Significance of the Science

Accurate and frequent monitoring of biomarkers like cholesterol is essential for managing chronic health conditions such as cardiovascular disease. Current diagnostic methods rely on centralized laboratories or optical detection systems, which can be costly, complex, and not easily adaptable for point-of-care or at-home testing. Field-effect transistor (FET) biosensors offer an alternative, but challenges such as variability between tests and interference from biological samples have limited their broader use. 

In this study, we use cholesterol as a model biomarker to demonstrate how combining FET sensors with paper-based cartridges and the computational power of deep learning can overcome these challenges. By capturing the full kinetics of biochemical reactions and analyzing them with neural networks, the system provides accurate, low-cost cholesterol testing with the potential to expand into broader biomarker applications .

Highlights

• Introduces a deep learning–based approach to analyze kinetic signals from FET biosensors

• Integrates paper-based analytical cartridges with reusable FET sensors for cost efficiency

• Delivers test results in under 5 minutes at a cost of less than $0.15 per test

• Demonstrates high accuracy (r² > 0.976) and low variation (CV < 6.46%) in cholesterol detection compared with a CLIA-certified clinical laboratory

• Establishes a platform adaptable to immunoassays for detecting other biomarkers beyond cholesterol

Summary

This work presents the integration of field-effect transistor (FET) sensors, paper-based analytical cartridges, and deep learning (DL) analysis for quantitative biosensing. The approach uses a porous sensing membrane with dried enzymes to generate electro-active signals, which are captured by FETs and translated into kinetic data. Deep learning models then process these data to minimize sample interference and variability between cartridges.

In proof-of-concept studies with clinical plasma, the platform quantified cholesterol concentrations with strong agreement to results from a CLIA-certified laboratory (r² > 0.976), while maintaining a coefficient of variation below 6.46%. Each disposable cartridge costs less than $0.15 and delivers results in under 5 minutes. Beyond cholesterol, this DL-enhanced FET system shows potential for broader applications, including immunoassays and other biomarker tests, advancing the accessibility and precision of point-of-care and at-home diagnostics.

Read more: https://pubs.acs.org/doi/10.1021/acsnano.4c02897