [*3D structured illumination microscope using a spinning disk [Invited]*](https://doi.org/10.1364/BOE.499181), published in *Biomedical Optics Express*. This tutorial explains the published method; it is not a separate peer-reviewed study.

[Read the open-access paper](https://pmc.ncbi.nlm.nih.gov/articles/PMC10659796/) · [Explore the MATLAB code](https://github.com/youchang-z/SD-3DSIM) · [Download this article as Markdown](/articles/single-shot-3d-sim.md)

## The idea in one minute

A conventional microscope blurs very small fluorescent structures. Structured illumination microscopy helps by lighting the sample with a known pattern: fine details that would normally be invisible become encoded in a larger, detectable pattern.

Conventional 3D SIM records a sequence of images with different illumination phases and orientations, then combines them on a computer. Our spinning-disk implementation uses the same physical disk twice: first to pattern the excitation light, then to filter the returning fluorescence. As the disk rotates, the camera integrates an optically reconstructed image.

The result is one exposure per optical section, with resolution improvement in both the lateral plane and the depth direction. The paper reports bead-image widths of about **129 nm laterally and 388 nm axially**, following Wiener filtering, and an acquisition capability of **up to 100 section images per second**. These describe the reported instrument, rather than a guarantee for every sample or microscope. [1]

## Why does conventional 3D SIM usually need 15 images?

Imagine photographing two fine meshes laid over one another. The meshes can create broad moiré bands even when their individual threads are difficult to resolve. SIM uses a related mixing effect: multiplying a fluorescent sample by a known illumination pattern shifts some of its fine-detail information into the microscope's detectable range.

One patterned image contains several overlapping contributions. A computer must separate them before putting the shifted information back in the correct place. Moving the pattern through several phases changes these contributions in a predictable way. Rotating it supplies information along additional lateral directions.

The classic three-beam 3D SIM acquisition uses **five phases at each of three orientations: 5 × 3 = 15 raw images at each axial position**. Reconstruction uses the axial stack as well as those phase and orientation measurements. The 15-image description refers to this conventional scheme; it is not a universal minimum for every technique called SIM. [2]

This explains why sample motion can be troublesome. The reconstruction expects the sequence to describe essentially the same structure. If a fluorescent feature moves while the sequence is being acquired, the measurements may no longer fit that assumption.

![Figure 1. Conventional three-beam 3D SIM records five phases at three orientations per axial position. The spinning-disk method integrates a rotating optical scan into one camera exposure per section. Both methods still sample depth to build a volume; the exposure bars are schematic and do not imply equal exposure times.](/images/single-shot-3d-sim/acquisition.png)

## How can a spinning disk do the reconstruction?

The disk is an **amplitude grating**: a thin opaque aluminum layer with a periodic array of transparent pinholes on a glass substrate. “Amplitude” means it controls how much light passes through each position.

The optical sequence has four steps:

1. **Pattern the excitation.** Laser light passes through the disk. The optical system admits one central diffraction order and four first orders, which interfere in the sample.
2. **Collect the fluorescence.** The objective collects the sample's emission and images it back onto the same disk.
3. **Modulate it again.** The pinhole pattern acts on the returning image before it reaches the camera. This second modulation supplies the matching spatial frequencies needed for optical demodulation.
4. **Average during rotation.** As the disk scans, matched contributions survive the exposure average while unmatched terms cancel under the uniform-scan approximation. The camera records the integrated result. [1, Section 2]

One way to picture this is a moving stencil that participates in both writing and reading a pattern. Because illumination and detection use the same disk, their spatial modulation is linked physically. The detailed mechanism is frequency mixing and averaging, rather than simply photographing a sharper projected pattern.

The disk has two distinct jobs: **optical demodulation recovers fine-detail information; physical masking reduces out-of-focus background**. Removing background alone would not explain the complete resolution mechanism.

### A little mathematics, with the terms explained

In the paper's idealized model, the effective optical transfer function can be written as a weighted sum of shifted copies of the ordinary microscope's transfer function:

$$
H_{\mathrm{SIM}}(\mathbf{k}) = \sum_p a_p^2 H(\mathbf{k}-\mathbf{k}_p).
$$

Here, $H$ describes which spatial frequencies the microscope transmits. Large spatial frequencies correspond to fine details. Each $\mathbf{k}_p$ is a frequency present in the illumination pattern, and $a_p^2$ gives its weight in this model. The shifted copies extend the range of detail the microscope can convey. [1, Eq. 9]

This expression assumes matched illumination and detection modulation and spatially uniform scanning. The actual rotating pattern changes orientation during the exposure; the paper therefore calculates the effective rotating-system transfer function numerically. You do not need the derivation to use the key idea: the optics carry out the frequency shifting and combination that a conventional SIM reconstruction performs digitally.

## What makes the illumination three-dimensional?

A flat drawing of the pinhole array can hide the most useful part of the design. The illumination varies in depth as well as across the focal plane.

The central beam travels along the optical axis. The four first-order beams enter at an angle. Their axial wave-vector components differ from that of the central beam, so their relative phases change as you move through depth. Interference therefore produces a **3D lattice of illumination intensity**. [1, Fig. 3]

![Figure 2. Five admitted excitation beams generate illumination that varies laterally and axially. These normalized scalar-model cross-sections illustrate the principle; they are not experimental images or a reproduction of the paper's vectorial simulation. The side view follows y = x, as a diagonal through the lateral lattice.](/images/single-shot-3d-sim/lattice.png)

The paper's vectorial simulation gives a lateral lattice period of approximately **500 nm** and an axial period of **1,072 nm**, with about **88% illumination contrast**. The resulting transfer function extends laterally and axially and fills the wide-field “missing cone”: a region of poorly transferred spatial information associated with weak depth discrimination. [1, Figs. 3–4]

There are two meanings to keep separate. **Axial resolution** describes how well structures at nearby depths can be distinguished. **Volume acquisition** means collecting information across a range of depths. Better axial resolution does not, by itself, make an entire volume appear in a single frame.

## What does single-shot actually mean here?

The camera records one section image while the disk moves through a scan. The experiment describes one reconstructed image per complete disk rotation. The illumination is therefore changing during that one exposure: “single-shot” describes the camera acquisition, not a frozen illumination pattern. [1, Section 3.1]

For a volume, the focus or sample position must still move through a series of axial positions. A 50-plane volume needs 50 section exposures in this method, before accounting for repeats, additional colours, or acquisition overhead.

For example, **if** every section could be acquired at 100 frames/s, 50 sections would take at least 0.5 seconds, giving an ideal ceiling of 2 volumes/s. This is simple acquisition arithmetic, not a measured volumetric performance result. Axial motion, settling, readout, and the photons needed for each image can make the real volume rate lower.

The method avoids the need to assemble a conventional sequence of phase-shifted raw frames for each section. It does not remove motion blur within the exposure, drift during the stack, or the need to synchronize acquisition with a sufficiently uniform disk scan.

## What did the experiment demonstrate?

The published instrument used 488 nm excitation, a 100× objective with numerical aperture 1.35, and an overall magnification of 167×. The reported field of view was approximately 80 × 80 µm. Resolution was assessed using nominally 100 nm fluorescent beads, and biological imaging used fixed, labelled rat cardiac myocytes. [1]

| Quantity | Wide-field comparison | Spinning-disk 3D SIM | How to interpret it |
|---|---|---|---|
| Lateral bead FWHM | 261 ± 11 nm | 129 ± 3 nm | Approximately twofold narrowing |
| Axial bead FWHM | 771 ± 47 nm | 388 ± 13 nm | Approximately twofold narrowing |
| Acquisition capability | No matched speed benchmark given here | Up to 100 section frames/s | Camera-limited capability of the reported setup |
| Out-of-focus rejection | Reference imaging mode | Approximately 90% | Geometric explanation based on the opaque disk area |
| Biological example | Fixed rat cardiac myocyte | RyR2-labelled cell, approximately 12 µm thick | Figure 6 includes a plane 6 µm below the surface |

FWHM means **full width at half maximum**: the width of a bead's image at half its peak brightness. A narrower bead image indicates less blur. These are reported Gaussian-fit image widths, not bead diameters or a universal minimum separation for two biological structures. The ± values are reproduced as reported in the paper; no additional statistical interpretation is assigned here.

![Figure 3. Reported fluorescent-bead FWHM values, replotted from Zhang et al., Biomedical Optics Express 14, 5710–5719 (2023), Section 4.3. Error bars reproduce the published plus/minus values. The paper compares Wiener-filtered spinning-disk SIM images with raw wide-field images; these bars do not isolate the optical contribution from subsequent processing.](/images/single-shot-3d-sim/resolution.png)

The cell experiment demonstrates that the method can reveal labelled structures within a thick fluorescent cell with improved contrast. Because these myocytes were fixed, that experiment should not be read as a demonstration of live-cell dynamics at 100 frames/s.

### Why a 33% opening ratio can reject about 90% of background

The paper defines the opening ratio as **pinhole diameter divided by array period**: 28 µm / 84 µm = 1/3. This is a ratio of lengths, not areas.

For circular pinholes on an ideal square lattice, the transparent area fraction is approximately

$$
f_{\mathrm{open}} = \frac{\pi}{4}\left(\frac{28}{84}\right)^2 \approx 0.087.
$$

That leaves about 91% opaque area, consistent with the paper's rounded description of roughly 90% background rejection. This geometric argument is useful for spatially spread out-of-focus light. It is not a promise that every sample loses exactly 90% of its background, nor that 90% of the desired in-focus fluorescence reaches the camera.

Smaller holes can suppress more background but also cost useful signal. The chosen opening ratio balances background rejection and signal-to-noise ratio. The disk also uses eight sectors with radial offsets to improve scanning uniformity; the paper reports a simulated increase from 72.7% to 94.4%. [1, Section 3.2]

## Does optical reconstruction mean no image processing?

No. The disk optically reconstructs the section image, but the paper also applies **Wiener filtering** to enhance weak high-frequency information and contrast. Its bead figure explicitly compares Wiener-filtered SIM images with raw wide-field images. [1, Section 4.2 and Fig. 5]

A Wiener filter balances sharpening against noise amplification using a model of the image transfer. This is a different operation from separating and recombining the phase and orientation images in conventional SIM. It remains digital processing, and its parameters influence the displayed result.

For quantitative work, keep the acquired images, document filtering parameters, and evaluate resolution and artifacts under the actual imaging conditions. A crisp-looking image alone is not sufficient evidence of accurately recovered structure.

## How does it relate to other fast microscopy methods?

Several approaches use rotating disks or optical processing, so the names can be confusing.

| Method | Central idea | Relationship to this method |
|---|---|---|
| Conventional three-beam 3D SIM | Record phase/orientation sequences and reconstruct computationally | Shares structured illumination and 3D bandwidth extension; uses a different acquisition and reconstruction workflow |
| Spinning-disk confocal microscopy | Scan pinholes to illuminate and detect many locations while rejecting background | A shared hardware concept; a rotating confocal disk alone does not establish this 3D SIM mechanism |
| Instant SIM (iSIM) | Perform operations including local image contraction and reassignment optically | Another established route to rapid super-resolution through analog image processing; not a synonym for this disk design [3] |
| Phase-modulated spinning-disk SIM | Use a phase-modulation layer with a pinhole disk to shape illumination | Our related 2023 optical-sectioning method; its disk design and resolution claims should not be substituted for the amplitude-disk 3D method [4] |

Optical reconstruction has earlier precedents. Hayashi and Okada described spinning-disk super-resolution imaging in 2015; our 3D paper builds on that optical-demodulation idea and develops the three-dimensional lattice and transfer-function extension. The contribution should be understood in that context, rather than as a claim that all single-exposure optical super-resolution began with this paper. [1, 5]

## When might this approach be useful?

It is worth investigating when multi-frame acquisition is a bottleneck, when out-of-focus fluorescence hides features inside a cell, and when an optical section with improved lateral and axial resolution is more useful than a conventional wide-field frame.

Practical implementation still needs attention to the following:

- **Photon budget.** Fewer camera frames do not automatically mean fewer excitation photons, less bleaching, or higher signal-to-noise ratio. Disk transmission and sample brightness matter.
- **Disk scan and camera timing.** An incomplete or uneven scan can leave spatial nonuniformity. A shorter exposure is not automatically a valid reconstruction.
- **Alignment and aberrations.** The same-disk arrangement links illumination and detection, but refractive-index mismatch and imperfect optics can still affect performance.
- **Axial sampling.** A full 3D experiment requires suitable z steps and enough sections. Section rate and volume rate answer different questions.
- **Conditions beyond the demonstration.** The paper's fixed-cell, single-colour implementation does not establish performance for every live sample, label, wavelength, or tissue depth.

These are reasons to match the instrument to the measurement. The acquisition advantage is most useful when enough photons can be collected during a properly completed scan and the sample remains sufficiently stable over that exposure.

## Frequently asked questions

### Can single-shot 3D SIM capture a whole volume in one exposure?

In this spinning-disk implementation, no. One camera exposure produces one optically reconstructed section. A volume requires sections collected at different axial positions.

### Does it need deep learning?

The published method does not require a learned reconstruction model. It uses physical illumination and detection modulation, followed by Wiener filtering in the reported processing workflow. The MATLAB code includes simulations and filtering. [1]

### Is it exactly 15 times faster than conventional 3D SIM?

Reducing the raw-frame count from 15 to one per axial position does not establish a 15-fold speed increase under matched image quality. Exposure time, photon efficiency, camera readout, pattern switching, and axial scanning all affect the comparison.

### Is this the same as single-shot 3D surface measurement?

No. Here, SIM means **structured illumination microscopy of fluorescent specimens**. Structured-light profilometry and other snapshot 3D methods can measure surfaces or volumes using different physics and detection arrangements.

### Where can I find the code and cite the method?

The authors' [SD-3DSIM repository](https://github.com/youchang-z/SD-3DSIM) contains MATLAB simulation and image-processing code. Consult its current documentation for inputs and requirements. For scientific use of the method, cite the original journal paper below rather than treating this tutorial as a new experimental result.

## Paper, code, and references

**Original paper:** Youchang Zhang, Parisa Asghari, David R. L. Scriven, Edwin D. W. Moore, and Keng C. Chou. “3D structured illumination microscope using a spinning disk [Invited].” *Biomedical Optics Express* **14**(11), 5710–5719 (2023). Published 11 October 2023. DOI: [10.1364/BOE.499181](https://doi.org/10.1364/BOE.499181).

[Open-access full text and supplementary links](https://pmc.ncbi.nlm.nih.gov/articles/PMC10659796/) · [MATLAB code](https://github.com/youchang-z/SD-3DSIM) · [Download BibTeX](/articles/single-shot-3d-sim.bib) · [Author's Google Scholar profile](https://scholar.google.com/citations?user=oL_MjNoAAAAJ&hl=en)

```bibtex
@article{Zhang2023SpinningDisk3DSIM,
  author = {Zhang, Youchang and Asghari, Parisa and Scriven, David R. L.
            and Moore, Edwin D. W. and Chou, Keng C.},
  title = {{3D} structured illumination microscope using a spinning disk [Invited]},
  journal = {Biomedical Optics Express},
  year = {2023},
  volume = {14},
  number = {11},
  pages = {5710--5719},
  doi = {10.1364/BOE.499181},
  url = {https://doi.org/10.1364/BOE.499181}
}
```

1. **Zhang et al. (2023).** [3D structured illumination microscope using a spinning disk [Invited]](https://doi.org/10.1364/BOE.499181). *Biomedical Optics Express* 14, 5710–5719. Primary source for the method and all instrument-specific results in this tutorial.
2. **Gustafsson et al. (2008).** [Three-dimensional resolution doubling in wide-field fluorescence microscopy by structured illumination](https://doi.org/10.1529/biophysj.107.120345). *Biophysical Journal* 94, 4957–4970. Conventional three-beam 3D SIM and its phase/orientation acquisition.
3. **York et al. (2013).** [Instant super-resolution imaging in live cells and embryos via analog image processing](https://doi.org/10.1038/nmeth.2687). *Nature Methods* 10, 1122–1126. A related optical-processing approach.
4. **Zhang et al. (2023).** [Structured illumination microscopy with a phase-modulated spinning disk for optical sectioning](https://doi.org/10.1364/OL.494655). *Optics Letters* 48, 3933–3936. Related phase-modulated disk design.
5. **Hayashi and Okada (2015).** [Ultrafast superresolution fluorescence imaging with spinning disk confocal microscope optics](https://doi.org/10.1091/mbc.E14-08-1287). *Molecular Biology of the Cell* 26, 1743–1751. Earlier spinning-disk optical demodulation.

The three illustrations in this tutorial were created for explanation. Figure 2 is an idealized scalar calculation; Figure 3 replots the numerical values reported in the original paper. No generated image is presented as a new experimental measurement.