An artificial intelligence-assisted, kilometer-scale wireless and wearable biochemical sensing platform for monitoring of key biomarkers in urine.

Abstract

Wearable biochemical sensors enabling non-invasive monitoring of biomarkers in bodily fluids play a pivotal role in advancing personalized healthcare. The state-of-the-art wireless and wearable biochemical sensors still suffer from large form factors, poor detection accuracy due to sample-to-sample variation, short and weak wireless communication, and difficulty to integrate with data processing algorithm on a system level. To solve these problems, this work develops an all-range wireless and wearable biochemical sensing platform which can be integrated in a diaper for monitoring four urine biomarkers (dimethylamine, creatinine, glucose, and H⁺) with two switchable wireless modes. To simplify the circuit design and reducing the form factor of the wearable sensing platform, this work develops flexible and passive potentiometric sensing interfaces for dimethylamine and creatinine detection by developing high-performance ion-selective electrode (ISE) with customized molecularly imprinted polymers (MIPs) as ionophores. The narrowband Internet of Things (NB-IoT) far-field wireless mode enables remote, and concurrent monitoring of urine biomarkers with a working range up to tens of kilometers, while the LC resonance near-field wireless mode is capable of battery-free and intermittent detection of urine biomarkers. The wearable sensor can be easily switched between the NB-IoT far-field wireless mode and the near-field wireless mode to fit different application scenarios. The wireless sensing platform enables system level integration of the wearable biochemical sensor with a multilayer perceptron data calibration system for data auto-calibration, which reduces the errors caused by varying pH and thus improves the detection accuracy, enabling deeper AI-wearable biochemical sensor fusion for next-generation healthcare applications.