Open Access
Add to BibSonomy
Abstract
To meet the increasing need for low-cost, compact imaging technology with cellular resolution, we have developed a microLED-based structured light sheet microscope for three-dimensional ex vivo and in vivo imaging of biological tissue in multiple modalities. All the illumination structure is generated directly at the microLED panel—which serves as the source—so light sheet scanning and modulation is completely digital, yielding a system that is simpler and less prone to error than previously reported methods. Volumetric images with optical sectioning are thus achieved in an inexpensive, compact form factor without any moving parts. We demonstrate the unique properties and general applicability of our technique by ex vivo imaging of porcine and murine tissue from the gastrointestinal tract, kidney, and brain.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Early detection and diagnosis of epithelial cancers is associated with improved patient outcomes and reduced mortality [1]. Computed tomography, ultrasound, and magnetic resonance imaging can provide in vivo information about tumor size and location, but there is a disparity in the availability of these imaging systems due to their complexity and cost. Optical imaging techniques provide a promising alternative for disease detection with enhanced portability and lower cost [2]. Abnormalities in nuclear morphology and intercellular architecture are common indications of cancer and therefore require imaging systems to have micron-scale resolution over a modest field of view (FOV) to be feasible for disease detection [3,4]. To comprehensively monitor cell groups in vivo, volumetric imaging is further required. Several optical imaging and spectroscopic techniques have been developed to expand examinations, including reflectance confocal microscopy [5,6], optical coherence tomography [7–9], two-photon microscopy [10,11], Raman spectroscopy [12,13], diffuse reflectance spectroscopy [14], and multispectral imaging [15,16]. These approaches typically rely on point scanning, expensive components, or bulky configurations, however, and therefore have limited reach in underserved communities. Attempts have been made to mitigate these issues by replacing mechanical scanning hardware with optical components. For example, digital micromirror devices and hyperchromatic objective lenses can be used to perform lateral and axial scanning, respectively, but these systems still commonly rely on upstream illumination optics and external light sources [17].
Light sheet microscopy (LSM) is an imaging technique wherein the sample is illuminated from the side by a thin plane of light that is coplanar with the microscope’s focal plane. Because illumination is delivered only to the sample region that is actively being imaged, out-of-focus background signal is avoided (i.e. optical sectioning) and light-induced damage to biological samples is minimized [18]. Further, LSM has shown promise as a high-throughput imaging modality capable of discriminating between healthy and cancerous tissue [19,20]. Many implementations rely on complex illumination configurations to achieve these benefits, often employing galvanometric scanners, Bessel beams, and cylindrical lenses, which ultimately limits their potential for clinical translation [21,22]. A system configuration with a reduced number of components would help accelerate the adoption of LSM into clinical workflows and expand access to healthcare in low resource settings.
MicroLED panels have benefitted from strong demand within the augmented and virtual reality (AR/VR) display industry and are in a state of constant improvement but have yet to be explored extensively for use in biomedical applications [23–25]. We previously demonstrated the utility of microLED panels for generating patterned illumination in chromatic confocal microscopy [26]. Here we extend this concept by presenting a structured light sheet microscope with a compact illumination arm free from moving parts. Structured illumination microscopy (SIM) enhances out-of-focus background rejection and can help compensate for light sheet broadening in turbid tissue, allowing for better evaluation of cellular morphology [27,28]. A common approach to combine SIM and LSM builds upon digital scanned laser LSM by incorporating an acousto-optical tunable filter that electronically modulates the laser beam as it sweeps across the sample, yielding structured light sheets [29]. Others have augmented this with Bessel beam illumination to create thinner light sheets that allow for isotropic resolution in three dimensions but at the cost of complexity, thereby reducing the likelihood for translation of this useful imaging modality to clinical and low-resource settings [30]. Alternatively, a pyramid prism can generate a three-dimensional lattice of structured light sheets via four beam interference [31]. In this configuration, pattern frequency is modified by varying the zoom ratio of a telescope and pattern phase is shifted by tilting a plane parallel plate, so elaborate optomechanical coordination is required. In our technique, the intensity of individual microLED pixels can be controlled digitally, so no additional hardware is required to modulate light sheets at dynamically tunable frequencies, representing a significant simplification of these previously reported methods.
Display brightness has hindered progress for microLED-based biomedical imaging in the past, but our results show empirically that interpretable scattering, fluorescence, and autofluorescence images may be captured using ex vivo animal tissue samples. We further highlight the versatility of our microscope by imaging a mouse brain prepared for optogenetics experimentation ex vivo.
2. Volumetric structured light sheet illumination with no moving parts
MicroLED panels are two-dimensional arrays of sub-10µm light emitting diodes (LEDs) that can be independently addressed to generate custom patterns and have shown promise as an emerging technology in the field of optical displays. Thanks to strong demand from the AR/VR industry, their development is in a state of constant acceleration toward improvements that will potentially benefit numerous other types of optical instrumentation. Many biomedical imaging modalities rely on patterned illumination to excite specific tissue regions, enhance resolution, and reject out-of-focus background signal, which presents an opportunity to use these small displays. Figure 1(a) shows one of the microLED panels used in these experiments (JBD AMµLED 0.13”). The pattern displayed on the panel highlights our ability to control intensity, frequency, and orientation digitally. Note that pixelation in the detailed panel view is due to the camera that captured the image, not individual microLED pixels.
Fig. 1. Experimental layout. (A) Blue microLED panel mounted in 30 mm standard cage plate (Thorlabs CP35) for scale. Inset shows sinusoidal pattern displayed on microLED panel that demonstrates flexibility to modulate pattern intensity, frequency, and orientation. Panel dimensions: 2.56 mm x 1.92 mm (640 × 480 pixels). (B) Basic optical layout of microLED-based structured LSM. µLED, microLED panel; CMO, collection microscope objective; EX, excitation filter; PMO, projection microscope objective; IMO, imaging microscope objective; EM, emission filter; TL, tube lens; sCMOS, scientific CMOS sensor. (C) Detailed view of light sheet imaging at the sample with reference coordinate system for images. (D) and (E) show orthogonal views of light sheet generation by activating a single row of microLED pixels. (F) shows the same view orientation as (E) with two rows of microLED pixels activated to demonstrate digital light sheet scanning. (G) Example pattern displayed on microLED panel for structured LSM; scale bar, 500µm. (H) Example structured light sheet projected into sample; scale bar, 200µm.
A schematic of our experimental setup is shown in Fig. 1(b). Light emitted by a microLED panel (µLED) is collimated by a collection microscope objective (CMO, 4x/0.12 American Optics), passed through a spectral excitation filter (EX), and projected onto the sample through a projection microscope objective (PMO, 10x/0.30W Olympus). Fluorescence signal emitted from the sample is collected by an imaging microscope objective (IMO, 20x/0.50W Olympus) oriented perpendicularly to the illumination path, spectrally filtered by an emission filter (EM), and directed by a tube lens (TL, Thorlabs AC254-200-A) onto a scientific Complementary Metal–Oxide–Semiconductor (sCMOS) camera (PCO Edge 5.5). Both the PMO and IMO are designed to be immersed in water during imaging to minimize the refractive index mismatch at the tissue surface, thereby reducing stray light in the system caused by Fresnel reflections. Note that label-free, scattered-light images may be collected by simply removing the spectral filters, and that higher scatter contrast may be achieved by inserting crossed linear polarizers in certain circumstances. This configuration will hereafter be referred to as the “scattering” modality and experimental images will be labeled accordingly. Figure 1(c) shows a detailed view of the optical configuration at the sample along with a reference coordinate axis to clarify dataset orientation for our experiments. The illumination and imaging paths are placed at a 45-degree angle relative to the sample to make our system more clinically relevant, as this is the required orientation for imaging bulk tissue samples. Our instrument is controlled by a custom software application written in Python within the Micro-Manager environment [32]. The microLED panel is connected via HDMI and controlled as an additional monitor by displaying full screen images with matching resolution.
The microLED panel in our system has a pixel pitch of 4 microns, a resolution of 640 × 480 pixels, and is conjugate with the sample plane. The projection magnification in our system is ∼0.395x, yielding a microLED pixel size of ∼1.58µm at the sample. Note that the microLED pixel size at the sample effectively sets the minimum light sheet spacing (i.e., sampling rate) along the X-axis. Because the lateral size of one camera pixel is 0.293µm in the YZ-plane, volumetric sampling is therefore not isotropic. The microLED panel is conjugate with the sample, so any pattern shown on the panel will be axially elongated through the sample due to the depth of field of the PMO. Light sheet illumination may therefore be generated by activating a single row of microLED pixels. Figure 1(d)-(f) illustrates how conjugate planes in the illumination train allow for motionless translation of the light sheet along the X-axis [inset of Fig. 1(c)], which corresponds to varying the axial position along the imaging optical axis, thereby enabling volumetric light sheet imaging. Further, structured light sheet illumination can be achieved by modulating the row of microLED pixels according to any desired function (e.g. sine wave, square wave), as shown in Fig. 1(g)-(h). One additional benefit of using a microLED panel as the light source is that it is incoherent, with a typical spectral full width at half-maximum (FWHM) of ∼20-30 nm, so speckle is negligible. Currently, we select between panels with center wavelengths of 405 nm, 450 nm, 520 nm, and 635 nm depending on the imaging task.
For traditional optical sectioning SIM, three images must be captured with sinusoidal illumination that are modulated in phase by 2π/3 relative to one another [33]. An optically sectioned image, ${I_P}$, may then be calculated by
where ${I_1},{I_2},$ and ${I_3}$ are structured illumination images captured with relative phases of 0, $2\pi /3$, and $4\pi /3$, respectively. Because non-zero spatial frequencies are attenuated with defocus, objects outside of the focal plane will be illuminated without any modulation in all three phase-shifted images and therefore be removed from the image by the subtraction operations in Eq. (1).In our system, this limits the smallest microLED display period to 6 pixels (i.e., three on, three off). This corresponds to a spatial frequency of 105 mm-1, although typically we use lower frequency modulation to yield higher contrast fringes and therefore better reconstruction quality. Lower frequency modulation also extends the distance over which there is detectable pattern contrast well beyond the depth of field of the PMO, allowing for optical sectioning to be achieved over a larger FOV. The tradeoff for this is that the optical transfer function bandwidth is negligibly extended so no meaningful resolution enhancement is gained, but this is considered acceptable in our tissue imaging applications where optical sectioning over a large FOV is required to properly assess intercellular morphological features.
The 45-degree imaging configuration of our system yields skewed image volumes when viewed from the perspective of the tissue surface normal, but that does not impact the validity of captured data since this may be corrected by simple affine transformations. To clarify the orientation of our volumetric data, Fig. 2 shows maximum intensity projections of a piece of fluorescent lens tissue paper color-coded both by depth both along the X-axis [inset of Fig. 2(b)] and relative to the surface normal. These affine transformations rely on interpolation to re-code the data, however, so we will use depth coding along the X-axis, as shown in Fig. 2(a)-(d), for the remainder of this manuscript.
Fig. 2. Orientation of volumetric data. (A) 3D image of lens tissue paper stained with proflavine and excited by 450 nm microLED panel. (B) Geometric layout of tissue paper and IMO with depth measured along the X-axis. Maximum intensity projections of data color-coded by depth along the X-axis in the (C) YZ plane and (D) XZ plane. (E-F) The same projections shown with depth color coding relative to surface normal. Visualization 1 shows a fly-through along the X-axis, demonstrating the left-to-right image shift along the Z-axis implied in (B). All scale bars, 100µm.
Because our light sheet is formed by imaging a square pixel, the three-dimensional point spread function is different than those from light sheets formed by cylindrical lenses or laser beam scanning. Following the development in Ref. 18 for incoherent imaging with a circular pupil, and in reference to the coordinate system of Fig. 1(c), the lateral profile (along the X-axis) of the light sheet is given by
The system lateral scattering PSF is given by the elementwise multiplication of the illumination and detection PSFs along the Z-axis:
Plugging in values for our system (with green microLED illumination, λ = 0.520µm) yields theoretical full width at half-maximum (FWHM) PSF values of 0.535µm and 1.562µm along the Z- and X-axis, respectively (i.e. lateral and axial). To evaluate the as-built resolution of our microscope against this theoretical model, we imaged 500 nm (sub-resolution size) beads suspended in 2% agarose gel. The measured data, when fit to a pseudo-Voigt profile [35], shows FWHMs of 0.608µm and 1.962µm along the Z- and X-axis, respectively, as shown in Fig. 3. Note that the axial bead image has been scaled using bicubic interpolation to yield equal pixel dimensions in both directions for visualization. Figure 3(e)-(f) shows a comparison of three different PSF fits: Gaussian, Lorentzian, and pseudo-Voigt. A pseudo-Voigt function is essentially a linear combination of Gaussian and Lorentzian functions, which yielded a lower root-sum-square error (RSSE) than either Gaussian or Lorentzian fits when tested independently, so it was selected to model the data to prevent over- or underestimating the resolution. The normalized pseudo-Voigt function used for fitting is given by
where ${f_G}({x;{\gamma_G}} )$ and ${f_L}({x;{\gamma_L}} )$ are the normalized Gaussian and Lorentzian functions, given byFig. 3. Resolution characterization. (A) 500 nm bead (sub-resolution) Y-Z section image and (B) lateral profile, FWHM = 0.608µm. (C) 500 nm bead X-Z section image and (D) axial profile, FWHM = 1.962µm. (E) Gaussian, Lorentzian, and pseudo-Voigt functions fit to measured lateral bead profile. The calculated $\boldsymbol{\eta }$ value indicates that the data follow a roughly equal superposition of Gaussian and Lorentzian profiles. (F) Fit quality comparison for lateral and axial PSF data; RSS error is minimized with the pseudo-Voigt fit for both lateral and axial profiles. All measurements taken in scattering modality with green microLED illumination. Scale bars, 2µm.
Increasing the light sheet modulation frequency could improve the spatial resolution. The maximum spatial frequency that our system can project is 105 mm-1 as determined by the microLED panel’s 4µm pixel pitch and the projection magnification. MicroLED panels provide incoherent light, however, so the super-resolution variant of SIM is inaccessible with this illumination scheme. There remains a tradeoff between spatial resolution and the usable field of view that must be considered for each application which we will now discuss.
The vertical field of view (VFOV) of our system (along the Y-axis) is set by the sensor magnification in the imaging arm as 632µm. The projected height of our light sheets is 760µm to ensure full illumination of the VFOV. While the horizontal field of view (HFOV) as seen from the sensor is 750µm, the range over which optical sectioning occurs is limited by the extent of the projected light sheet along the Z-axis. See the inset of Fig. 1(c) for axis definitions. To quantify the HFOV, we imaged the scattering from an agarose gel with optical sectioning SIM (square wave modulation) and measured the average horizontal profile across the entire height of the image. Scattering makes the light sheet modulation visible, so in the processed images only the regions with strong pattern contrast appear as bright signal. The remaining image noise along the vertical axis is due to a combination of shadows being cast by impurities in the gel suspension and small residual nonuniformity in the microLED panel. The HFOV width was determined by finding the full width of the normalized intensity profile at the 1/e point (Fig. 4). For spatial frequencies of 52.7mm-1, 26.3mm-1, and 13.2mm-1, the HFOV was measured to be 69.8µm, 131.9µm, and 254.3µm, respectively. This result clearly illustrates the tradeoff between light sheet modulation frequency and field of view. Side lobes appear in the HFOV profile due to the Talbot-like patterning of the light sheet through focus, visible in Fig. 4(a)-(c). We typically use a spatial frequency of 26.3 mm-1 for tissue imaging since it has empirically demonstrated a good compromise between FOV and optical sectioning quality, yielding a typical useful FOV of 632µm x 132µm.
Fig. 4. HFOV characterization for structured LSM. (A-C) Captured images of agarose gel under square wave structured light sheet illumination for spatial frequencies of 52.7 mm-1, 26.3 mm-1, and 13.2 mm-1. (D-F) Processed SIM images for each spatial frequency. (G) Horizontal (Z-axis) cross section of processed SIM images, averaged over all rows. The useful HFOV is calculated where the signal drops to 1/e, demonstrating the tradeoff between modulation frequency and HFOV width. Scale bars, 100µm.
One way to increase the HFOV is by placing an electrically tunable lens (ETL; Optotune EL-10-30-TC) in infinity space between the CMO and PMO in the illumination path (see Fig. 1) to perform axially swept light sheet imaging [36,37]. In this configuration, a change in ETL voltage corresponds to a displacement of the light sheet along the illumination optical axis (Z-axis of Fig. 1(c) inset). If the light sheet position is synchronized with the sCMOS rolling shutter, the light sheet appears to cover the entire HFOV and the full sensor resolution will appear to be optically sectioned. Under these conditions, a lateral FOV of 632µm x 750µm is achieved. A comparison of images captured under static and axially swept structured light sheet illumination is shown in Fig. 5, where it can be seen that the static illumination FOV is limited to the central region of the image. SIM processing clearly visualizes the usable FOV, since features illuminated by defocused portions of the light sheet have low modulation contrast and are therefore removed from the optically sectioned image. In the case of axially swept structured light sheet illumination, modulation contrast is high across the entire FOV and a full-field optically sectioned image is the result. Depth into the cucumber increases from left to right in the images, so the right side is dimmer due to increased scattering and absorption along the optical path within the sample.
Fig. 5. Extending the HFOV via axially swept structured light sheet illumination. (A) Processed image of cucumber with static light sheet. (B) Processed image of cucumber with axially swept light sheet, revealing details at image edges that were removed in (A) by SIM processing.
3. Multimodal tissue imaging
To highlight the versatile capabilities of our compact, low-cost, microLED-based structured light sheet microscope, we imaged a variety of biomedical samples in multiple modalities. Label-free contrast is preferable to exogenous fluorescence for in vivo imaging since the latter relies on introducing fluorophores to the patient through either injection or topical application, so we first demonstrate that our microscope can resolve relevant cellular features in the scattering configuration. To further explore what can be achieved with microLED illumination, we show fluorescence and autofluorescence images at multiple excitation wavelengths.
3.1 Porcine gastrointestinal imaging
We first performed scattering and fluorescence imaging of porcine esophageal tissue with the blue microLED panel (450 nm). The tissue was stained with proflavine, a fluorescent dye that stains cell nuclei and cytoplasmic structures, which has been shown to be safe for clinical use [38]. Nuclei shape, size, and arrangement are common features that are relevant to cancer diagnosis, so nuclear fluorescence was used as a ground truth to correlate features in the scattering image to specific, useful cellular morphology. Figure 6 displays an overlay of the two modalities, with arrows pointing out examples where the nucleus can be clearly resolved and interpreted. Further, cell boundaries are clearly defined and can therefore be used to assess disease progression. For example, Barrett’s esophagus is a common precursor to esophageal cancer that is initiated by a change in the normal stratified squamous epithelial cells that line the distal esophagus to a metaplastic columnar epithelium; formal diagnosis is based on the presence of this metaplastic columnar epithelium extending above the gastroesophageal junction into the esophagus [39]. This would present in our images as a morphological change in the cell boundaries and arrangement, so our microscope can feasibly detect this condition. For better visualization of details, the image was processed in Fiji (an augmented version of ImageJ) as described in the Materials and Methods section [40].
Fig. 6. Scattering and fluorescence imaging of porcine esophageal mucosa. The sample was stained with proflavine and fluorescence was excited by 450 nm microLED illumination. (A) Overlay of scattering (gray) and fluorescence (green) images; red arrows indicate nuclei. Detailed views of a single cell in scattering (B) and fluorescence (C) are shown to demonstrate that nuclei can be visualized in scattering modality.
Next, we applied the same imaging modalities to porcine anal tissue. HIV-positive patients are at risk of developing anal cancer precursor lesions, so there is a need for high resolution image-based monitoring as part of a comprehensive treatment plan [41]. Figure 7(a) shows an overlay image of proflavine fluorescence from nuclei and scattering signal. The blue microLED panel was again used to excite fluorescence, but the scattering image was captured with red microLED illumination (635 nm) to take advantage of the notable increase in imaging depth that occurs for biological tissue around 600 nm [42]. Figures 7(b)-(c) show scattering images captured from a different sample region under blue and red illumination, respectively. This side-by-side comparison experimentally demonstrates that red illumination enhances feature fidelity at depth and decreases haze near the surface thanks to the improved penetration depth of longer wavelengths. The yellow arrows indicate examples of features that are more detectable under red illumination.
Fig. 7. Scattering and fluorescence imaging of porcine anal tissue. (A) Overlay image of scattering (gray) captured with 635 nm illumination and proflavine fluorescence (green) excited by 450 nm illumination. Comparison of scattering images taken with (B) 450 nm and (C) 635 nm illumination highlights the increase in imaging depth with longer wavelength. Yellow arrows indicate example deep features that are more clearly seen under red illumination.
3.2 Mouse kidney
Autofluorescence is another useful tool for clinical imaging and has been shown to aid the detection of esophageal squamous neoplasia, Barrett’s esophagus, gastric neoplasia, colon polyps, and colitis-associated neoplasia, among other conditions [43]. While disadvantageous for deep penetration in tissue, the shorter wavelength microLED panels provide two good options for autofluorescence excitation (450 nm and 405 nm) allowing for excitation of collagen, elastin, flavins, protoporphyrin IX, and lipofuscins [44]. During kidney transplant, for example, this information can help surgeons determine the level of ischemic injury and assess organ viability prior to operating on the patient [45]. To demonstrate our microscope’s ability to excite and detect autofluorescence with microLED illumination, we imaged a healthy, unlabeled mouse kidney with 450 nm microLED excitation. A 500 nm cut-on long pass filter (Edmund 62983) was used to isolate the autofluorescence signal, and then removed from the optical path to observe the scattering signal. Figure 8 shows an overlay of autofluorescence and scattering images alongside a representative histologic image for qualitative comparison. The spatial extent and orientation of kidney tubules are clearly seen in the autofluorescence signal and are located relative to the tissue surface by the scattering signal. Although short wavelength excitation was used, we are still able to detect meaningful detail 100µm below the tissue surface.
Fig. 8. Comparison of microLED autofluorescence LSM with histology. (A) Overlay of scattering (gray) and autofluorescence (yellow) captured under 450 nm illumination. (B) Representative histological image showing similar tubule structure.
3.3 Mouse brain
We next imaged a 100 µm thick slice of a mouse brain hippocampus, in which protein kinase C delta (PKCδ) neurons were transgenically labeled with tdTomato. To enable fluorescent tdTomato expression in PKCδ neurons, Jackson Lab transgenic mouse strains 028437 and 007908 were crossed.
Fluorescence was excited by the green microLED (520 nm) and collected through a 600/50 bandpass emission filter (Edmund 84-785). Figure 9 shows a 632µm x 132µm x 113µm volumetric image (maximum intensity projection) that is color-coded by depth to show the relative position of neurons in space. No further image processing was applied. A side-by-side comparison shows that structured light sheet illumination increases rejection of background haze originating from features outside of the focal plane, yielding an image with enhanced contrast. Notably, in addition to labeled soma, neuronal projections extending from individual soma are clearly resolved and assigned depth information. These are typically on the order of 1µm in diameter and therefore not resolved by many low-cost approaches to mouse brain imaging. Visualizing axonal and dendritic structure and orientation provides important information about inter-neuron communication, connectivity, neural plasticity, and stimulus responses. In addition to improving understanding of connectivity in the brain itself, the ability to assess fine neuronal projections may also be useful for evaluating disease states involving denervation, including ALS, age-related muscle atrophy, and diabetic neuropathy [46–48]. Assessment of neuronal connectivity and peripheral innervation is critical for rigorous neuroscience experimentation. With access to this information, researchers could adapt our microLED structured LSM system to serve a variety of health-related applications.
Fig. 9. Volumetric fluorescence imaging of 100µm mouse brain hippocampus slice labeled with tdTomato. Full FOV images are shown under (A) standard light sheet and (B) structured light sheet illumination. Detail views of the red boxed region are shown for (C) standard and (D) structured illumination, revealing intricate neuron projection morphology. The color-coded depth scale extends beyond 100µm due to the 45-degree orientation of the microscope’s imaging axis relative to the sample surface normal. Visualization 2 shows a fly-through along the X-axis (see Fig. 1(c) for coordinate system).
4. Conclusion
In summary, we have developed a simplified version of three-dimensional structured light sheet microscopy by leveraging a microLED panel as the source. The primary advantage of our system is that it has a very compact illumination arm wherein all the illumination structure is generated at the source. This yields a small package that is more cost efficient and less prone to mechanical error than other implementations of structured light sheet microscopy, while maintaining the flexibility of dynamically tunable light sheet width, axial location, and modulation frequency. While previous approaches accomplish this through a reliance on complex optical components, such as acousto-optical tunable filters, galvo mirrors, and spatial light modulators, we require only a single microLED panel to offer this versatility with no moving parts. Although other parts of the system still rely on relatively expensive hardware, such as the imaging camera, our simplified illumination path represents a notable cost reduction compared to conventional structured light sheet microscopes. We have shown that the current system can provide three-dimensional optically sectioned images that reveal the morphological features most commonly used for disease detection and diagnosis (e.g. nucleus and cell boundaries). Neuroscience research is another area where this microscope can find use, as our results indicate sufficient resolution to identify and locate neuron projections. The reduced footprint and lower cost may be attractive to labs where benchtop space is a premium and budgets are restricted. The complexity of existing structured light sheet microscopes has precluded their adoption into clinical settings, but our results show that this novel configuration can yield images that may prove useful for disease screening.
The imaging objective focus mechanism is the only remaining mechanical motion in the system, which could potentially be supplanted by an electrically tunable lens or computational refocusing algorithm in future work. Currently, wavelength switching is achieved by swapping out microLED panels; this could be simplified by combining multiple panels with a beamsplitter cube to enable simultaneous multi-color imaging. Displays with red, blue, and green pixels do exist, but the projected illumination spatial resolution would suffer due to the Bayer pattern and pseudo-continuous modulation would no longer be possible. Thanks to its simple illumination train, this microscope could be packaged into a handheld device and tested for clinical skin and oral imaging, where regions of interest are relatively accessible. Freeform optics could be utilized to further miniaturize the device for endoscopic imaging of the gastrointestinal tract and cervix. These same optimizations would also lend themselves to the development of head-mounted brain imaging devices for neuroscience studies.
MicroLED display technology is in a state of rapid development thanks to strong interest from the AR/VR industry. Fortunately, there is a strong overlap between the specifications that benefit the performance of head mounted displays and biomedical imaging devices. It is expected that microLED panels will continue driving toward higher power outputs in the coming years, allowing for increased imaging depth and shorter exposure times. The constant push toward smaller pixels will yield higher resolution illumination engines as time goes on. In our case, this would enable better optical sectioning and decreased photodamage thanks to thinner light sheets, and possibly improve spatial resolution through higher frequency pattern modulation. These two advancements will address some of the limitations of this technique in its current state.
Because microLED panels have been developed principally for AR/VR applications to date, available wavelengths are biased toward the visible spectrum. This makes it difficult to leverage windows of low absorption and scattering in the near- and mid-infrared bands that would significantly increase imaging depth in biological tissue. Further, workhorse imaging modalities such as two-photon fluorescence and second harmonic generation are inaccessible due to the additional need for femtosecond pulsing with high peak powers. It is our hope, however, that our results in the visible spectrum make a compelling case for biomedical imaging as a fruitful application of microLED panels that will inspire researchers to work toward overcoming these limitations. For now, replacing the microLED panel with a digital micromirror device and external light source could expand the wavelength capabilities of our system at the expense of increased complexity, but the peak powers required for multiphoton phenomena may remain elusive.
MicroLED-based structured light sheet microscopy takes advantage of emerging technologies in the display industry to provide a compact, low-cost means to volumetrically image biomedical samples in both labeled and label-free modalities. This configuration can theoretically be applied wherever traditional light sheet microscopy is used—as we have demonstrated for ex vivo imaging of animal tissue and neurons—and is amenable to miniaturization for handheld clinical devices. Many fields within the life sciences may benefit from the simplicity and flexibility of this approach to structured light sheet microscopy.
5. Materials and methods
5.1 Sample preparation
To create the fluorescent sample in Fig. 2, a piece of lens tissue paper was mounted on the surface of a 2% agarose gel (IVYX Scientific), topically stained with proflavine for 1-2 minutes (0.01% w/v in distilled water, AdooQ Bioscience), and immersed in ultrasound gel (Aquasonic Clear) for imaging.
Nuclear staining in Fig. 6 and 7 was accomplished by submerging the tissue samples in a solution of proflavine (0.01% w/v) and distilled water for approximately 5 minutes. Once removed, the samples were rinsed with 1x PBS and imaged immediately.
To prepare the sample shown in Fig. 9, a 2.5-month-old male mouse was put under isoflurane anesthesia and cardiac puncture was used to euthanize and perfuse 1x PBS through the circulation. Next, 4% PFA was perfused to fix the tissues. The brain was removed and placed in 4% PFA, at 4C, overnight. The brain was then rinsed 3x with 1xPBS and sliced in 1x PBS using a GS429 vibratome (100µm sections). Slices were mounted using Vectashield Plus mounting media and stored in -20C. The IACUC protocol number for animal research shown in this report is 15-059.
5.2 Image processing
For better visualization of details, the image in Fig. 6 was processed in Fiji (an augmented version of ImageJ) using “Remove Outliers” (radius = 2.0, threshold = 100, bright), “Gaussian Blur” (radius = 3.0, weight = 0.5), “Unsharp Mask” (radius = 3.0, weight = 0.5), and contrast-limited adaptive histogram equalization (CLAHE) (blocksize = 5, bins = 256, max slope = 3.0, mask = none). In Fig. 7(a), nuclei contrast was improved in Fiji using “Subtract Background” (radius = 50) and CLAHE (blocksize = 5, bins = 256, max slope = 3.0, mask = none). CLAHE was used in Fig. 7(b)-(c) to highlight the difference in local contrast. The images in Fig. 8(a) were processed similarly to those in Fig. 6.
Funding
National Institute of Biomedical Imaging and Bioengineering (R21EB029599); National Institute of Dental and Craniofacial Research (R01DE030682, R21DE028734); National Institutes of Health (S10OD018061).
Acknowledgments
Research reported in this publication was supported by the National Cancer Institute of the National Institutes of Health under award number P30 CA023074. Histology services were provided by the Tissue Acquisition and Cellular/Molecular Analysis Shared Resource (TACMASR) at the University of Arizona Cancer Center. MicroLED development kits were provided by Jade Bird Display (JBD).
Disclosures
R.L. is the founder of Light Research Inc., which didn’t support this research.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. R. D. Neal, P. Tharmanathan, B. France, et al., “Is increased time to diagnosis and treatment in symptomatic cancer associated with poorer outcomes? Systematic review,” Br. J. Cancer 112(S1), S92–S107 (2015). [CrossRef]
2. N. Bedard, M. Pierce, A. El-Naggar, S. Anandasabapathy, A. Gillenwater, and R. Richards-Kortum, “Emerging roles for multimodal optical imaging in early cancer detection: a global challenge,” Technol. Cancer Res. Treat. 9(2), 211–217 (2010). [CrossRef]
3. E. G. Fischer, “Nuclear morphology and the biology of cancer cells,” Acta Cytol. 64(6), 511–519 (2020). [CrossRef]
4. L. He, L. R. Long, S. Antani, and G. R. Thoma, “Histology image analysis for carcinoma detection and grading,” Comput. Methods Programs Biomed. 107(3), 538–556 (2012). [CrossRef]
5. M. Rajadhyaksha, A. Marghoob, A. Rossi, A. C. Halpern, and K. S. Nehal, “Reflectance confocal microscopy of skin in vivo: from bench to bedside,” Lasers Surg. Med. 49(1), 7–19 (2017). [CrossRef]
6. J. Champin, J.-L. Perrot, E. Cinotti, B. Labeille, C. Douchet, G. Parrau, F. Cambazard, P. Seguin, and T. Alix, “In vivo reflectance confocal microscopy to optimize the spaghetti technique for defining surgical margins of lentigo maligna,” Dermatol. Surg. 40(3), 247–256 (2014). [CrossRef]
7. Y. Li, Z. Zhu, J. J. Chen, J. C. Jing, C.-H. Sun, S. Kim, P.-S. Chung, and Z. Chen, “Multimodal endoscopy for colorectal cancer detection by optical coherence tomography and near-infrared fluorescence imaging,” Biomed. Opt. Express 10(5), 2419–2429 (2019). [CrossRef]
8. K. Li, W. Liang, J. Mavadia-Shukla, H.-C. Park, D. Li, W. Yuan, S. Wan, and X. Li, “Super-achromatic optical coherence tomography capsule for ultrahigh-resolution imaging of esophagus,” J. Biophotonics 12, e201800205 (2019). [CrossRef]
9. B. G. Muller, D. M. de Bruin, M. J. Brandt, W. van den Bos, S. van Huystee, D. J. Faber, D. Savci, P. J. Zondervan, T. M. de Reijke, M. P. Laguna-Pes, T. G. van Leeuwen, and J. J. M. C. H. de la Rosette, “Prostate cancer diagnosis by optical coherence tomography: First results from a needle based optical platform for tissue sampling,” J. Biophotonics 9(5), 490–498 (2016). [CrossRef]
10. S. W. Perry, R. M. Burke, and E. B. Brown, “Two-photon and second harmonic microscopy in clinical and translational cancer research,” Ann. Biomed. Eng. 40(2), 277–291 (2012). [CrossRef]
11. R. Pal, P. Villarreal, X. Yu, S. Qiu, and G. Vargas, “Multimodal widefield fluorescence imaging with nonlinear optical microscopy workflow for noninvasive oral epithelial neoplasia detection: a preclinical study,” J. Biomed. Opt. 25(11), 116008 (2020). [CrossRef]
12. Z. Huang, M. S. Bergholt, W. Zheng, K. Lin, K.-Y. Ho, M. The, and K.-G. Yeoh, “In vivo early diagnosis of gastric dysplasia using narrow-band image-guided Raman endoscopy,” J. Biomed. Opt. 15(3), 037017 (2010). [CrossRef]
13. J. Desroches, M. Jermyn, M. Pinto, F. Picot, M.-A. Tremblay, S. Obaid, E. Marple, K. Urmey, D. Trudel, G. Soulez, M.-C. Guiot, B. C. Wilson, K. Petrecca, and F. Leblond, “A new method using Raman spectroscopy for in vivo targeted brain cancer tissue biopsy,” Sci. Rep. 8(1), 1792 (2018). [CrossRef]
14. A. Garcia-Uribe, J. Zou, M. Duvic, J. H. Cho-Vega, V. G. Prieto, and L. V. Wang, “In vivo diagnosis of melanoma and nonmelanoma skin cancer using oblique incidence diffuse reflectance spectrometry,” Cancer Res. 72(11), 2738–2745 (2012). [CrossRef]
15. M. Halicek, H. Fabelo, S. Ortega, G. M. Callico, and B. Fei, “In-vivo and ex-vivo tissue analysis through hyperspectral imaging techniques: revealing the invisible features of cancer,” Cancers 11(6), 756 (2019). [CrossRef]
16. S. Ortega, H. Fabelo, D. K. Iakovidis, A. Koulaouzidis, and G. M. Callico, “Use of hyperspectral/multispectral imaging in gastroenterology. shedding some–different–light into the dark,” J. Clin. Med. 8(1), 36 (2019). [CrossRef]
17. S. Li and R. Liang, “DMD-based three-dimensional chromatic confocal microscopy,” Appl. Opt. 59(14), 4349–4356 (2020). [CrossRef]
18. J. Mertz, Introduction to Optical Microscopy, 2nd ed. (Cambridge University Press, 2019).
19. A. K. Glaser, N. P. Reder, Y. Chen, E. F. McCarty, C. Yin, L. Wei, Y. Wang, L. D. True, and J. T. C. Liu, “Light-sheet microscopy for slide-free non-destructive pathology of large clinical specimens,” Nat. Biomed. Eng. 1(7), 1–10 (2017). [CrossRef]
20. K. B. Patel, W. Liang, M. J. Casper, V. Voleti, W. Li, A. J. Yagielski, H. T. Zhao, C. Perez Campos, G. S. Lee, J. M. Liu, E. Philipone, A. J. Yoon, K. P. Olive, S. M. Coley, and E. M. C. Hillman, “High-speed light-sheet microscopy for the in-situ acquisition of volumetric histological images of living tissue,” Nat. Biomed. Eng. 6(5), 569–583 (2022). [CrossRef]
21. F. O. Fahrbach, P. Simon, and A. Rohrbach, “Microscopy with self-reconstructing beams,” Nat. Photonics 4(11), 780–785 (2010). [CrossRef]
22. S. L. Reidt, D. J. O’Brien, K. Wood, and M. P. MacDonald, “Polarised light sheet tomography,” Opt. Express 24(10), 11239–11249 (2016). [CrossRef]
23. T. Wu, C.-W. Sher, Y. Lin, C.-F. Lee, S. Liang, Y. Lu, S.-W. Huang Chen, W. Guo, H.-C. Kuo, and Z. Chen, “Mini-LED and Micro-LED: promising candidates for the next generation display technology,” Appl. Sci. 8(9), 1557 (2018). [CrossRef]
24. Z. Chen, S. Yan, and C. Danesh, “MicroLED technologies and applications: characteristics, fabrication, progress, and challenges,” J. Phys. D: Appl. Phys. 54(12), 123001 (2021). [CrossRef]
25. Y. Huang, E.-L. Hsiang, M.-Y. Deng, and S.-T. Wu, “Mini-LED, Micro-LED and OLED displays: present status and future perspectives,” Light: Sci. Appl. 9(1), 105 (2020). [CrossRef]
26. S. Li, B. Song, T. Peterson, J. Hsu, and R. Liang, “MicroLED chromatic confocal microscope,” Opt. Lett. 46(11), 2722–2725 (2021). [CrossRef]
27. T. Breuninger, K. Greger, and E. H. K. Stelzer, “Lateral modulation boosts image quality in single plane illumination fluorescence microscopy,” Opt. Lett. 32(13), 1938–1940 (2007). [CrossRef]
28. J. Mertz and J. Kim, “Scanning light-sheet microscopy in the whole mouse brain with HiLo background rejection,” J. Biomed. Opt. 15(01), 016027 (2010). [CrossRef]
29. P. J. Keller, A. D. Schmidt, A. Santella, K. Khairy, Z. Bao, J. Wittbrodt, and E. H. K. Stelzer, “Fast, high-contrast imaging of animal development with scanned light sheet–based structured-illumination microscopy,” Nat. Methods 7(8), 637–642 (2010). [CrossRef]
30. T. A. Planchon, L. Gao, D. E. Milkie, M. W. Davidson, J. A. Galbraith, C. G. Galbraith, and E. Betzig, “Rapid three-dimensional isotropic imaging of living cells using Bessel beam plane illumination,” Nat. Methods 8(5), 417–423 (2011). [CrossRef]
31. M. Lei and A. Zumbusch, “Structured light sheet fluorescence microscopy based on four beam interference,” Opt. Express 18(18), 19232–19241 (2010). [CrossRef]
32. A. D. Edelstein, M. A. Tsuchida, N. Amodaj, H. Pinkard, R. D. Vale, and N. Stuurman, “Advanced methods of microscope control using µManager software,” J. Biol. Methods 1(2), e10 (2014). [CrossRef]
33. M. a. A. Neil, R. Juškaitis, and T. Wilson, “Method of obtaining optical sectioning by using structured light in a conventional microscope,” Opt. Lett. 22(24), 1905–1907 (1997). [CrossRef]
34. K. Becker, S. Saghafi, M. Pende, I. Sabdyusheva-Litschauer, C. M. Hahn, M. Foroughipour, N. Jährling, and H.-U. Dodt, “Deconvolution of light sheet microscopy recordings,” Sci. Rep. 9(1), 17625 (2019). [CrossRef]
35. T. Ida, M. Ando, and H. Toraya, “Extended pseudo-Voigt function for approximating the Voigt profile,” J. Appl. Crystallogr. 33(6), 1311–1316 (2000). [CrossRef]
36. K. M. Dean, P. Roudot, E. S. Welf, G. Danuser, and R. Fiolka, “Deconvolution-free subcellular imaging with axially swept light sheet microscopy,” Biophys. J. 108(12), 2807–2815 (2015). [CrossRef]
37. P. N. Hedde and E. Gratton, “Selective plane illumination microscopy with a light sheet of uniform thickness formed by an electrically tunable lens,” Microsc. Res. Tech. 81(9), 924–928 (2018). [CrossRef]
38. N. Pantano, B. Hunt, R. A. Schwarz, S. Parra, K. Cherry, J. C. Possati-Resende, A. Longatto-Filho, J. H. T. G. Fregnani, P. E. Castle, K. Schmeler, and R. Richards-Kortum, “Is Proflavine Exposure Associated with Disease Progression in Women with Cervical Dysplasia? A Brief Report,” Photochem. Photobiol. 94(6), 1308–1313 (2018). [CrossRef]
39. S. J. Spechler, P. Sharma, R. F. Souza, J. M. Inadomi, and N. J. Shaheen, “American Gastroenterological Association Technical Review on the Management of Barrett’s Esophagus,” Gastroenterology 140(3), e18–e52 (2011). [CrossRef]
40. J. Schindelin, I. Arganda-Carreras, E. Frise, V. Kaynig, M. Longair, T. Pietzsch, S. Preibisch, C. Rueden, S. Saalfeld, B. Schmid, J.-Y. Tinevez, D. J. White, V. Hartenstein, K. Eliceiri, P. Tomancak, and A. Cardona, “Fiji: an open-source platform for biological-image analysis,” Nat. Methods 9(7), 676–682 (2012). [CrossRef]
41. F. Gimenez, I. T. da Costa-e-Silva, A. Daumas, J. de Araújo, S. G. Medeiros, and L. Ferreira, “The value of high-resolution anoscopy in the diagnosis of anal cancer precursor lesions in hiv-positive patients,” Arq. Gastroenterol. 48(2), 136–145 (2011). [CrossRef]
42. J. Zhao, N. Kulkarni, E. Dobo, M. J. Khan, E. Yang, D. Kang, and D. Kang, “Investigation of different wavelengths for scattering-based light sheet microscopy,” Biomed. Opt. Express 13(7), 3882–3892 (2022). [CrossRef]
43. L.-M. Wong Kee Song, S. Banerjee, D. Desilets, D. L. Diehl, F. A. Farraye, V. Kaul, S. R. Kethu, R. S. Kwon, P. Mamula, M. C. Pedrosa, S. A. Rodriguez, and W. M. Tierney, “Autofluorescence imaging,” Gastrointest. Endosc. 73(4), 647–650 (2011). [CrossRef]
44. A. C. Croce and G. Bottiroli, “Autofluorescence spectroscopy and imaging: a tool for biomedical research and diagnosis,” Eur. J. Histochem. 58, 2461 (2014). [CrossRef]
45. J. T. Fitzgerald, A. P. Michalopoulou, C. D. Pivetti, R. N. Raman, C. Troppmann, and S. G. Demos, “Real-time assessment of in vivo renal ischemia using laser autofluorescence imaging,” J. Biomed. Opt. 10(4), 044018 (2005). [CrossRef]
46. P. J. Blijham, H. J. Schelhaas, H. J. Ter Laak, B. G. M. van Engelen, and M. J. Zwarts, “Early diagnosis of ALS: the search for signs of denervation in clinically normal muscles,” J. Neurol. Sci. 263(1-2), 154–157 (2007). [CrossRef]
47. K. H. J. Park, “Mechanisms of muscle denervation in aging: insights from a mouse model of amyotrophic lateral sclerosis,” Aging Dis. 6(5), 380–389 (2015). [CrossRef]
48. C.-T. Shun, Y.-C. Chang, H.-P. Wu, S.-C. Hsieh, W.-M. Lin, Y.-H. Lin, T.-Y. Tai, and S.-T. Hsieh, “Skin denervation in type 2 diabetes: correlations with diabetic duration and functional impairments,” Brain 127(7), 1593–1605 (2004). [CrossRef]
