A single gas sensor assisted by machine learning algorithms for breath-based detection of COPD: A pilot study

Abstract

Chronic obstructive pulmonary disease (COPD) is a widespread respiratory disorder with a high mortality rate. Current diagnostic methods have limitations, necessitating innovative diagnostic approaches. Breath analysis using electronic noses (e-noses) is promising but still faces challenges. The single temperature-modulated sensor (STMS) is one of the optimized models of e-noses. It uses a single sensor instead of an array of multiple sensors, which enhances commercial viability and ease of use. This pilot study explores the feasibility of using a STMS device for COPD detection. Breath samples were collected from 34 healthy individuals and 33 COPD patients. Features extracted from the sensor's response were analyzed, and a novel feature selection method was developed to discriminate between the two groups. Using this method and a linear support vector machine (SVM) classifier, a single commercial sensor, driven with half-sine-staircase and staircase heater waveforms, achieved 80.60 % accuracy, 78.79 % sensitivity, and 82.35 % specificity in differentiating between COPD patients and healthy controls. These results demonstrate the potential of an appropriately configured STMS to capture comprehensive breath information for COPD detection. Furthermore, our study highlights how advancements in machine learning can maximize the effectiveness of this approach. These promising pilot study results warrant further investigation with larger, more diverse cohorts to validate this approach for broader clinical implementation.