Single-pixel imaging (SPI) is an advanced computational imaging technique that uses a single-pixel detector to reconstruct two-dimensional images^[1]. Unlike traditional systems with pixel-array detectors that directly capture spatial information, SPI employs spatial encoding patterns to modulate the optical field. These patterns convert spatial data into one-dimensional temporal sequences, which are measured by the single-pixel detector. Advanced computational algorithms then process these signals to reconstruct the images. By eliminating the need for pixel-array detectors, SPI is particularly effective in spectral domains where such detectors are challenging to fabricate such as terahertz^[2], X-ray^[3], or infrared wavelengths^[4].

The reliance on a single-pixel detector allows SPI to excel in challenging conditions such as environments with strong scattering or limited light availability^[5,6], which often compromise traditional imaging methods. Moreover, the computational nature of SPI enables the integration of advanced data acquisition techniques such as compressive sensing^[7], which significantly reduces the number of measurements required for high-quality image reconstruction. This makes SPI particularly effective for imaging through scattering media^[5], where traditional approaches struggle to produce clear images, or in weak-light conditions^[8], where efficient light detection is critical.

In addition to these advantages, the modular design of SPI allows seamless integration with advanced optical configurations and reconstruction algorithms, further enhancing its versatility. For instance, SPI can be adapted for applications such as hyperspectral imaging^[9], polarization imaging^[10], and dynamic imaging of time-varying scenes^[11,12], significantly expanding its scope. These capabilities emphasize SPI’s potential as a transformative imaging modality, addressing complex challenges across a range of fields including biomedical imaging, industrial inspection, and environmental monitoring.

Historically, SPI research has predominantly focused on macroscopic objects and scenes, demonstrating its diverse imaging capabilities in applications such as environmental monitoring, industrial inspection, and object recognition. Recently, however, only a few pioneering research groups have shifted their attention toward developing microscopic versions of SPI, achieving remarkable resolutions down to the micrometer scale^[13-18]. This advancement has significantly expanded the utility of SPI, enabling it to capture intricate structural details in biological specimens with unprecedented clarity. Notable examples include the holographic imaging of insect wings^[19], which reveals fine surface patterns, as well as the visualization of mouse tissues^[15], blood smears^[16], and pumpkin stem samples^[16], offering insights into their microstructures. Beyond imaging amplitudes and phases, researchers have also explored polarization imaging at micron-level resolution^[18], successfully differentiating between normal and cancerous esophageal tissues. This functional enhancement highlights SPI’s growing potential as a diagnostic tool. Alternatively, SPI employs wide-field illumination, which offers a significant advantage in signal-to-noise ratio compared to conventional point-by-point scanning microscopy. This advantage is particularly important for biological specimens with limited illumination intensity.

In this work, we utilize a custom-built high-resolution SPI system developed by our research group to image a diverse range of biological specimens. Specifically designed for microscopic applications, this system integrates heterodyne holography with advanced spatial encoding techniques and computational algorithms, achieving a high space-bandwidth-time product of 41,667 pixel/s and an exceptional lateral resolution of 4–5 $\mathrm{\mu m}$. Our experimental results highlight the versatility of SPI in resolving fine structural details across various biological specimens, from tissue slices to cellular-scale structures. The system’s adaptability, enabled by the enhanced resolution and flexibility provided by heterodyne holography, underscores its potential for diverse biological and medical imaging applications. These capabilities position it as a promising tool for advancing both fundamental research and clinical diagnostics.

The experimental setup is illustrated in Fig. 1. A continuous-wave laser (Verdi G5, Coherent) operating at a wavelength of 532 nm was used as the light source for the system. The optical path was split into a reference beam and a signal beam with an adjustable power ratio using a half-wave plate and a polarizing beam splitter. The polarization state of each beam was further refined using corresponding half-wave plates. Both beams were modulated by acousto-optic modulators (AOM-505AF1, IntraAction), generating a beat frequency that facilitated natural phase shifting over time. Each beam was subsequently expanded through a pair of lenses with focal lengths of 8 and 250 mm, respectively, before being directed onto a digital micromirror device (DMD).

The DMD used in this system has a resolution of $768\mathrm{pixel}\times 1024\mathrm{pixel}$, a pixel size of 13.68 $\mathrm{\mu m}$, and a refresh rate of 22 kHz (V7001 DLP7000 & DLPC410, Vialux & Texas Instruments). For this experiment, only $768\mathrm{pixel}\times 768\mathrm{pixel}$ were utilized, employing a $3\times 3$ binning strategy to enhance operational efficiency. Hadamard bases with orders up to 256 were selected for pattern projection. Notably, the DMD cannot directly generate Hadamard bases but instead displays binary patterns (0 or 1). These non-orthogonal patterns can introduce artifacts such as an exaggerated intensity in the first pixel and a uniform background across the entire image. However, these issues are well-documented in the literature and can be effectively mitigated using established correction techniques.

After being reflected by the DMD, the patterned light was reduced in size by a factor of 10 using a 4f optical system consisting of two lenses with focal lengths of 300 and 30 mm. The optical parameters of this setup were consistent with those described in Refs. [15,18], achieving a lateral resolution of 4–5 $\mathrm{\mu m}$. These values are comparable to the performance of our previously reported holographic and polarization microscope systems. Further enhancement of the spatial resolution can be achieved by increasing the numerical aperture of the illumination, which may be accomplished through the use of combinations of objective lenses. After interacting with the biological specimen, the transmitted light was combined with the similarly scaled reference beam and collected using a lens with a focal length of 125 mm. Finally, the combined optical power was measured by a photodetector (DET36A2, Thorlabs) with a 25 MHz bandwidth, and the signal was digitized using a data acquisition card (USB-6251, National Instruments) at a sampling rate of 1.25 Ms/s.

Benefiting from the heterodyne holography employed in this system, which allows the signal to evolve over time, both amplitude and phase information can be acquired with a single DMD display. Given that the refresh time of the DMD was set to 0.048 ms, the space-bandwidth-time product, defined as the effective information throughput collected per unit time, can be estimated as $2\text{pixel/}0.048\mathrm{ms}=41,667\text{pixel/}\mathrm{s}$.

We first demonstrate the capability of amplitude imaging using a stained biological specimen. Figure 2(a) depicts a slice of epithelial cells, where the red staining effectively highlights distinct cellular structures, including densely packed clusters and well-defined cellular boundaries. This image was captured using a commercial microscope system. Within the stained sample, the yellow dashed diamond marks a region of interest selected for its rich structural complexity and visible cellular heterogeneity. This specific region was subsequently imaged using the developed SPI system, and the results are presented in Figs. 2(b) and 2(c), showcasing the system’s ability to resolve fine structural details with high precision.

The SPI results present high-resolution amplitude images of the selected region, unveiling fine structural details with exceptional clarity. Subtle intensity variations across the area indicate differences in cellular density and staining concentration, reflecting the inherent biological heterogeneity within the tissue. Distinct cell boundaries and intracellular features are clearly discernible, offering valuable insights into the spatial organization and structural complexity of the specimen. For stained specimens, we did not provide phase images as they closely resembled their amplitude counterparts.

We further demonstrate the capability of the SPI system by performing amplitude imaging on stained esophageal cancer cells, highlighting its potential for analyzing complex tissue structures. In Fig. 3(a), a stained tissue section reveals key anatomical features, such as muscle fibers, which are prominently visible in the red-stained regions. The yellow dashed diamond outlines a region of interest rich in structural details, selected for detailed analysis using the SPI system. The corresponding SPI amplitude image in Fig. 3(b) captures intricate spatial patterns within the marked area, with distinct contrasts indicating variations in tissue density and composition.

In Fig. 3(c), a stained tissue section highlights macrophage cells within the tumor microenvironment, emphasizing regions of cellular heterogeneity and dense biological structures. The yellow dashed diamond delineates a specific area selected for detailed analysis using the SPI system. The corresponding SPI amplitude image in Fig. 3(d) reveals fine structural details, effectively resolving cellular boundaries and subcellular features within the marked region. Variations in amplitude intensity across the region provide valuable insights into the local microstructure, reflecting differences in cellular distribution and staining properties.

These results demonstrate the high-resolution capabilities of the SPI system in visualizing critical biological features at the microscopic scale. The ability to clearly distinguish features such as muscle fibers and macrophage cells demonstrates the system’s potential for detailed morphological analysis of cancerous tissues. This capability opens avenues for further exploration of tumor microenvironments and cellular interactions, highlighting the versatility of SPI for both diagnostic and research applications in biomedical imaging.

For unstained biological specimens, phase imaging serves as a crucial complement to amplitude imaging, offering additional insights into structural and morphological variations that may not be discernible with amplitude imaging alone. We demonstrated the holographic imaging capabilities of our high-resolution SPI system by imaging an unstained sample associated with human epidermal growth factor receptor 2 (HER2) amplification. HER2 is a protein that promotes cell growth and division, and its overexpression or gene amplification is commonly linked to aggressive forms of breast cancer. This biomarker is routinely analyzed using fluorescence in situ hybridization in molecular diagnostics, aiding in the prognosis and treatment planning for patients.

Figure 4(a) displays a bright-field image of the unstained HER2 amplification sample, with two regions of interest marked by red dashed diamonds, selected for further analysis using the SPI system. The corresponding SPI amplitude and phase images are presented in Figs. 4(b) and 4(c), respectively. These images effectively capture subtle structural variations within the sample such as cellular boundaries, intracellular regions, and fine morphological features. Phase imaging, in particular, provides additional contrast and information beyond what is offered by amplitude imaging, making it an invaluable tool for studying the intricate details of unstained biological specimens.

We further demonstrated the versatility of the SPI system by imaging unstained mouse brain tissue slices, with a focus on its phase imaging capabilities for uncovering detailed structural information. Figure 5(a) shows a bright-field image of the brain tissue, with two regions of interest marked by red dashed diamonds. These regions were carefully chosen to emphasize key structural features including distinct tissue boundaries and subtle textural variations within the sample.

The SPI-generated amplitude and phase images of the marked regions are presented in Figs. 5(b) and 5(c), respectively. Together, these images provide a comprehensive view of the refractive index distribution within the tissue, offering enhanced contrast between different regions. The amplitude images capture variations in structural density, while the phase images reveal localized phase shifts, uncovering intricate details along tissue boundaries and within the surrounding areas. This dual imaging approach highlights the capability of the SPI system to resolve fine-scale structural differences and spatial organization in unstained biological tissues, reinforcing its potential for advanced morphological and structural analysis.

In this study, we have demonstrated the versatility and effectiveness of our high-resolution SPI system for imaging both stained and unstained biological specimens. For stained specimens, including epithelial and esophageal cancer tissues, the SPI system effectively captured detailed amplitude images, providing high contrast and fine structural resolution comparable to conventional microscopy methods. For unstained samples such as HER2 amplification specimens and mouse brain tissue slices, the SPI system further showcased its capability by generating complementary phase images. These phase images revealed subtle refractive index variations and intricate structural details that were often undetectable in amplitude images, highlighting the system’s potential for label-free imaging.

It is also worth noting that the imaging region demonstrated in this study has a diamond shape, which is consistent with previous reports^[15,18]. This is due to the need to consider the diffraction effects of the DMD grating at high resolution. To maximize diffraction efficiency and avoid artifacts, a specific angle between the DMD and the incident light is chosen, resulting in the diamond-shaped imaging region. While it is possible to transform the imaging region into a square through coordinate transformations, this is beyond the scope of this work, and thus we do not discuss it further.

The ability of the SPI system to achieve high-resolution imaging through both amplitude and phase modalities underscores its broad applicability in biomedical imaging. Its adaptability to diverse sample types and imaging conditions makes it a valuable tool for various applications including tumor diagnostics, tissue characterization, and fundamental neuroscience research. Overall, the developed SPI system represents a significant advancement in computational imaging, providing a promising platform for future innovations in both clinical diagnostics and research settings.