Detection of COPD and Lung Cancer with electronic nose using ensemble learning methods

Area of Science:

• Respiratory Medicine
• Biomedical Engineering
• Analytical Chemistry

Background:

• Electronic-nose (e-nose) devices utilize chemical gas sensor arrays and machine learning to analyze volatile organic compounds (VOCs) in exhaled breath.
• Detecting and differentiating respiratory ailments through breath analysis offers a non-invasive diagnostic approach.

Purpose of the Study:

• To develop and validate an e-nose system for distinguishing between chronic obstructive pulmonary disease (COPD), lung cancer, and healthy controls.
• To assess the efficacy of machine learning algorithms in classifying respiratory diseases based on VOC profiles.

Main Methods:

• A portable, cost-effective e-nose system with TGS gas sensors was developed.
• Exhaled breath samples were collected from 199 participants (93 controls, 55 COPD patients, 51 lung cancer patients).
• The XGBoost ensemble learning method was employed for data analysis and classification.

Main Results:

• The XGBoost model achieved 79.31% accuracy in predicting lung cancer and 76.67% accuracy in predicting COPD.
• The developed e-nose system demonstrated robustness and portability.
• Distinct VOC profiles were identified between patients with pulmonary diseases and healthy individuals.

Conclusions:

• The portable, low-cost e-nose system provides rapid response and shows potential as a diagnostic tool for lung diseases.
• Breath VOC analysis using e-nose technology can effectively differentiate pulmonary diseases from healthy controls.
• The findings support the use of e-nose technology for non-invasive diagnosis of respiratory conditions.