Portable, smartphone-linked, and miniaturized photonic resonator absorption microscope (PRAM Mini) for point-of-care diagnostics

Abstract

We report the design, development, and characterization of a miniaturized version of the photonic resonator absorption microscope (PRAM Mini), whose cost, size, and functionality are compatible with point-of-care (POC) diagnostic assay applications. Compared to previously reported versions of the PRAM instrument, the PRAM Mini components are integrated within an optical framework comprised of an acrylic breadboard and plastic alignment fixtures. The instrument incorporates a Raspberry Pi microprocessor and Bluetooth communication circuit board for wireless control and data connection to a linked smartphone. PRAM takes advantage of enhanced optical absorption of ∼80 nm diameter gold nanoparticles (AuNP) whose localized surface plasmon resonance overlaps with the ∼625 nm resonant reflection wavelength of a photonic crystal (PC) surface. When illuminated with wide-field low-intensity collimated light from a ∼617 nm wavelength red LED, each AuNP linked to the PC surface results in locally reduced reflection intensity, which is visualized by observing dark spots in the PC-reflected image with an inexpensive CMOS image sensor. Each AuNP in the image field of view can be easily counted with digital resolution. We report upon the selection of optical/electronic components, image processing algorithm, and contrast achieved for single AuNP detection. The instrument is operated via a wireless connection to a linked mobile device using a custom-developed software application that runs on an Android smartphone. As a representative POC application, we used the PRAM Mini as the detection instrument for an assay that measures the presence of antibodies against SARS-CoV-2 infection in cat serum samples, where each dark spot in the image represents a complex between one immobilized viral antigen, one antibody molecule, and one AuNP tag. With dimensions of 23 × 21 × 10 cm³, the PRAM Mini offers a compact detection instrument for POC diagnostics.

1. Introduction

Point-of-care testing (POCT) is emerging as a compelling alternative to laboratory-based in vitro diagnostic tests for a variety of diseases and states of health as a means toward reducing healthcare costs, reducing health disparities, and bringing sophisticated molecular diagnostics to parts of the world with limited resources. POCT is important not only for more effective management of human health but also for detecting the presence of animal disease in farm environments and veterinary clinics. POCT is also valuable in the context of environmental monitoring, where measuring the presence of pathogens and contaminants in water resources and indoor environments can assist in counteracting disease outbreaks [1]. The COVID-19 pandemic underscores the pivotal role of POCT home testing, highlighting the significance of rapid, accurate, and affordable testing accessibility [2]. POCT encompasses diverse technological frameworks, including fluorescence [3–5], colorimetry [6], electrochemistry [7], lateral flow assays [8], optical-based [9], photonic crystal-based [10], and smartphone-based platforms [11–13]. In recent reports, we described the development of a novel form of biosensor microscopy called Photonic Resonator Absorption Microscopy (PRAM), in which a photonic crystal (PC) surface serves to enhance the absorption contrast of gold nanoparticles (AuNP) tags that are used to label biological analytes such as proteins and microRNAs (miRNA) [14]. Unlike ordinary dark-field imaging or scattering microscopy, enhanced contrast for AuNPs is achieved in PRAM through optical interaction between the PC surface resonant reflection, the localized surface plasmon resonance (LSPR) absorption wavelength of the AuNP, and collimated illumination from a low-power LED that overlaps with the PC resonant wavelength. From a practical standpoint, the enhanced absorption in PRAM results in greater signal-to-noise contrast for easily observing and counting individual AuNPs linked to the PC surface, which provides robust detection with a simple and inexpensive instrument. Unlike fluorescent-based biodetection approaches, PRAM does not suffer from photobleaching or quenching, does not require a high-intensity laser, and does not require a sensitive, costly image sensor. Importantly, each AuNP in a PRAM image can represent the presence of only one detected molecule, thus enabling “digital” counting without sample partitioning (such as in droplet digital PCR) or enzymatic amplification. To utilize the digital counting capabilities of the PRAM instrument, we previously reported assay protocols for the detection of nucleic acid and protein analytes that linked AuNPs to the PC surface [14–19].

While the initially reported version of the PRAM instrument was implemented as a separate hyperspectral imaging line-scanning approach on a commercially available microscope [15,19] with a total system cost of ∼$200,000 USD, we demonstrated a more compact and inexpensive version of the instrument by eliminating the need for a spectrometer and precision line-scanning components by simply illuminating the PC with a large field of view while gathering reflected images with an image sensor [14,16–18]. This version of the PRAM could be considered “portable,” although it is assembled from expensive and heavy metal components such as an optical breadboard, precision optical brackets, and precision holding fixtures for stable alignment of components. This version of the instrument also incorporates computer-controlled translation stages and linear motors for functions that include auto-focus, image tiling, and sequenced detection from many separate sensors within a single microfluidic device (such as PC-integrated multi-well devices) [20].This portable PRAM instrument has a component cost of ∼$6,900 USD and may be most appropriate for laboratory environments where high throughput and automation are very important.

Here, we report the design, construction, and characterization of a PRAM instrument that is optimized for POC environments. The main design considerations are reduction of cost, complexity, size, and weight. Compared to the “portable PRAM” reported previously, the “PRAM Mini” in this work incorporates a sample holding stage for holding one PC biosensor in the detection/readout area to simply measure one PRAM field of view with an adjustable linear translation stage. The PRAM Mini operates on the same detection principle as previous versions of the PRAM instruments for observing enhanced absorption of AuNP tags used in biomolecular assays. Wherever possible, we utilize less expensive optical components for the microscope objective, LED, polarizing filters, and lenses. Also, the PRAM Mini replaces metal-based precise optical fixtures with tapped screw holes with an inexpensive acrylic sheet as the breadboard, with drilled holes for accurate placement of optical components that are held in place by plastic fixtures (made by 3D printing), fastened with adhesive. Despite the reduced cost, the PRAM Mini remains capable of imaging and counting AuNP tags with digital precision and acceptable optical contrast, and thus, the performance of assays (in terms of the detection limit for molecules) is not impacted. We envision that the PRAM Mini cost (particularly should it be transitioned to a mass-manufactured product) and size will be compatible with many POC applications in health clinics, farms, physician offices, veterinary facilities, food safety quality control, and environment monitoring scenarios.

In this paper, we briefly review the function of the PC biosensor and the operating principle of PRAM. Next, we describe the rationale for selecting the optical components used in the PRAM Mini for illumination, collimation, polarization, and sensing, as well as the electronic components used for managing the function of optical components and communicating data to a linked wireless device (smartphone). We describe the image processing workflow for converting raw images into AuNP counts and demonstrate the contrast obtained for individual AuNPs in PRAM images. We conclude by demonstrating a representative blocking biosensor assay for PRAM-based detection of antibodies against SARS-CoV-2 infection in cat serum samples.

2. Background

PC biosensors have been demonstrated as a highly versatile technology for a variety of label-free assays, including high-throughput screening of small molecule-protein interactions, characterization of protein-protein interactions, and measurement of cell attachment modulation by drugs [21–23]. A PC is a sub-wavelength grating structure consisting of a periodic arrangement of a low refractive index material coated with a high refractive index layer. When the PC is illuminated with a broadband light source, high-order diffraction modes couple light into and out of the high index layer, destructively interfering with the zeroth-order transmitted light [24]. At a particular resonant wavelength and incident angle, complete interference occurs, and no light is transmitted, resulting in nearly 100% reflection efficiency. The resonant reflectance magnitude is modulated by the addition of absorbing material, such as an AuNP, upon the PC surface, resulting in a large reduction in the reflected intensity. By measuring the resonant reflected intensity on a pixel-by-pixel basis across the PC, images of attached AuNPs may be gathered by illuminating the structure with collimated light that incorporates the PC-reflected wavelength through the transparent substrate, while the front surface of the PC is immersed in aqueous media. While the resonant wavelength of the PC is back-reflected, all other wavelengths propagate through the PC. Localized reduction in reflected intensity is caused by optical absorption of the resonant electromagnetic standing wave caused by the surface-captured AuNPs, using resonance-enhanced absorption as a contrast mechanism that is not possible in standard microscopies such as dark field imaging.

Our prior reports of PRAM described the operating principles for how AuNPs can be visualized with enhanced contrast when attached to a PC surface [19]. We showed that PC biosensor microscopy can be used to detect AuNPs that are substantially smaller than the pixel size and that metallic AuNPs produce highly localized effects upon the PC resonant reflection spectrum that enable individually attached particles to be easily observed for particles as small as 30 × 30 × 60 nm³. We observed that the optical absorption of NPs results in a highly localized reduction in the resonant reflectivity magnitude. Importantly, we demonstrated that the selection of AuNPs with extinction wavelength that matches the PC resonant wavelength results in a large signal-to-noise ratio in reflected images for the detection of individually adsorbed AuNPs, and thus AuNPs are a useful imaging contrast agent for visualizing the presence of biomolecules without background, photobleaching, blinking, or quenching. Leveraging this capability to count individual nanoparticles significantly lowers the detection threshold of our technique to a single-molecule detection level, showcasing its utility in detecting molecules such as miRNA [14–17,25] and antibodies [18].

The PCs (Moxtek, Orem, UT) utilized in this work utilize a linearly etched glass structure with a 400-nm grating period and a depth of 97 nm, coated with TiO₂ as a high refractive index thin film (n = 2.44, thickness = 98.5 nm). The design of this PC ensures an 85% efficiency of resonant reflection at 625 nm with a full width at half-maximum Δλ=4.34 nm. The PCs were diced into 10 × 12 mm² chips and attached to microscope slides using optical adhesive (NOA63, Edmund Optics) for further experiments.

3. Experimental

3.1. PRAM Mini point-of-care detection principle

The PRAM Mini detection principle was implemented using specific components chosen for their spectral characteristics: a photonic crystal (PC) with a resonant reflection at 625 nm, gold nanoparticles (AuNPs) with absorption centered at ∼620 nm, and an illumination source provided by a 617 nm fiber-coupled LED. The spectral overlap of these components is illustrated in Fig. 1(A). To showcase this detection principle, a blocking biosensor assay for detecting host antibodies against SARS-CoV-2 was conducted, as depicted in Fig. 1(B). Subsequent to the assay, the image was captured by PRAM Mini and processed an image processing algorithm (Fig. 1(C)), resulting in the visualization of AuNP counts (Fig. 1(D)). For any biomedical assays, we seek to avoid overcounting (false positives) while a bit of undercount is acceptable as long as a result is proportional to the concentration of the analytes. Hence, the threshold of the PRAM mini AuNP counting algorithm is tuned to be slightly preserved to avoid overcount and will end up with some missed AuNP spots (whose contrast is usually less than the counted AuNPs), as shown in Fig. 1(D). The algorithm worked based on the fact that the AuNPs’ attachment to the sensor surface would provide a measurable localized drop in reflected light intensity compared to the area where no AuNP is bound. Due to the size of the AuNP being much smaller than the resolution of the diffraction-limited imaging system, the sensor, in fact, captures the point spread function of the absorptive AuNPs, which are shown as dark circles. Based on this knowledge, the algorithm used shape, circularity, and area, with or without holes (Euler number), as the criteria to screen out the AuNPs from other non-AuNP features to provide accurate counting results.

Fig. 1.

The point-of-care principle of the PRAM Mini system involves several key steps: (A) A spectral overlap plot demonstrating the interaction between the fiber-coupled LED, PC, and AuNP within the PRAM Mini setup. (B) A blocking biosensor assay designed to detect antibodies against SARS-CoV-2. This assay utilized conjugated monoclonal antibodies (mAbs) tagged with AuNPs, which were specifically bound to the immobilized SARS-CoV-2 nucleocapsid (N) protein on the surface of the PC. (C) Subsequent to the assay, images of AuNPs bound to the PC surface were captured by the PRAM Mini and processed using an image processing algorithm. (D) The processed images revealed the presence of AuNPs as indicated by red dots, providing a clear representation of the results obtained through the image processing algorithm.

3.2. Selection of optical and electronic components

The PRAM Mini was designed to serve as a point-of-care detection device by miniaturizing its size and maintaining the ability to detect AuNPs on the PC. Figure 2(A) provides an overview of the optical setup of the PRAM Mini, showcasing its components and their functions. The PC utilized in our setup exhibits a resonant reflection wavelength of 625 nm. Consequently, we strategically chose an illumination wavelength that overlaps with the PC's resonance using a fiber-coupled 617 nm LED (M617F2, Thorlabs). The beam from this LED is collimated by a commercially available collimation package (F810SMA-635, Thorlabs). Utilizing well-collimated light for illumination reduces background noise from stray light and improves the optical alignment of the system. The collimated light is then directed by enhanced aluminum mirrors (#43-792, Edmund optics) towards a linear polarizer (XP42-200, Edmund optics) to ensure that only polarized light from the LED excites the TM mode of the PC. This polarized beam is then converged by a plano-convex lens acting as an L1 condenser (LA1251-N-BK7, Thorlabs) and focused onto the back focal plane of a 20x plan achromatic objective lens (MT02513431, BoliOptics). The PC sensor is securely placed on a custom-made aluminum sample holder, connected to a manually adjusted linear translation stage with 1 µm precision (450A, Newport) for z-axis focusing. Subsequently, the collimated light interacts with the PC, and the reflected light passes back through the objective lens to a 50:50 beam splitter (B08FBFMWFL, Jingliang, China).

Fig. 2.

(A) Optical design diagram of the miniaturized Photonic Resonator Absorption Microscopy (PRAM Mini) working principle where P: Polarizer, L1: Condenser, BS: Beam splitter, OBJ: Objective lens, PC: Photonic crystal, L2: Doublet lens, and CAM: Camera, (B) CAD diagram of the PRAM Mini construction with 3D-printed fixtures in a black acrylic enclosure, (C) Image of actual PRAM Mini instrument with the size of 23 × 21 × 10 cm³.

The reflected light from the PC is projected to form an image using an achromatic doublet lens that serves as an L2 (AC254-300-A, Thorlabs) to a 64-MP, 9152 × 6944 resolution CMOS camera (Pi Hawk-eye, Arducam, China). The camera is controlled by a single-board computer (Raspberry Pi module 4, Raspberry Pi, UK) with optimized conditions of 500 µs shutter speed, 1.5 contrast, 1.1,1 auto white-balance gain, and 1 gain, ensuring the acquisition of suitable images for subsequent analysis.

To streamline the integration of all components within the PRAM Mini enclosure and eliminate the need for expensive optical mounts, breadboards, and cage systems, we custom-designed (SolidWorks, Fig. S1) fixtures for each component and 3D printed them (Form 3B+, Formlabs) as depicted in Fig. 2(B). The enclosure of the PRAM Mini utilized lightweight black acrylic sheets (ACRY20250.187PM24X48, ePlastics). The sheets were precisely laser cut (Fusion Edge 12, Epilog) to match the dimensions of the customized 3D-printed optical fixtures, as illustrated in Fig. 2(C). The final dimensions of the PRAM Mini are 23 × 21 × 10 cm³ (Fig. S2). Notably, the total cost of the PRAM Mini is $2,100 USD, and the most expensive component is the 617 nm fiber-coupled LED and its driver, with a combined single unit price of ∼$804 USD. For convenience, we chose to use a high-quality z-linear translation stage with a micrometer resolution to focus the PRAM Mini images, with a component cost of $381 USD. Table S1 provides a detailed list of components, including part numbers, commercial sources, and prices. Our hypothesis is that a manufactured version of this instrument could be built at a cost of ∼$500 by sourcing components from original equipment manufacturers. In addition to the cost of the instrument, the consumables for each bioassay mainly consist of PCs and AuNPs. One 10 × 12 mm² PC (∼$25 USD, Moxtek, Orem, UT) can be used for up to 10 bioassays, resulting in a cost of approximately $2.50 USD per bioassay. The AuNPs are synthesized in-house, and based on our protocol, the cost is approximately $0.0004 USD per bioassay. Therefore, the total cost of consumables per bioassay is approximately $2.5004 USD, demonstrating the cost-effectiveness of this point-of-care diagnostic tool.

3.3. Image processing development

The image processing algorithm for PRAM Mini is informed by our previous work for automated portable PRAM [20]. For PRAM Mini, some adjustments were implemented to adapt it to the specific configurations, and the workflow of the image processing algorithm is shown in Fig. 3. The pixel size of the raw PRAM Mini image is 9152 × 6944 (approximately 30 times larger than reported in our previous work [14]. Because dark spots indicating attached AuNPs are approximately 100 pixels in diameter, numerous sources of noise (fine structure of pixel intensity induced by shot noise) will rise within the region of the spots, which hinders the accurate recognition of AuNPs. Therefore, we first reduce the pixel size of the image to 1788 × 1356 to smooth the PRAM image and facilitate further processing. Then, the resized image is filtered by the Tophat transform, which efficiently reduces the unevenness of the background. The next step is to apply a simple noise filter (Wiener filter) to minimize random noise from the former steps. The processed image demonstrates a sufficiently high signal-to-noise ratio and transforms the recognition of AuNPs into an easier binary thresholding task. However, because the point spread function of the AuNP is not a perfectly circular shape due to the unevenness of the evanescent field distributed on the PC surface, we apply a Maximally Stable Extremal Regions (MSER) algorithm to differentiate AuNPs from non-AuNP features by considering the shape, circularity, and area, with or without holes (Euler number) as the criteria. With fixed and finely tuned parameters for those criteria, we enumerated the attached AuNPs from the PRAM image, as shown in Fig. 5.

Fig. 3.

Workflow of the image processing for PRAM Mini. Left: 2D images. Right: 1D images of pixel grayscale along y direction at x = 90.2 µm. (A) Raw image (9152 × 6944) captured from PRAM Mini. The high resolution induces interrupting fine structure in the pixel intensity, as shown in the right figure. (B) Step 1: Resize the image size to 1788 × 1356 to smooth those fine structures within the AuNPs image region. (C) Step 2: Filtering the nonuniform illumination background using Tophat transformation. By applying a 9-pixel radius disk-shaped structural element, the nonuniformed ‘V’ shape background is perfectly removed, and the signal is more obvious to be recognized. (D) Step 3: Applying a Wiener filter to reduce random noise. We applied a 5 × 5-pixel square as the neighborhood to estimate the local image mean and standard deviation. The red-shaded bars indicate the position of AuNPs recognized by the algorithm. Scale bar: 25 µm.

Fig. 5.

Optical characterization of PRAM Mini (A) SEM image of AuUPs on the PC surface (Scale bar = 1 µm), (B) PRAM Mini image of AuNPs on the PC surface after image processing algorithm (Scale bar = 2 µm), (C) Intensity surface plot of the corresponding image showing the grey value of the AuNPs, (D) 1951 USAF Target image taken under PRAM Mini, showing elements in group 7 (Scale bar = 10 µm), (E) Zoom-in image of group 7, element 6 area with the line across the element, (F) Plot between grey value and distance (in pixels and µm) across the line from Fig. 4(E).

3.4. Smartphone application design and programming

The PRAM Mini is architecturally organized into two core systematic components (Fig. 4): firmware and software. The firmware framework (Fig. 4(A)) is responsible for programming the single-board computer (SBC) that operates within the PRAM Mini. This encompasses a suite of functionalities essential for system performance, executed through Python scripting. Key firmware functions include operating the high-resolution camera module, extracting and storing monochromatic images converted from RGB images for analysis, and managing Bluetooth connectivity. This connectivity is instrumental in interfacing with the customized app, allowing for data transfer, reception, and notification to peripheral Bluetooth devices. Open-source Python libraries, such as PyBluez for Bluetooth communication, PiCamera for camera operations, and OpenCV for image processing, were employed to enable these capabilities.

Fig. 4.

Block diagram of the PRAM Mini system and the application homepage. The system can be organized into two core blocks: (A) firmware and (B) software. Firmware handles functionalities essential for system performance, while software enhances the user interaction with the PRAM Mini system for the point-of-care application. (C) the primary interface of the PRAM Mini application.

Complementing the firmware, an Android App was developed to enhance user interaction with the PRAM Mini system for POC application. This software component (Fig. 4(B)) was designed with an emphasis on user experience, providing an intuitive interface for device operation and data handling. Android Studio was used in the development of the PRAM Mini application. Upon launch, the application directs users to its primary interface, where they can access the PRAM Mini control center and history of detected tests (Fig. 4(C)). The user interface has a swipeable function that makes it easy to move between areas without manually pushing buttons. Users must activate the Bluetooth feature on their smartphone to connect to the PRAM Mini device through the control center area. Upon successful connection, users gain access to the control panel, enabling them to issue commands to the PRAM Mini. The PRAM Mini starts detecting when it receives the “CAPTURE” instruction. It then takes a picture of the detection area on the PC, sends it back to the application for more analysis, and uploads the data to a cloud storage provider (Firebase, Google Cloud, US). Users can view a chronological record of their prior detections stored in the cloud by accessing the detection history section (Fig. S3). Users can access comprehensive information about the associated detection, including the collected image and analysis results, by choosing a specific date from the list.

3.5. Blocking biosensor assay preparation

The monoclonal antibody (mAb)-based blocking enzyme-linked immunosorbent assay (bELISA) is a technique for indirectly detecting molecules (e.g. host antibodies induced by a viral infection) based on antigen-mAb binding ability. It demonstrates a sensitivity similar to traditional indirect ELISAs (iELISA) while offering a notably increased level of specificity. Since the bELISA test readout is based on the mAb binding ability, unlike the iELISA, a species-specific secondary antibody is not required. Yuan et al. developed a bELISA method using a specific mAb targeting the SARS-CoV-2 nucleocapsid (N) protein, demonstrating exceptional specificity and repeatability [26]. Despite its improved specificity and capability of multi-species detection, bELISA still requires a laboratory setting. Innovative approaches are being explored to broaden its use to more POC scenarios. Recently, researchers have adapted the bELISA into the AuNP detection platform, creating a new blocking biosensor assay. This innovative approach combines the advantages of both methods, offering a fast and dependable serological test that would enable a POC diagnostic test to facilitate the rapid detection of SARS-CoV-2 exposure in diverse animal species and human.

Pre-treatment and silane functionalization

To begin the cleaning process, the PCs were sonicated in acetone, isopropyl alcohol (IPA), and Milli-Q water for 2 minutes, respectively. This is followed by drying with compressed nitrogen and a hotplate at 100 °C for at least 10 minutes to allow complete drying. The cleaning process is completed by performing oxygen plasma treatment (200 W) at 99% power for 10 minutes (Pico Plasma System, Diener Electronic). This creates a layer of hydroxyl groups across the titanium dioxide surface to bind to silane. For the silane functionalization process, the PCs were then immediately submerged in a solution containing 2.5 mL of 3-aminopropyltriethoxysilane (APTES) diluted in 47.5 mL of tetrahydrofuran (THF) on a shaker for 1 hour at room temperature. Following this, the PCs were sonicated in THF, acetone, ethanol, and Milli-Q water for 2 minutes each, then dried with compressed nitrogen. After completing this step, the PCs were stored in a desiccator for future experiments.

Monoclonal antibody and gold nanoparticle conjugation

Prior to conjugating mAb onto AuNPs, it is essential to attach the mAb to an NHS-PEG-SH linker. The NHS group on this linker will react with the amine group on the mAb and leave the thiol end ready to be covalently conjugated with the surface of AuNPs. To begin, mix 20 ug mAb with 1 mg/mL 10.5 uL NHS-PEG-SH linker solution, then shake it on a shaker for 2 hours at room temperature. This was followed by washing off unbound linkers using a 50 k centrifugal filter at centrifugation conditions of 14000 rcf, 4° C, 10 minutes three times.

The AuNP synthesis protocol was previously described [16,27]. To prepare the AuNPs for mAb conjugation, we wash the AuNPs with 0.1% Pluronic F-127 surfactant (Sigma-Aldrich Inc, P2443-250 G), centrifuging at 0.8 rcf for 7 minutes two times. Following the wash, a NanoDrop Microvolume Spectrophotometer (ThermoFisher, USA) is used to verify that the absorbance and optical density are approximately 620 nm and 1.2, respectively. The AuNPs are then mixed with the previously prepared mAb and left to incubate on a rotator at 4 °C overnight to enable mAb and AuNPs conjugation. After incubation, the mAb-AuNPs conjugates are verified by measuring their diameter using dynamic light scattering (Beckman Coulter) to be approximately 65 nm. Any unbound mAb is then washed away using 0.1x PBST (Thermo Scientific, 28352). To prevent non-specific binding, the mAb-AuNPs conjugates are incubated with 2% BSA (Sigma-Aldrich, 05470) for 30 minutes on a rotator. Finally, any unbound BSA is removed by washing with 0.1x PBST, completing the preparation of the AuNPs for subsequent assays.

SARS-CoV-2 N protein immobilization and detection of anti-N antibodies in cat serum

The capture molecule for this biosensor assay is SARS-CoV-2 nucleocapsid protein (N) to selectively bind with anti-N antibody. The PC surface is activated by drop-casting a 5 mg/mL solution of dimethyl sulfoxide (DSC, Sigma Aldrich) across the entire surface, allowing it to sit for 20 minutes before rinsing with water and drying with compressed nitrogen. This is followed by adhering silicone rubber gaskets made with polydimethylsiloxane (PDMS) onto the PC, then depositing 10 µL of N protein into each well and incubating overnight at 4° C.

After overnight incubation, the liquid was removed, and 1x PowerBlock (Thermo Fisher, part number needs to be checked) was added for 1 hour at 37 °C to block non-specific bindings. Next, the PC was washed three times with 1x PBST 3 times before adding cat serum samples; in which the negative sample was collected from a non-infected control cat, while the positive samples were collected from SARS-CoV-2 infected experimental cats at 14 days post-infection [26]. The negative sample was diluted 4 times, whereas the positive samples were serially diluted 4, 8, 16, and 32 times with 1x PowerBlock. The samples are then incubated for 1 hour at 37 °C on a shaker at 120 rpm to enable complete binding of antibodies in the sample to the N protein on the PC surface. After incubation, the sample solution was removed, and 10 µL mAb-AuNPs conjugates were added to the PC surface, and the PC was imaged after 15 minutes with the PRAM Mini. In each well, 3 different FOVs will be imaged and analyzed to obtain the average number of AuNP counts in the well.

According to the principle of the blocking biosensor assay, samples with higher concentrations of anti-SARS-CoV-2 antibody lead to a higher percentage of inhibition as it blocks the binding activity of the mAb-AuNP tag to the N protein, resulting in fewer AuNP counts. The assay’s results are quantified using the percentage of inhibition (PI) equation:

$$\text{Percentage}\text{of}\text{Inhibition}=\left(1-\frac{\text{Positive}\text{signal}}{\text{Negative}\text{signal}}\right)\times 100$$ (1)

where the positive signal is the AuNP counts from the positive sample (4- to 32-time dilutions), and the negative signal is the AuNP counts from the negative sample (4-time dilution). The detection of anti-SARS-CoV-2 antibody in cat serum was performed with three independent replicates (n = 3) to ensure accuracy and reproducibility.

4. Results and discussion

4.1. Device characterization

The optical characterization of PRAM Mini, aimed at demonstrating its ability as a point-of-care detection device, encompassed several key analyses. Figure 5(A) presents a scanning electron microscope (SEM) image illustrating the distribution of AuNPs on the PC surface. This SEM image depicts a PC grating period of approximately 400 nm and AuNPs with an 80 nm diameter. Utilizing the operational principles of PRAM, real-time imaging similar to SEM results was achieved without the complexities associated with sample spin coating or the high cost typically associated with SEM equipment. Figure 5(B) showcases a selected region from the PRAM Mini image post-image processing, revealing the presence of AuNPs on the PC surface. Notably, the apparent size of AuNPs in the PRAM Mini image appears larger than 80 nm, caused by the diffraction limit inherent in optical imaging techniques. Further analysis was conducted through intensity surface plotting (Fig. 5(C)), validating the precise localization of AuNPs on the PC surface as indicated by the previous PRAM Mini image. This analysis demonstrates that the image obtained from PRAM Mini shows distinguishable features of the AuNPs, highlighting the core capability of PRAM Mini.

To determine the field of view (FOV) of PRAM Mini, a rigorous assessment using a 1951 USAF resolution target (R3L1S4P, Thorlabs) was performed. Figure 5(D) displays the highest achievable and available resolution, group 7 and elements 1-6, quantified at 228.1 line pairs (lp)/mm or 4.4 µm/lp. This resolution was further validated by correlating pixel values to spatial dimensions. Comparing the spatial dimension of a drawn line across a group 7 element 6 object (Fig. 5(E)) to the spatial grey value plot (Fig. 5(F)) obtained from the image demonstrates that 37.27 pixels in the PRAM Mini image correspond to 1 µm. The determined FOV of the PRAM Mini image, encompassing 9152 × 6944 pixels, translates to a spatial area of 245 µm × 186 µm. This comprehensive analysis confirms the imaging capabilities of PRAM Mini and establishes its spatial resolution and field coverage, essential metrics for its utility in POC applications requiring precise and localized detection of analytes.

4.2. Image processing for PRAM Mini digital counting

PRAM images are processed using the counting algorithm mentioned above. Figure 6 shows the step-by-step outcomes of the original PRAM image as well as single nanoparticle’s image (Fig. 6(A)). We observed that the image is less noisy after resizing (Fig. 6(B)) and the signal-to-noise ratio of a single AuNP image is considerably enhanced after implementing the Tophat (Fig. 6(C)) and noise filtering (Fig. 6(D)). After applying MSER, the dark spots which satisfied all the criteria are tagged by a red mask (Fig. 6(E)) and finally AuNPs are represented by identical red circles at the location of their centroid (Fig. 6(F)), improving the visualization of the recognized AuNPs for non-signal processing personnel to measure the outcome.

Fig. 6.

Digital counting results for PRAM Mini. (A) Raw image (9152 × 6944) captured from PRAM Mini. (B) Resize the image size to 1788 × 1356 to avoid shot noise within the AuNPs image region. (C) Filtering the nonuniform illumination background using Tophat transformation. (D) Applying a Wiener filter to reduce random noise. We applied a 5 × 5-pixel square as the neighborhood to estimate the local image mean and standard deviation. (E) AuNPs were recognized using MSER by considering the shape, circularity, and area, with or without holes (Euler number) as the criteria: step size between intensity threshold levels: 0.1, maximal area variation between extremal regions: 0.5, circularity: > 0.5, area: < 250 and > 100 pixels, intensity threshold: < 0.95 (normalized), Euler number: > 0. (F) Identified AuNPs were represented by red circles with the same radius, and the count of the AuNPs was reported accordingly. Scale bar in the original-size image: 25 µm, scale bar in the zoomed-in image: 2.5 µm.

4.3. Detection assay for antibodies against SARS-CoV2

To demonstrate the ability of PRAM Mini to detect AuNPs, we performed a blocking biosensor assay for anti-SARS-CoV-2 antibody detection in cat serum samples. As previously discussed, the SARS-CoV-2 N protein was immobilized on the PC surface overnight to allow full coverage of the biorecognition element. Followed by the blocking step to inhibit any nonspecific binding on the PC surface. The 4-time dilution standard negative cat serum and 4- to 32-time dilution standard positive cat serum were added and incubated for one hour to promote the binding activity of the anti-SARS-CoV-2 antibody to the N protein. After the incubation and sample removal, mAb-AuNP tags were introduced into each well to allow the binding activity of the mAb to N protein and the conjugated AuNP tag, which will be detectable by PRAM. A 15-minute assay time was selected as the optimal imaging timeframe. Each well was imaged 3 times and analyzed by the counting algorithm to determine the number of AuNP counts. Figure 7(A) shows the matrix of representative images for anti-SARS-CoV-2 antibody detection ranging from 1:4 to 1:32 dilutions (n = 3) at 15 minutes, where the 1:4 dilution exhibits the lowest AuNP counts and the 1:32 dilution has the highest AuNP counts. The bar graph in Fig. 7(B) presented the PI value for each dilution (n = 3), detailing the average PI values and the standard deviations.

Fig. 7.

The detection of anti-SARS-CoV-2 antibodies in cat serum using PRAM Mini. (A) An FOV representative image of AuNP binding on the PC surface as a function of sample dilution (4- to 32-times dilutions) at 15 minutes after processed with counting algorithm, (B) A bar chart showing percent inhibition (PI) of each dilution in triplicate together with standard deviation. (Each bar represents a standard deviation of n = 3 independent PC samples)

The assay results exhibit a dose-response correlation, with the PI value decreasing as the serum samples become increasingly diluted. Notably, for the 1:32 dilution, both the average PI value and standard deviation are above zero, suggesting that the assay has the potential to detect even lower concentrations. This study demonstrated the proportional correlation between the number of AuNPs and the diluted SARS-CoV2 samples. In the future, another thorough study on the dose-response curve is to be investigated to provide a quantitative analysis of the target concentrations using PRAM technology. We also demonstrated the temporal detection of the blocking assay with our newly developed dynamic counting algorithm [20], where the kinetic binding and unbinding events, as well as the kinetic rate, were recorded, as shown in Supplementary Figures S4, S5, and Visualization 1 ^{(143.8MB, avi)} .

This proof-of-concept experiment is a pivotal demonstration of the operational feasibility and efficacy of PRAM Mini as a POC detection device for the determination of immune status through the detection of anti-SARS-CoV-2 antibody. The ability of the compact, portable, and miniaturized PRAM Mini to produce clear images during this assay demonstrates its effectiveness and practicality for POC applications. Its performance is comparable to bulkier, non-portable imaging systems while also being significantly lighter and having a lower cost. The PRAM Mini is a flexible and convenient tool suitable for not only detecting serological antibodies but also for a wide range of other biomolecules.

5. Conclusion

The development of the miniaturized Photonic Resonator Absorption Microscope (PRAM Mini) in this work represents a significant advancement in POC diagnostics, offering a compact, smartphone-linked instrument for rapid and accurate detection. The PRAM Mini, leveraging the optical coupling of PC surfaces and AuNP tags, has demonstrated its effectiveness in detecting anti-SARS-CoV-2 antibodies (from 1:4 to 1:32 dilution of standard cat serum), showcasing its potential for a wide range of diagnostic applications. The compact design, smartphone integration, and cost-effectiveness make it a promising tool for enhancing healthcare accessibility and patient care. One of the key strengths of the PRAM Mini is its ability to achieve digital resolution detection of biomolecular interactions, enabling precise quantification of target analytes. This digital resolution quantification feature sets it apart from traditional diagnostic methods, offering a more advanced and reliable diagnostic platform without microdroplet sample partitioning or enzymatic amplification. The demonstration of the PRAM Mini in detecting anti-SARS-CoV-2 antibodies with the simple blocking biosensor assay underscores its relevance in addressing pressing healthcare challenges, such as pandemics. Its integration into POC settings can streamline diagnostic workflows, reduce turnaround times, and improve healthcare system efficiency.

We envision exploring the potential of PRAM Mini in several aspects, including expanding the menu of detected analytes, detection of arrays of assays in the form of printed spots within the imaging field of view, and precise AuNP counting with machine learning image processing algorithms. The instrument can offer a sensitive, quantitative, simple, and inexpensive molecular diagnostic system for resource-limited settings, health clinics, and physician offices.

Funding

National Institute of Health (R01AI166791); National Science Foundation10.13039/100000001 (PFI-TT1919015).

Disclosures

B. Cunningham is a founder of Atzeyo Biosensors, which holds a license for the PRAM technology.

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 ^{(2.9MB, pdf)} for supporting content.