Significance of the Science

Heart disease remains the leading cause of death worldwide, and predicting cardiovascular risk often relies on measuring C-reactive protein (CRP), a blood marker of inflammation. Current high-sensitivity CRP (hsCRP) tests require expensive laboratory instruments, making them inaccessible in many settings. Paper-based tests are simpler and cheaper but typically lack the precision needed for reliable CRP measurement. This study bridges that gap by combining paper-based sensors with deep learning algorithms, creating a portable, low-cost system capable of high-quality hsCRP testing at the point of care.

Highlights

• Developed a paper-based vertical flow test and a smartphone-based reader for high-sensitivity CRP, a key heart health biomarker

• Combined the test with deep learning to interpret results more accurately

• Achieved 11.2% variation and 0.95 linearity in a blind test of 85 human samples, compared to results of FDA-approved assay

• Delivered results in under 12 minutes at a cost of less than $0.50 per test

Summary

This study presents a computationally enhanced paper-based vertical flow assay (VFA) for high-sensitivity C-reactive protein (hsCRP) testing, used to assess cardiovascular disease risk. By integrating deep learning with a multiplexed immunoreaction membrane, the system was trained to identify the most reliable sensing spots and accurately infer CRP concentrations. In a blind clinical study of 85 patient serum samples, the test achieved a coefficient of determination (R²) of 0.95 compared with an FDA-approved reference assay, with an average coefficient of variation of 11.2%. The method also successfully avoided false readings caused by the hook effect, extending the dynamic range of detection. With results delivered in under 12 minutes and a materials cost below $0.50, this approach demonstrates how machine learning and low-cost paper diagnostics can be combined to create portable, accurate, and affordable point-of-care tests for cardiovascular risk assessment.

Read more (open access): https://www.nature.com/articles/s41746-020-0274-y