Scholarlink

Background <p>Tobacco smoking is a leading cause for lung cancer, even for non-smokers. Assessing exposure, particularly to passive or secondhand smoke, is challenging due to reliance on self-reporting and the invasiveness of traditional biomarker tests. Exhaled breath analysis via Proton Transfer Reaction Time-of-Flight Mass Spectrometry (PTR-TOF-MS) provides a non-invasive approach for detecting relevant volatile organic compounds (VOCs). This study aimed to develop and validate a PTR-TOF-MS-based model to classify individuals according to smoking status and exposure level.</p> Methods <p>Exhaled breath from 3942 participants (active smokers, passive smokers, non-smokers) were collected by a standardized offline breath sampler and subsequently analyzed using high-resolution PTR-TOF-MS. Spectral m/z features were utilized to train, validate, and test seven machine learning models (6:2:2 data split). Performance was evaluated using accuracy, sensitivity, specificity, and Area Under the Curve (AUC), with optimized hyperparameters.</p> Results <p>Distinct VOC profiles were identified for active, passive, and non-smokers, indicating PTR-TOF-MS sensitivity to varying exposure levels. XGBoost model demonstrated superior performance, effectively differentiating active smokers from non-smokers with high test set accuracy of 92.08% (95% CI: 87.91%–95.17%), sensitivity of 90.83% (95% CI: 84.19%–95.33%), specificity of 93.33% (95% CI: 87.29%–97.08%), and AUC of 0.9678 (95% CI: 0.9438–0.9867). Key discriminatory mass spectral features (m/z), forming a distinct VOC spectral fingerprint associated with smoke exposure, were identified.</p> Conclusion <p>Exhaled breath analysis using PTR-TOF-MS combined with machine learning accurately differentiates active, passive, and non-smokers via distinct VOC signatures. This non-invasive method demonstrates sensitivity to different smoke exposure levels, including passive exposure, offering significant potential as an objective assessment tool for clinical research and public health monitoring.</p> Trial registration <p>Our study was registered with the Chinese Clinical Trial Registry on April 30,2025, registration number: ChiCTR2500101879.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Screening and diagnosis of combustion smoking with exhaled breath analysis via proton transfer reaction time-of-flight mass spectrometry

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月20日

查看全文
  • Chen Huang,
  • Jiayi Li,
  • RuiCen Li,
  • Jiajin Li,
  • Hanlu Yue,
  • Yanting Yang,
  • Yan Huang

摘要

Background

Tobacco smoking is a leading cause for lung cancer, even for non-smokers. Assessing exposure, particularly to passive or secondhand smoke, is challenging due to reliance on self-reporting and the invasiveness of traditional biomarker tests. Exhaled breath analysis via Proton Transfer Reaction Time-of-Flight Mass Spectrometry (PTR-TOF-MS) provides a non-invasive approach for detecting relevant volatile organic compounds (VOCs). This study aimed to develop and validate a PTR-TOF-MS-based model to classify individuals according to smoking status and exposure level.

Methods

Exhaled breath from 3942 participants (active smokers, passive smokers, non-smokers) were collected by a standardized offline breath sampler and subsequently analyzed using high-resolution PTR-TOF-MS. Spectral m/z features were utilized to train, validate, and test seven machine learning models (6:2:2 data split). Performance was evaluated using accuracy, sensitivity, specificity, and Area Under the Curve (AUC), with optimized hyperparameters.

Results

Distinct VOC profiles were identified for active, passive, and non-smokers, indicating PTR-TOF-MS sensitivity to varying exposure levels. XGBoost model demonstrated superior performance, effectively differentiating active smokers from non-smokers with high test set accuracy of 92.08% (95% CI: 87.91%–95.17%), sensitivity of 90.83% (95% CI: 84.19%–95.33%), specificity of 93.33% (95% CI: 87.29%–97.08%), and AUC of 0.9678 (95% CI: 0.9438–0.9867). Key discriminatory mass spectral features (m/z), forming a distinct VOC spectral fingerprint associated with smoke exposure, were identified.

Conclusion

Exhaled breath analysis using PTR-TOF-MS combined with machine learning accurately differentiates active, passive, and non-smokers via distinct VOC signatures. This non-invasive method demonstrates sensitivity to different smoke exposure levels, including passive exposure, offering significant potential as an objective assessment tool for clinical research and public health monitoring.

Trial registration

Our study was registered with the Chinese Clinical Trial Registry on April 30,2025, registration number: ChiCTR2500101879.