Poster V16 - A high-sensitivity vertical flow assay enhanced by deep learning for rapid troponin detection

Introduction: Cardiovascular diseases (CVDs), leading to high mortality worldwide, disproportionately affect low- and middle-income countries. As a major subtype of CVDs, acute myocardial infarction (AMI) demands rapid diagnosis due to its severe, life-threatening nature. Cardiac troponin I (cTnI) is essential for diagnosing AMI. Still, it is challenging to detect due to its low abundance in the bloodstream, especially in the early stages of cardiac muscle damage. High-sensitivity laboratory assays, while accurate, can be slow, expensive, and not universally accessible, making timely intervention difficult. Point-of-care testing (POCT) offers a promising alternative by providing rapid, affordable, on-site diagnostics. However, existing cTnI-POCT assays lack the necessary sensitivity and precision, failing to cover the clinically required dynamic range of concentration, which is crucial for timely diagnosis and intervention in AMI cases. This highlights the need for an advanced POCT for cTnI that is affordable, rapid, reliable, and highly sensitive. Here, we introduce a high-sensitivity vertical flow assay (hs-VFA) enhanced by deep learning to address these limitations, providing an accessible, rapid, and accurate method for high-sensitivity cTnI detection at the point of care.

Materials and

Methods: The hs-VFA system integrates a paper-based sensor with rapid nanoparticle amplification chemistry, a portable and cost-effective reader with time-lapse imaging capabilities, and a neural network-based computational algorithm (Figure 1a). Akin to the level of a rapid at-home test, our setup allows for the highly sensitive and accurate quantification of cTnI, using 50 µL of serum, delivering results within 15 min. The assay operation is streamlined into two steps in the VFA platform: the immunoassay and signal amplification. The signal amplification process was captured using time-lapse imaging, and computational algorithms (i.e., virtual control indicator and outlier analysis) were used to ensure the accuracy of the data. The assay's performance was validated with both cTnI-spiked serum samples and clinical serum samples obtained from UCLA Health. A neural network-based analysis was carried out to quantify the concentration of cTnI in clinical serum samples. The quantification results were compared to the ground truth values obtained from an FDA-cleared benchtop analyzer.

Results, Conclusions, and Discussions: The hs-VFA assay powered by a rapid nanoparticle amplification strategy achieved a detection limit of 0.2 pg/mL with a coefficient of variation (CV) of < 7%. It covers a wide detection range of over six orders of magnitude in cTnI concentration, up to 100 ng/mL, sufficient to meet the clinical needs for hs-cTnI testing. Integrating computational algorithms, virtual control indicator (VCI) and outlier analysis (OA), played a vital role in refining the assay data to improve systemic precision. The VCI served as an internal benchmark to validate each test run by comparing the signals from control spots against predefined thresholds. For example, the VCI identified and excluded four outlier samples from the clinical serum-related sample pool (N=62). The OA allowed for the exclusion of data points that deviated significantly from the statistical norm, improving assay uniformity and reproducibility. Using trained neural networks, the blinded test results showed accurate quantification of cTnI in clinical serum samples and demonstrated a high correlation (Pearson’s r = 0.965) with FDA-cleared analyzers, confirming the assay’s accuracy (Figure 1b and c). Our paper-based sensor costs $3.86 per test, and the mobile reader costs $170 per prototype, making them a cost-effective POC alternative to laboratory testing.

In summary, we developed a deep learning-enhanced paper-based hs-VFA that achieved significant advancements in POCT. With a low detection limit (0.2 pg/mL) and a high precision (an average CV of < 7%), our hs-VFA demonstrates the capability for highly sensitive and precise quantification of cTnI at point-of-care. This remarkable performance enhancement does not compromise the critical advantages of POCT, making it still affordable, simple, fast, and digitally connected. This innovation can transform diagnostic testing by making high-quality assays more accessible through cost-effective and easy-to-use POCT platforms, previously only available in well-equipped laboratories. This democratization of high-sensitivity diagnostic testing can help bridge the gap in healthcare accessibility, ensuring more equitable access to diagnostic testing, especially for challenging disease biomarkers. We envision this technology being widely used to test challenging disease biomarkers and expedite diagnostic equity globally.

Acknowledgements (Optional): The authors acknowledge the support of the US National Science Foundation (NSF) PATHS-UP Engineering Research Center (NSF #1648451). G.-R.H. acknowledges the Basic Science Research Program through the National Research Foundation of South Korea (NRF) funded by the Ministry of Education (NRF-2021R1A6A3A14039885).