Sniffing Disease: Chemiresistive Metal-Oxide Electronic Noses for Noninvasive Disease Diagnosis
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
This review examines recent advances in chemiresistive metal-oxide semiconductor (MOS) gas sensors fornoninvasive disease diagnosis through the detection of volatile organic compounds (VOCs) in exhaled breath and skin gas. Theoperating principle, based on oxygen chemisorption and desorption, is first outlined along with strategies for enhancing sensorperformance, including nanostructuring, catalytic functionalization, and heterojunction engineering, which enable detection limitssuitable for clinical analysis (sub-ppb range). To address limitations in sensitivity and selectivity, the integration of cross-reactivesensor arrays (electronic noses) with artificial intelligence–based pattern recognition, particularly deep neural networks (DNNs), ishighlighted as a critical approach for classifying complex VOC signatures. Clinical case studies across major diseases—such as lungcancer (aromatic and alkane markers), diabetes (acetone), asthma (H2S), COVID-19 (multi-array DNN systems), and tuberculosis(multiple VOCs)—demonstrate high diagnostic accuracy, validating the technology’s potential as a rapid and low-cost screening tool. However, successful clinical implementation requires overcoming key challenges, including the standardization of sampling and pretreatmentmethods (e.g., end-tidal breath collection and humidity control), cross-site data generalization, mitigation of confoundingvariables, and improvement of long-term sensor stability. Future research should focus on advanced material systems and robustmachine learning frameworks to realize universally applicable point-of-care diagnostic platforms.
Abstract
This review examines recent advances in chemiresistive metal-oxide semiconductor (MOS) gas sensors fornoninvasive disease diagnosis through the detection of volatile organic compounds (VOCs) in exhaled breath and skin gas. Theoperating principle, based on oxygen chemisorption and desorption, is first outlined along with strategies for enhancing sensorperformance, including nanostructuring, catalytic functionalization, and heterojunction engineering, which enable detection limitssuitable for clinical analysis (sub-ppb range). To address limitations in sensitivity and selectivity, the integration of cross-reactivesensor arrays (electronic noses) with artificial intelligence–based pattern recognition, particularly deep neural networks (DNNs), ishighlighted as a critical approach for classifying complex VOC signatures. Clinical case studies across major diseases—such as lungcancer (aromatic and alkane markers), diabetes (acetone), asthma (H2S), COVID-19 (multi-array DNN systems), and tuberculosis(multiple VOCs)—demonstrate high diagnostic accuracy, validating the technology’s potential as a rapid and low-cost screening tool. However, successful clinical implementation requires overcoming key challenges, including the standardization of sampling and pretreatmentmethods (e.g., end-tidal breath collection and humidity control), cross-site data generalization, mitigation of confoundingvariables, and improvement of long-term sensor stability. Future research should focus on advanced material systems and robustmachine learning frameworks to realize universally applicable point-of-care diagnostic platforms.