Biomedical Signal Processing and Control · Nov 1, 2026
Exhaled VOCs and AI for early lung cancer diagnosis: Advances and barriers
By Chen Z, Hua F, Xue Y
Lung cancer is the leading cause of cancer-related mortality worldwide, and early diagnosis is critical for improving prognosis. Exhaled volatile organic compound (VOC) analysis, a non-invasive technique, has gained increasing attention for early lung cancer screening, especially when combined with artificial intelligence (AI). This narrative review searched the PubMed, Scopus, and IEEE Xplore databases for literature published between January 2015 and July 2026. Our analysis identifies three major challenges. First, technical standardization is lacking. Sampling, detection, and data-analysis steps lack unified protocols, limiting comparability across studies. Second, biomarker specificity remains insufficiently supported. Most discriminatory VOCs lack mechanistic validation at the metabolic-pathway level, and after rigorous adjustment for confounders, some markers lose discriminative power. Third, the level of clinical validation evidence is low. Single-center, small-sample case–control studies often report high diagnostic performance, but performance tends to decline in multicenter prospective validations. Low-dose computed tomography (LDCT), the current screening gold standard, has been validated in large randomized trials but suffers from high false-positive rates. Exhaled VOCs, radiomics, and liquid biopsy each offer distinct advantages in terms of non-invasiveness or molecular information, but most approaches remain investigational. While exhaled VOCs with AI show promise for early lung cancer diagnosis, clinical translation requires solutions to standardization and validation gaps. Future research should focus on establishing end-to-end standardized workflows, conducting multicenter prospective validations, and exploring integrated multi-technology clinical pathways.