Prediction of lung cancer with a sensor array based e-nose system using machine learning methods

• Published in: Microsystem technologies : sensors, actuators, systems integration actuators, systems integration, 2024-11, Vol.30 (11), p.1421-1434
• Main Authors: Binson, V. A., Subramoniam, M., Mathew, Luke
• Format: Article
• Language: English
• Subjects: Age groups; Algorithms; Artificial intelligence; Biomarkers; Breath tests; Chemical sensors; Chronic obstructive pulmonary disease; Classification; Cost analysis; Data analysis; Data processing; Discriminant analysis; Electronic noses; Electronics and Microelectronics; Engineering; Gas sensors; Health care; Instrumentation; Lung cancer; Lung diseases; Machine learning; Mechanical Engineering; Medical materials; Medical research; Metal oxides; Nanotechnology; Patients; Reproducibility; Sampling methods; Sensor arrays; Sensors; Surface acoustic waves; Technical Paper; Tobacco; VOCs; Volatile organic compounds
• ISSN: 0946-7076; 1432-1858
• DOI: 10.1007/s00542-024-05656-5

Lung cancer diagnosis with breath volatile organic compounds (VOC) analysis using electronic nose (e-nose) is an emerging area in the medical electronics field. Numerous chemical gas sensors were developed for the analysis of human breath VOCs. Even though the VOC gas sensors are developed with modern materials and techniques, as of now, these gas sensors are still not widely used in clinical applications because of their adverse performance in some cases. This paper discusses the design and development of an innovative artificial intelligence (AI) based e-nose system, which can detect lung cancer by detecting the volatile organic compounds in the exhaled human breath. We fabricated an e-nose system with five VOC gas sensors and tested the system with breath samples of 22 lung cancer patients and 40 healthy controls. This work, details the sensor selection process, fabrication of e-nose system, sampling methods, and sensor data analysis methods. Among the three classification algorithms used in the study, linear discriminant analysis have shown a better classification accuracy of 93.14% and AUC of 0.98. This algorithm provided a sensitivity and specificity of 88.63% and 95.62% respectively. Briefly, the sensor array system developed with TGS gas sensors was non-invasive, low cost, and gave a rapid response. Even though the attained results were good, further examinations are essential to enhance the sensor array system, investigate the long run reproducibility and repeatability, and enlarge its relevancy.