Early sepsis diagnosis using machine learning–augmented plasmon-enhanced biosensors

Abstract

Rapid and prompt diagnosis of sepsis is crucial in alleviating death and timely provision of therapeutic care. This work describes a multimodal biosensing solution based on the plasmon-enhanced biosensing system combined with machine learning (ML) algorithms to detect sepsis early biomarkers multiplexically and at the point of care. In this study, a hybrid surface plasmon resonance (SPR) and localized surface plasmon resonance (LSPR) biosensor was developed and built, which was bioreceptor-receptor engineered to detect a panel of biomarkers, such as procalcitonin, interleukin-6, lactate, and tumor necrosis factor-alpha. Bench top experiments, as well as simulated clinical data, indicate that the ML-augmented plasmonic sensor has high sensitivity and specificity allowing it to give early-warning capabilities that are better than those of conventional sensors. A microfluidic means of integration and edge computing to perform quick analysis also support point-of-care deployment within this integrated platform.