ABSTRACT

Introduction

This paper describes the research work done toward the development of a breath analyzing electronic nose (e-nose), and the results obtained from testing patients with lung cancer, patients with chronic obstructive pulmonary disease (COPD), and healthy controls. Pulmonary diseases like COPD and lung cancer are detected with MOS sensor array-based e-noses. The e-nose device with the sensor array, data acquisition system, and pattern recognition can detect the variations of volatile organic compounds (VOC) present in the expelled breath of patients and healthy controls.

Materials and methods

This work presents the e-nose equipment design, study subjects selection, breath sampling procedures, and various data analysis tools. The developed e-nose system is tested in 40 patients with lung cancer, 48 patients with COPD, and 90 healthy controls.

Results

In differentiating lung cancer and COPD from controls, support vector machine (SVM) with 3-fold cross-validation outperformed all other classifiers with an accuracy of 92.3% in cross-validation. In external validation, the same discrimination was achieved by k-nearest neighbors (k-NN) with 75.0% accuracy.

Conclusion

The reported results show that the VOC analysis with an e-nose system holds exceptional possibilities in noninvasive disease diagnosis applications.

中文翻译：

---

使用基于 MOS 传感器阵列的电子鼻通过呼吸分析无创检测 COPD 和肺癌

摘要

介绍

本文介绍了为开发呼吸分析电子鼻 (e-nose) 所做的研究工作，以及对肺癌患者、慢性阻塞性肺病 (COPD) 患者和健康对照组的测试结果。使用基于 MOS 传感器阵列的电子鼻检测 COPD 和肺癌等肺部疾病。带有传感器阵列、数据采集系统和模式识别的电子鼻设备可以检测患者和健康对照者排出的呼吸中存在的挥发性有机化合物 (VOC) 的变化。

材料和方法

这项工作介绍了电子鼻设备设计、研究对象选择、呼吸采样程序和各种数据分析工具。开发的电子鼻系统在 40 名肺癌患者、48 名慢性阻塞性肺病患者和 90 名健康对照者中进行了测试。

结果

在将肺癌和 COPD 与对照组区分开来时，具有 3 倍交叉验证的支持向量机 (SVM) 在交叉验证中的准确率为 92.3%，优于所有其他分类器。在外部验证中，k-最近邻（k-NN）以 75.0% 的准确率实现了相同的区分。

结论

报告的结果表明，使用电子鼻系统进行 VOC 分析在无创疾病诊断应用中具有特殊的可能性。