This study explores the fusion of a field-effect transistor (FET), a paper-based analytical cartridge, and the computational power of deep learning (DL) for quantitative biosensing via kinetic analyses. The FET sensors address the low sensitivity challenge observed in paper analytical devices, enabling electrical measurements with kinetic data. The paper-based cartridge eliminates the need for surface chemistry required in FET sensors, ensuring economical operation (cost < $0.15/test). The DL analysis mitigates chronic challenges of FET biosensors such as sample matrix interference, by leveraging kinetic data from target-specific bioreactions. In our proof-of-concept demonstration, our DL-based analyses showcased a coefficient of variation of <6.46% and a decent concentration measurement correlation with an r2 value of >0.976 for cholesterol testing when blindly compared to results obtained from a CLIA-certified clinical laboratory. These integrated technologies have the potential to advance FET-based biosensors, potentially transforming point-of-care diagnostics and at-home testing through enhanced accessibility, ease-of-use, and accuracy.
Keywords: FET biosensors; cholesterol; deep learning; dry chemistry; paper cartridge.