Online breath analysis using metal oxide semiconductor sensors (electronic nose) for diagnosis of lung cancer. (23rd October 2019)

• Record Type: Journal Article
• Authors: Kononov, Aleksandr; Korotetsky, Boris; Jahatspanian, Igor; Gubal, Anna; Vasiliev, Alexey; Arsenjev, Andrey; Nefedov, Andrey; Barchuk, Anton; Gorbunov, Ilya; Kozyrev, Kirill; Rassadina, Anna; Iakovleva, Evgenia; Sillanpää, Mika; Safaei, Zahra; Ivanenko, Natalya; Stolyarova, Nadezhda; Chuchina, Victoria; Ganeev, Alexandr
• Is Part Of: Journal of breath research. Volume 14:Number 1(2020:Mar.)
• Publication Date: 2019-10-23
• Subjects: breath analysis -- early diagnostics -- lung cancer -- electronic nose -- metal oxide sensors -- volatile organic compounds; Volatile organic compounds -- Analysis -- Periodicals; Clinical chemistry -- Periodicals; Bad breath -- Periodicals; Bad breath -- Treatment -- Periodicals; Bad breath -- Diagnosis -- Periodicals; 616.0756
• DOI: 10.1088/1752-7163/ab433d
• Languages: English
• ISSNs: 1752-7155

Abstract

Abstract: The analysis of exhaled breath is drawing a high degree of interest in the diagnostics of various diseases, including lung cancer. Electronic nose (E-nose) technology is one of the perspective approaches in the field due to its relative simplicity and cost efficiency. The use of an E-nose together with pattern recognition algorithms allow 'breath-prints' to be discriminated. The aim of this study was to develop an efficient online E-nose-based lung cancer diagnostic method via exhaled breath analysis with the use of some statistical classification methods. A developed multisensory system consisting of six metal oxide chemoresistance gas sensors was employed in three temperature regimes. This study involved 118 individuals: 65 in the lung cancer group (cytologically verified) and 53 in the healthy control group. The exhaled breath samples of the volunteers were analysed using the developed E-nose system. The dataset obtained, consisting of the sensor responses, was pre-processed and split into training (70%) and test (30%) subsets. The training data was used to fit the classification models; the test data was used for the estimation of prediction possibility. Logistic regression was found to be an adequate data-processing approach. The performance of the developed method was promising for the screening purposes (sensitivity—95.0%, specificity—100.0%, accuracy—97.2%). This shows the applicability of the gas-sensitive sensor array for the exhaled breath diagnostics. Metal oxide sensors are highly sensitive, low-cost and stable, and their poor sensitivity can be enhanced by integrating them with machine learning algorithms, as can be seen in this study. All experiments were carried out with the permission of the N.N. Petrov Research Institute of Oncology ethics committee no. 15/83 dated March 15, 2017.