A Novel Hybrid Sensor Array with Enhanced Sensitivity and Selectivity for Biomarker Detection of Multiple Respiratory Diseases

Abstract

Electronic nose is an emergent technique for noninvasive disease detection via breath analysis, which is, however limited by the sensitivity and selectivity of the sensor array to complex biomarkers. In this study, we developed a hybrid sensor array to identify various biomarkers of different respiratory diseases, providing a rapid and convenient diagnostic method. Five on-chip microsensors were fabricated and combined with 11 commercial sensors to achieve enhanced sensitivity and complementary selectivity for multiple biomarkers. The array performances were significantly improved by globally optimizing the operating temperatures with machine learning, which enabled the precise identification of six key biomarkers (isoprene, n-propanol, toluene, acetaldehyde, acetone, and nitric oxide) from three respiratory diseases (lung cancer, asthma, and COVID-19). The classification mean accuracy for these biomarkers under concentration variations reached 98.9% with 5-fold cross-validation, and the R2 values exceeded 97% for concentration prediction. Furthermore, tests on simulated exhaled air validated our array’s effectiveness on disease diagnosis, achieving mean classification accuracy of 100% and 98.3% under dry and high-humidity conditions with 5-fold cross-validation, respectively. The exceptional capability of the hybrid sensor array in gas discrimination and disease-specific pattern recognition highlights its potential for exhaled breath tests in clinical and home healthcare.

Keywords: Hybrid gas sensor array, Array optimization, Enhanced sensitivity and selectivity, Biomarker detection, Respiratory disease diagnosis