Response Transient-Aware Artificial Olfaction for Nitrogenous Biomarkers Recognition and Asthma Diagnosis

Abstract

Noninvasive disease diagnosis through the exhaled nitric oxide (NO) biomarker is of great significance for the real-time screening of asthma patients. Herein, a high-performance artificial olfactory system (E-nose) based on porous In2O3 nanorods functionalized with Pt, Pd, and PtPd was designed and prepared for the precise recognition of the asthma biomarker NO and the noninvasive diagnosis of asthma. The sensor array demonstrates a limit of detection (LOD) of approximately 1.33 ppb (PtPd/In2O3) for NO and maintains high selectivity. To overcome the prolonged response/recovery times of room-temperature gas sensors and the indistinguishable response values for structurally similar molecules, a novel response transient-aware prototypical network model is proposed. By dynamically extracting transient features from early response stages, the model achieves a high classification accuracy of 94.44% for structurally similar nitrogen oxides (NO, NO2, and N2O). On a clinical dataset comprising 61 participants (42 healthy volunteers and 19 asthma patients), the E-nose system achieved a high accuracy of 93.75% based on the full time-series response stage. Notably, the accuracy of the early-response stage reached 92.85%, validating the feasibility of rapid diagnosis by using early-response segments. These results demonstrate that the artificial olfactory system based on metal-functionalized porous In2O3 nanorods has great potential in the precise and non-invasive diagnosis of asthma. It also offers useful guidance for rapid detection, even when sensors with slow response and recovery times are employed.