Open Access
Add to BibSonomy
Abstract
Point-of-care testing (POCT) plays an increasingly important role in biomedical research and health care. Quantitative phase microscopes (QPMs) with good contrast, no invasion, no labeling, high speed and automation could be effectively applied for POCT. However, most QPMs are fixed on the optical platform with bulky size, lack of timeliness, which remained challenging in POCT solutions. In this paper, we proposed a plug-and-play QPM with multimode imaging based on the quantitative differential phase contrast (qDPC) method. The system employs a programmable LED array as the light source and uses the GPU to accelerate the calculation, which can realize multi-contrast imaging with six modes. Accurate phase measurement and real-time phase imaging are implemented by the proposed qDPC algorithms for quantitative phase targets and biomedical samples. A 3D electric control platform is designed for mechanical control of field of view and focusing without manual operations. The experimental results verify the robustness and high performance of the setup. Even a rookie could finish the POCT scheme for biomedical applications at the scene using the QPM with a compact size of 140 × 165 × 250 mm3.
© 2024 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Point-of-care testing (POCT), used to describe diagnostic testing performed with “field” conditions like outside laboratory or hospital, plays a critical role in biomedical detections [1]. The aim of POCT is to collect samples at or near the location of patients and obtain results in a short time to make the necessary assay [2,3]. Cheaper, faster, and more intelligent POCT devices have been widely adopted in health care [4,5].
As an essential detection method, microscopic imaging with high contrast, no labeling, and no invasion attracts increasing attention in biomedical research and clinical diagnosis [6], which can be implemented by quantitative phase imaging (QPI) technology [7]. In recent years, there have been some commercial quantitative phase microscopes (QPMs) based on a typical QPI method, holography, with integrated structure and high-precision phase imaging, including SLIM [8], GLIM [9], wDPM (PhiOptics, USA), DHM (Lyncee Tec, Switzerland), DSX-1000 (Olympus, Japan), DPXM12000 (DeltaPix, Demark), and FM-DHM500 (SCILAB, China) [10]. However, those QPMs are high-cost and limited on the optical platform to keep the stable interferograms, which makes it hard to meet the requirements of timeliness and portability for POCT.
Benefitting from its simplicity and extensibility, differential phase contrast (DPC) imaging shows good potential in POCT solutions. A smartphone-based multi-contrast microscope was proposed using the color-complexing DPC (cDPC) imaging [11,12]. Furthermore, the deep learning technique was introduced to improve DPC imaging on mobile phones [13]. However, none of these portable DPC imaging devices mentioned above had phase measurement capabilities, and the imaging contrast was far less than those fixed on the optical platform. Quantitative DPC (qDPC) phase reconstruction solves the quantitative measurement problems with improved resolution using partially coherent illumination [14,15]. Various QPMs based on qDPC were presented, including aberration correction qDPC [16], motion-resolved qDPC [17], optimal-illumination-designed qDPC [18–20], and polarization-resolved qDPC [21,22]. Nevertheless, most of them were modified by large-scale commercial microscopes, which are bulky, high-cost, and still fixed on the optical platform. It may limit the application of qDPC imaging for POCT.
Since the cDPC strategy contributes to single-shot qDPC imaging [23,24], it is possible to provide plug-and-play QPI for POCT [25]. In this work, we proposed a low-cost qDPC microscope with multimode imaging, including bright field (BF), dark field (DF), oblique lighting (OL), Rheinberg illumination, DPC, and qDPC imaging. The phase reconstruction algorithms [26,27] deployed on the graphics processing unit (GPU) of a laptop were used for phase denoising, maintaining fidelity, and real-time QPI. With a compact size of 140 × 165 × 250 mm3 and working robustness, our system could provide a plug-and-play POCT solution for biomedical applications with good contrast, high fidelity and resolution, including multimode imaging, quantitative measurement of cell phases, long-term observation of living cells in incubator, and real-time detection with QPI. It is convenient and suitable for POCT and medical treatment in remote areas since even a rookie can use a laptop and the QPM to meet the biomedical requirements at the scene.
2. Methods and materials
2.1 Design of the self-developed QPM
Our portable self-developed QPM is designed to compress the imaging path and dramatically reduce the overall volume of the system, as shown in Figure 1(a). With a volume of 140 × 165 × 250 mm3, our setup is portable to work in different scenarios. According to modular design, the setup is divided into four parts: a programmable LED illumination module, a two-dimensional (2D) electronic control module, axial focus module, and an ordinary imaging module.
Fig. 1. The self-developed QPM. (a) Overview and prototype design of the system. (b) Schematics and corresponding images of different imaging modes, including BF, OL, and DF. The scale bar is 20 µm. (c) GPU deployment of L0-qDPC and DSP-qDPC algorithms.
The programmable LED illumination module contains a programmable LED array as the light source with a plug-in design. The array with 93 LEDs (WS2812B, China) is controlled by the micro control unit (MCU) (Arduino UNO, USA) from a personal computer (PC). The LED array is fixed on the top of the setup as the light source without any other optical revision, as seen in the dotted inset of Figure 1(a). It just provides the downward incident light. Each LED can realize the control of 0-255 levels of the RGB channels individually (2563 combinations of color in total). The wavelength is 620-630 nm for the red channel, 515-530 nm for the green channel, and 465-475 nm for the blue channel, respectively. It is convenient for users to set the illumination patterns according to the control signals triggering MCU. The three-dimensional (3D) displacement between LED and rack can be adjusted with a range of ±5 mm for the alignment between light source and the principal axis of the setup. Besides, it also has a locking function after secondary calibration of the optical axis.
The electronic control platform module is designed as a three-layer structure. The system shell is connected by the bottom layer determining the direction. The medium layer is an X-axis control platform, which can drive Y-axis control platform (the top layer) to reduce the friction between the two axes. The core components are a guide rail and a screw motor (20H33, China), in which the guide rail connects the upper and lower layers and plays the role of fixed direction combined with the slot. The screw motor provides transmission so the platform can complete automatic 2D control.
The axial focus module employs a one-dimensional high-precision displacement control platform (XM50H-25-PU, China) combined with an eccentric objective bracket. The center of the objective is matched with that of the achromatic tube lens. The objective can be positioned via control signals.
The imaging module is implemented as follows. The illumination light through the sample is collected by an infinity-corrected objective, then passes through an achromatic lens (GCL-010656, Daheng, China) with a focal length of 175 mm, and is reflected twice by silver-plated mirrors M1 and M2. Finally, the sample is imaged on a CMOS camera (MV-SUA133GC-T, MindVision, China) connected to the PC. All used objectives include 10× magnification with NA of 0.3 (UPLFLN10X2, Olympus, Japan), 20× magnification with NA of 0.5 (UPLFLN20X, Olympus, Japan), and 40× magnification with NA of 0.75 (UPLFLN40X, Olympus, Japan). The calculation device is a high-performance laptop (CPU of AMD R7 5800 H and GPU of RTX3070 with 32GB RAM). More details about the prototype and assembly are shown in Visualization 1. For more technical details and performance, please see Supplement 1.
2.2 Illumination modes
Figure 1(b) depicts the optical layout for multimode imaging. For each layer of circular LED, the illumination numerical aperture (NA) has the relationship of
where R is the radius of the LED and D is the distance between the light source and the sample plane. In our system, both D and R can be adjusted accordingly.BF imaging is the most common method in microscopy [28], implemented by uniform illumination at symmetrical angles. When the inner three layers of the LED array are lightened, NAillu < NAobj, both background illumination light and forward scattering light from the sample can be collected to the objective lens. In this case, the image background is full of background illumination components.
In OL mode, the sample is illuminated by oblique incident light (assumed by plane wave). Only the third layer of the LED array needs to be lightened to obtain the optimal imaging contrast, meeting NAillu = NAobj[14]. The background components still exist, and moreover, due to the frequency filtering of the objective lens, the image of the sample has a ‘relief’ effect.
When the fourth layer of the LED array is lightened, NAillu > NAobj, our system is switched to DF mode, where the background illumination cannot be collected by the objective lens, and only forward scattering light, carrying the high-frequency information of the sample, is collected by the objective lens.
Rheinberg illumination mode is implemented by the combination of different lighting patterns (including position and wavelength). Taking the illumination mode in this manuscript as an example, the improved DF imaging, rainbow-DF mode, is implemented by lightening the third layer LEDs with different colors, including red (255, 0, 0), orange (255, 165, 0), yellow (255, 255, 0), green (0, 255, 0), cyan (0, 255, 255), blue (0, 0, 255), and purple (139, 0, 255). The LEDs in several subregions with different colors are lightened at the same time. The illumination NA requirements can be referred to the DF mode.
2.3 Algorithm deployment
The DPC image is differentiated by two OL images
where I1 and I2 are the asymmetrical OL images along the axis. Based on weak object approximation, the phase of target can be determined by a deconvolution of the DPC images and the phase transfer function [14]Since cell samples and quantitative phase target (QPT) are usually sparse, we used L0-qDPC algorithm to implement a non-tuning and high-contrast phase imaging for sparse samples [26]. Setting the penalty term $\psi$(φ) to L0-norm, the loss function is
The half-quadratic splitting (HQS) method [29] is used to solve L0-norm problem, a nondeterministic polynomial hard (NP-hard) problem. Moreover, we prove that the DPC images have dark-field sparse prior (DSP) based on simulations and mathematical derivations [27]. To achieve better qDPC phase reconstruction based on DSP characteristics, the L0-norm is imposed on the forward model of DPC to improve sparsity and denoising, and the L1-norm is introduced on the phase gradient to smooth the high-frequency information. The loss function is written as
To improve the temporal resolution of qDPC imaging, we rewrite the L0-qDPC and DSP-qDPC algorithms based on GPU parallel computing acceleration. First, the data-independent parameter initialization and preprocessing deployment are completed on GPU, including spatial sampling, settings of pupil and illumination, and the PTF models, as shown in the gray block in Figure 1(c). Then the LED array controlled by MCU generates the required illumination patterns, including four orthogonal asymmetric OL images and three RGB equipartition OL images.
Next, those images are transmitted to GPU and processed by 2D fast Fourier transform (2D-FFT) and the pre-deployed PTF model (L0-qDPC or DSP-qDPC). After a 2D inverse fast Fourier transform (2D-IFFT), the QPI results can be obtained. Finally, the phase image is retransmitted from GPU to the central processing unit (CPU) for preservation and other operations. As for time cost, the CPU-based and GPU-based algorithms are compared in Supplement 1, S3. Owing to the better imaging contrast and high temporal resolution, we chose the L0-cDPC algorithm based on GPU as the real-time QPI.
2.4 Sample preparation
A quantitative phase target (QPT, Benchmark, USA) [26] was used for phase imaging evaluation, including a focus star and a USAF with a thickness of 150 nm and 50 nm. A micrometer (JNOEC, China) with a range of 1 mm and an accuracy of 10 µm was applied for displacement measurement.
A 20 mm confocal dish (NEST, China, 1.0 × 105 cells per dish) was treated with 100 µl Poly-D-lysine (PDL) (100 µg/ml), put in the incubator for 2 to 3 hours, and dried on a super-clean bench. Then, L929 cells, mouse fibroblasts, were suspended in 2 ml Dulbecco's modified Eagle medium (DMEM, Biosharp, China) containing 10% FBS and 1% Penicillin-Streptomycin solution, inoculated into the treated confocal dish, and cultured at 37 °C and 5% CO2.
Human semen samples were donated by individual patients. The studies involving human participants were reviewed and approved by the Medical Ethics Committee of Anhui Medical University. The sperm gradient separation solution (10138 and 10139, SpermGrad, Vitrolife, Sweden) and liquefied semen were added to the centrifuge tube. After 20 minutes of centrifugation at 600 rpm, the bottom layer contained viable sperm. 50 µl of semen sample was injected into a centrifuge tube and was mixed evenly with 50 µl medium for sperm preparation (SpermRinse, Vitrolife, Sweden). Finally, the solution was flowed into a counting chamber with 10 µm (ML-CASA, Mai Lang, China).
1% 1-phenyl-2-thiourea (PTU, Shanghai Feixi Biotechnology, China) was added to the zebrafish embryo (Shanghai Feixi Biotechnology, China) to inhibit melanin growth, and the 5 ml culture medium was added. Then, the juvenile fish were observed after being cultured at 28 °C for 3 days. The blood flow velocity of juvenile zebrafish was reduced at 4 °C, and a confocal plate with a glass bottom diameter of 20 mm was added to perform the phase imaging.
3. Results
3.1 Multimode imaging of the biological samples
The multimode images of different unlabeled samples are presented in Figure 2. Under the imaging with the 20× objective, each mode corresponds to the illumination pattern at the bottom right. Figure 2 (a) are the ordinary BF images. The DF images in Figure 2 (b) present the cell outlines. Figure 2 (c) and (d) are the DPC images by the difference of the two asymmetrical OL images along the orthogonal directions. The sample morphology is shown in detail with the ‘relief’ effect, while there is different information between the two orthogonal phase gradients. Figure 2(e) are the improved rainbow-DF images with the rainbow-ring pattern, where thick cells and cell edges are stained with RGB. Figures 2 (f) and (g) are the optical stained images with Rheinberg illumination, where the thickness information can be denoted by the shade of color. The thicker the cells, the brighter the color. Since thickness corresponds to the optical path difference, the topography map can be demonstrated by the QPI method, as shown in Figure 2 (h).
Fig. 2. Multimode imaging of different biological samples, (1) are diatoms and (2) are red blood cells. (a) BF images. (b) DF images. (c) DPC images along the vertical axis of symmetry. (d) DPC images along the horizontal axis of symmetry. (e) Rainbow-DF images. (f-g) Rheinberg images with the corresponding illumination patterns at the bottom right. (h) Phase images. The scale bar is 20 µm.
The adherent cells, weak absorption objects, were also imaged. Figure 3 presents the multimode imaging of L929 and oral epithelial cells with flat distributions and transparent boundaries. In BF mode, the edges of two kinds of cells are observed. As for the nucleuses, L929 cells are hard to distinguished in Figure 3 (a1) due to the weaker optical absorption, and oral epithelial cells in Figure 3 (a2) can be observed but not clear enough. Moreover, those membranes with little intracellular substance are also fuzzy, while it is more prominent in DF imaging (Figure 3 (b)). According to the change of the phase gradient of oblique lighting [30], the DPC images in Figure 3 (c) and (d) of such samples show the ‘relief’ effect like DIC [31]. The results of Rheinberg modes in Figure 3 (e) - (g) do not reveal the fine structures of adherent cells since the samples are flat with tiny structures. For phase imaging as shown in Figure 3 (h1) and (h2), not only the cell edge and nucleuses but also tiny structures have been distinguished clearly with high contrast.
Fig. 3. Multimode imaging of different biological samples, (1) L929 cells and (2) are oral epithelial cells. (a) BF images. (b) DF images. (c) DPC images along the vertical axis of symmetry. (d) DPC images along the horizontal axis of symmetry. (e) Rainbow-DF images. (f-g) Rheinberg images with the corresponding illumination patterns at the bottom right. (h) Phase images. The scale bar is 20 µm.
3.2 Phase imaging and quantitative measurement
According to measurement of QPT with 10× objective in Supplement 1, our QPM can work with high phase fidelity and resolution. The reconstructed phase of the QPT is in line with the ground truth, while the phase imaging also shows good contrast.
We further presented the quantitative phase measurement of the human sperms with a 40× objective, including thickness and dry mass distribution. Figure 4 (a) and (b) show the BF and DF imaging, respectively. A round cell may be detached from the samples of patients with epididymal inflammation.
Fig. 4. Imaging and measurement of human sperms. (a) BF image. (b) DF image. (c) Phase image. The scale bar is 20 µm. (d) Topography map of the sperm. (e) Dry mass map of the sperm.
In the phase image of Figure 4 (c), the sperm and the round cell are seen clearly with high contrast. Furthermore, we measured the topography and dry mass map of the sperm in the yellow box according to the reconstructed phase, as seen in Figure 4 (d) and (e). The thickness of the sperm is around 1.03 µm, and the dry mass distribution is 0.93 pg/µm2, similar to the existing report [32].
Thus, our system can provide multi-contrast unlabeled imaging for sample observation using those objectives with different magnifications. Moreover, the setup can also measure the phase, thickness, optical volume, and other physiological indices for further studies.
3.3 QPI of the living cells
In this section, we presented the ability to observing living biological cells over a long period of time with the 20× objective. The setup worked in the incubator at 37°C and 5% CO2. After the automatic timing acquisition was set up by the system control platform, the one-click DSP-qDPC reconstruction was performed every 7.5 minutes, and 120 phase images were collected (More details can be seen in Visualization 2).
Figure 5 shows the time-lapse phase images of several regions of interest in Visualization 2, where L929 cells have apparent changes. Cells A and B underwent cytokinesis successively from the insets with the yellow box. Specifically, cell A was divided into two cells, a1 and a2, and a1 took the migration. In contrast, there was little change in the morphology of cell C, but its nuclear substances aggregated. Figure 5 (b1-b8) and (c1-c8) also present the cytokinesis process. Notably, when cell migration causes, there is an alternating process of extension of the pseudopod in the cell head, establishment of new adhesions, and contraction of the cell body tail in space and time, and some proteins (as pointed out with the white box) will be gathered near the cells to promote cell migration and proliferation. The results show that our self-developed microscope has good stability and robustness in QPI, which can be applied to narrow working environments such as incubators and can provide POCT solutions for biomedical research.
Fig. 5. Time-lapse phase images of the L929 cells. (a) First frame with FOV of of 204.8 × 204.8 µm (unit: rad). The insets with the yellow box are the zoom-in time-lapse phase images in (a), which are at 0, 412.5, 420, 427.5, 472.5, 495, 652.5, 900 min, respectively. (b1-b8) Zoom-in time-lapse phase images corresponding to the blue box in (a), which are at 0, 300, 315, 330, 352.5, 375, 382.5, 870 min, respectively. (c1-c8) Zoom-in time-lapse phase images corresponding to the red box in (a), which are at 0, 510, 517.5, 540, 570, 607.5, 682.5, 900 min, respectively. All scale bars are 10 µm.
3.4 Real-time QPI and in-vivo applications
After comparing the time cost of different DPC-based algorithms accelerated by different computing devices and demonstrating the real-time QPI potential on moving objects (more details in Supplement 1, S3), we further presented the real-time QPI of the tail vessels of zebrafish in vivo with the 10× objective.
Since the living sample could move and leave the field of view, the zebrafish was cooled to reduce the motor ability. The real-time transport of red blood cells is presented in Visualization 7, Visualization 8. Notably, the comparisons in Supplement 1, S3 show the time cost of full FOV (with 409.6 * 409.6 µm) phase reconstruction. To improve the frame rate, we chose the region of interest with 192 * 192 µm. Thus, the two videos are captured at 90 fps, Visualization 7 is the result of cooling for 5 min, while Visualization 8 is that of cooling for 15 min. Although the irregular motion of the sample at the initial stage of cooling leads to a slight wobble in the field of view, this does not affect the final imaging and measurement results. Figure 6 shows the direction of blood flow in the zebrafish tail.
Those results prove that our system can realize real-time QPI with good contrast for moving targets and has potential in POCT of high motility samples, including sperm detection, cytometry, and other in-vivo researches.
4. Discussion
For the POCT solution, we proposed a plug-and-play DPC-based QPM. In Supplement 1, we gave the prototype and assembly of the QPM in Visualization 1, and presented the performance on reconstructed phase fidelity, imaging contrast and resolution, and working robustness at different scenarios and temperatures. First, we tested the phase fidelity. According to the visualization based on the phase profile in Figure S1 (Supplement 1) and the quantitative rSNR score of 12.18, our QPM can realize quantitative phase imaging and measurement with good phase fidelity of biological samples.
The phase imaging in Fig. S2 demonstrates that the lateral resolution of 1.10 µm is close to the incoherent limit with the 10× magnification (NA of 0.3). The setup can provide high-contrast multimode imaging with more sample details to meet the different biomedical needs. Furthermore, the QPM provided the axial resolution (phase sensitivity) of 50 nm to achieve imaging and quantitative measurement of thin samples in Figure S3. Whether on an optical platform or an ordinary desk, the phase of a focus star with 50 nm can be measured.
For different environmental temperature, our microscope works with low drift and sufficient mechanical stability, as shown in Figure S4. Its lateral displacement control is 3.16 ± 0.45 µm and the axial control is 0.91 ± 0.24 µm. Furthermore, we demonstrated the drift of electromechanical control at different temperatures. Although the start-up and operation of the motor are affected at a low temperature (10 °C), there is little difference in displacement precision and return error at room temperature (20-30 °C). In addition, the working robustness and the ease of use are verified by a typical comparison of axial resolution on-desk measurements between the untrained and trained operators in Figure S5. We also presented the anti-vibration ability of the setup by the real-time QPI of the blood flow of the zebrafish tail in vivo in Visualization 7, as the motion of the living zebrafish could be regarded as sudden vibrations.
The visualizations of different phase reconstruction algorithms deployed on CPU or GPU are shown in Visualization 3, Visualization 4, Visualization 5, Visualization 6. In addition, the time costs are concluded in Supplement 1, S3. As a result, the L0-cDPC algorithm based on GPU acceleration is chosen as the real-time QPI scheme.
Since the above results prove the robustness of device, the manuscript listed the specific applications of the QPM in biomedicine. Multimode with high resolution and automation will help users acquire different information about the samples easily (Section 3.1); good phase fidelity for more measurements of physiological indices (Section 3.2); mechanical robustness does good to work in different scenarios like in-incubator or on-desk (Section 3.3 and Visualization 2); moreover, high-speed QPI can capture the fast-moving objects for real-time detection (Section 3.4, Visualization 7, Visualization 8). Even a rookie can choose the region of interest by automatically moving FOV through the GUI software, and the multimode images, including one-click qDPC phase reconstruction, are captured in an unattended manner after setting a fixed time interval on an ordinary desk.
Something could be improved in this study. From those QPI results, tDPC imaging is superior to cDPC in imaging contrast, phase fidelity, and target information in our system. There are two main reasons. On the one hand, although the PTF models of the two methods are similar, cDPC uses three different wavelengths of light based on the color-multiplexing strategy to realize single-shot phase reconstruction. Different wavelengths will cause chromatic aberration, introducing nonzero amplitude term into the forward model of ideal DPC and resulting in much phase noise [33]. On the other hand, since the self-developed system is designed for a plug-and-play POCT solution, we omit the calibration step of the camera color response before QPI, which may also cause a gap between the two methods. Furthermore, our microscope has few lenses for aberration correction except an achromatic glued lens. The factors mentioned above may affect the performance of the microscope on real-time QPI. This limit will be broken in the future, benefitting from the deep learning method [19,20,33,34].
In the future, the device will continue to be optimized and iterated with continuous theorem [35] and solver [36] improvements in the direction of integration [37], intelligence [38,39], and simplification [40].
5. Conclusion
In this paper, we presented a low-cost plug-and-play qDPC imaging system. Our system has portability with a compact size and can work in mobile medicine, cell incubator, and other scenarios. Axial focusing and FOV moving can be realized by the 3D electronic control platform without manual intervention. Through high-quality QPI, the setup can capture the sample with high imaging contrast and measure other physiological indices according to the quantitative phase information.
Compared to the existing QPMs modified from commercial microscopes, our device costs less than one-tenth (If the prototype is put into mass production, its cost will be further reduced). We hope it will contribute to POCT solutions for biomedical applications.
Funding
Key Research and Development Program of Anhui Province (2022a05020028); Natural Science Foundation of Anhui Province (2208085MC54); Research Fund of Anhui Institute of Translational Medicine (2021zhyx-B16); Key Scientific Research Foundation of Education Department of Anhui Province (2022AH050676, 2023AH040083); Open Research Topics of Anhui Provincial Engineering Technology Research Center for Biomedical Optical Instrument (2023BMP02).
Acknowledgments
Thanks for the donation of human semen samples used for experiments from Dr. Jun Li, Reproductive Medicine Center, Hefei BOE Hospital.
Disclosures
The authors declare no conflicts of interest.
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.
Supplemental document
See Supplement 1 for supporting content.
Reference
1. P. B. Luppa, C. Müller, A. Schlichtiger, et al., “Point-of-care testing (POCT): Current techniques and future perspectives,” TRAC 30(6), 887–898 (2011). [CrossRef]
2. P. Datta, “Methodological issues in point of care testing devices,” in Accurate Results in the Clinical Laboratory (Second Edition), A. Dasgupta and J. L. Sepulveda, eds. (Elsevier, 2019), Chapter 30, pp. 451–458.
3. C. Florkowski, A. Don-Wauchope, N. Gimenez, et al., “Point-of-care testing (POCT) and evidence-based laboratory medicine (EBLM) - does it leverage any advantage in clinical decision making?” Critical Reviews in Clinical Laboratory Sciences 54(7-8), 471–494 (2017). [CrossRef]
4. J. G. Toffaletti and C. R. Rackley, “Models for point-of-care testing of critical care analytes,” in Blood Gases and Critical Care Testing (Third Edition), J. G. Toffaletti and C. R. Rackley, eds. (Academic Press, 2022), Chapter 12, pp. 231–241.
5. T. Gao, Z. J. Smith, T. Y. Lin, et al., “Smart and fast blood counting of trace volumes of body fluids from various mammalian species using a compact, custom-built microscope cytometer,” Anal. Chem. 87(23), 11854–11862 (2015). [CrossRef]
6. Y. Wang, X. Zhang, J. Xu, et al., “The development of microscopic imaging technology and its application in micro- and nanotechnology,” Front. Chem. 10, 931169 (2022). [CrossRef]
7. Y. Park, C. Depeursinge, and G. Popescu, “Quantitative phase imaging in biomedicine,” Nature Photonics 12(10), 578–589 (2018). [CrossRef]
8. PhiOptics, “CellVista SLIM”, retrieved https://phioptics.com/cellvista-slim/.
9. PhiOptics, “CellVista GLIM”, retrieved https://phioptics.com/cellvista-glim/.
10. Y. Fan, J. Li, L. Lu, et al., “Smart computational light microscopes (SCLMs) of smart computational imaging laboratory (SCILab),” PhotoniX 2(1), 19 (2021). [CrossRef]
11. D. Jung, J.-H. Choi, S. Kim, et al., “Smartphone-based multi-contrast microscope using color-multiplexed illumination,” Scientific Reports 7(1), 7564 (2017). [CrossRef]
12. S. Kheireddine, Z. J. Smith, D. V. Nicolau, et al., “Simple adaptive mobile phone screen illumination for dual phone differential phase contrast (DPDPC) microscopy,” Biomed. Opt. Express 10(9), 4369–4380 (2019). [CrossRef]
13. Y. Bian, Y. Jiang, Y. Huang, et al., “Smart-phone phase contrast microscope with a singlet lens and deep learning,” Optics & Laser Technology 139, 106900 (2021). [CrossRef]
14. L. Tian and L. Waller, “Quantitative differential phase contrast imaging in an LED array microscope,” Opt. Express 23(9), 11394–11403 (2015). [CrossRef]
15. Y. Fan, Q. Chen, J. Sun, et al., “Review of the development of differential phase contrast microscopy,” Infrared and Laser Engineering 48, 603014 (2019). [CrossRef]
16. M. Chen, Z. F. Phillips, and L. Waller, “Quantitative differential phase contrast (DPC) microscopy with computational aberration correction,” Opt. Express 26(25), 32888–32899 (2018). [CrossRef]
17. M. Kellman, M. Chen, Z. F. Phillips, et al., “Motion-resolved quantitative phase imaging,” Biomed. Opt. Express 9(11), 5456–5466 (2018). [CrossRef]
18. Y. Fan, J. Sun, Q. Chen, et al., “Optimal illumination scheme for isotropic quantitative differential phase contrast microscopy,” Photon. Res. 7(8), 890–904 (2019). [CrossRef]
19. M. R. Kellman, E. Bostan, N. A. Repina, et al., “Physics-based learned design: optimized coded-illumination for quantitative phase imaging,” IEEE Trans. Comput. Imaging 5(3), 344–353 (2019). [CrossRef]
20. R. Cao, M. Kellman, D. Ren, et al., “Self-calibrated 3D differential phase contrast microscopy with optimized illumination,” Biomed. Opt. Express 13(3), 1671–1684 (2022). [CrossRef]
21. S. Hur, S. Song, S. Kim, et al., “Polarization-sensitive differential phase-contrast microscopy,” Opt. Lett. 46(2), 392–395 (2021). [CrossRef]
22. R. Kalita, W. Flanagan, J. Lightley, et al., “Single-shot phase contrast microscopy using polarisation-resolved differential phase contrast,” J Biophotonics 14(12), e202100144 (2021). [CrossRef]
23. Z. F. Phillips, M. Chen, and L. Waller, “Single-shot quantitative phase microscopy with color-multiplexed differential phase contrast (cDPC),” PLoS One 12(2), e0171228 (2017). [CrossRef]
24. W. Lee, D. Jung, S. Ryu, et al., “Single-exposure quantitative phase imaging in color-coded LED microscopy,” Opt. Express 25(7), 8398–8411 (2017). [CrossRef]
25. W. Lee, J. H. Choi, S. Ryu, et al., “Color-coded LED microscopy for quantitative phase imaging: Implementation and application to sperm motility analysis,” Methods 136, 66–74 (2018). [CrossRef]
26. T. Peng, Z. Ke, S. Zhang, et al., “Quantitative differential phase contrast phase reconstruction for sparse samples,” Optics and Lasers in Engineering 163, 107478 (2023). [CrossRef]
27. S. Zhang, T. Peng, Z. Ke, et al., “High-fidelity quantitative differential phase contrast deconvolution using dark-field sparse prior,” in Proc. SPIE, 2023), 127450X.
28. L. Xue, “Optical microscopy imaging and its application in bio-sample display and measurement,” Dissertation (Nanjing University of Science and Technology, Nanjing, Jiangsu Province, China, 2013).
29. Y. Wang, J. Yang, W. Yin, et al., “A new alternating minimization algorithm for total variation image reconstruction,” SIAM J. Imaging Sci. 1(3), 248–272 (2008). [CrossRef]
30. D. K. Hamilton and T. Wilson, “Two-dimensional phase imaging in the scanning optical microscope,” Appl. Opt. 23(2), 348–352 (1984). [CrossRef]
31. W. B. Amos, S. Reichelt, D. M. Cattermole, et al., “Re-evaluation of differential phase contrast (DPC) in a scanning laser microscope using a split detector as an alternative to differential interference contrast (DIC) optics,” Journal of Microscopy 210(2), 166–175 (2003). [CrossRef]
32. I. Barnea, L. Karako, S. K. Mirsky, et al., “Stain-free interferometric phase microscopy correlation with DNA fragmentation stain in human spermatozoa,” Journal of Biophotonics 11(11), e201800137 (2018). [CrossRef]
33. Y.-C. Lin, Y. Luo, Y.-J. Chen, et al., “Single-shot quantitative phase contrast imaging based on deep learning,” Biomed. Opt. Express 14(7), 3458–3468 (2023). [CrossRef]
34. A. C. Li, S. Vyas, Y. H. Lin, et al., “Patch-based U-Net model for isotropic quantitative differential phase contrast imaging,” IEEE Trans. Med. Imaging 40(11), 3229–3237 (2021). [CrossRef]
35. Y. Fan, J. Sun, Y. Shu, et al., “Accurate quantitative phase imaging by differential phase contrast with partially coherent illumination: beyond weak object approximation,” Photon. Res. 11(3), 442 (2023). [CrossRef]
36. S. Zhang, H. Wu, T. Peng, et al., “Pupil-driven quantitative differential phase contrast imaging,” arXiv, arXiv:2306.17088 (2023). [CrossRef]
37. Y. C. Yu, J. A. Yeh, and Y. Luo, “FPGA control module for quantitative differential phase contrast microscope,” in Frontiers in Optics + Laser Science 2021, Technical Digest Series (Optica Publishing Group, 2021), JW7A.115.
38. L. Kreiss, S. Jiang, X. Li, et al., “Digital staining in optical microscopy using deep learning - a review,” PhotoniX 4(1), 34 (2023). [CrossRef]
39. K. Wang, L. Song, C. Wang, et al., “On the use of deep learning for phase recovery,” Light: Science & Applications 13(1), 4 (2024). [CrossRef]
40. S. Daetwyler and R. P. Fiolka, “Light-sheets and smart microscopy, an exciting future is dawning,” Communications Biology 6(1), 502 (2023). [CrossRef]
