Objective To explore a novel method for early lung cancer screening based on exhaled breath analysis. MethodsThis study enrolled patients with suspected pulmonary malignancies and healthy individuals undergoing physical examinations at Sir Run Run Shaw Hospital, Zhejiang University School of Medicine (Qingchun and Qiantang campuses) from September 2023 to June 2024. Enrolled subjects were categorized into a lung cancer group, a benign nodule/tumor group, and a healthy control group. Exhaled breath samples were collected using a sensor array constructed from multiple graphene composite materials to capture breath fingerprints. Based on the collected data, screening and diagnostic models for lung cancer were developed and their performance was evaluated. ResultsA total of 4 580 subjects were included. Among them, 3 195 were pathologically diagnosed with pulmonary malignancies, including 1 394 males and 1 801 females with a mean age of (58.93±12.37) years, 599 were diagnosed with benign nodules/tumors including 339 males and 260 females with a mean age of (57.10±11.06) years, and 786 were healthy controls with no pulmonary nodules detected on chest CT including 420 males and 366 females with a mean age of (29.75±9.32) years. There were 4 031 patients in the training set and 549 patients in the external testing set. The screening model for high-risk populations (distinguishing patients with lung cancer/high-risk pulmonary nodules from healthy individuals) demonstrated excellent performance, with an area under the receiver operating characteristic curve (AUC) of 0.926. At the optimal Youden’s index (cutoff threshold of 63.5%), the external testing set achieved a specificity of 85.2%, a sensitivity of 88.4%, and an accuracy of 86.8%. The diagnostic model (distinguishing patients with lung cancer/premalignant lesions from those with benign pulmonary nodules/healthy individuals) achieved an AUC of 0.818. At its optimal Youden’s index (cutoff threshold of 47.0%), the external testing set showed a specificity of 71.7%, a sensitivity of 77.3%, and an accuracy of 74.5%. ConclusionThe non-invasive breath analysis platform based on a sensor array, developed in this study, can achieve rapid and relatively accurate lung cancer screening by analyzing breath fingerprints. This confirms the feasibility of this technology for early lung cancer screening and holds promise for facilitating the early detection and intervention of lung cancer.
With the increasing use of low-dose computed tomography (LDCT) in lung cancer screening and health examinations, the clinical goal of pulmonary nodule management has shifted from detecting nodules to identifying lesions that warrant intervention while avoiding overdiagnosis and overtreatment of low-risk disease. Artificial intelligence (AI) has consequently evolved from computer-aided detection and segmentation to benign-malignant differentiation, longitudinal growth assessment, invasiveness prediction, risk stratification, and closed-loop workflow support. This review outlines the developmental trajectory of AI for pulmonary nodules and summarizes advances in detection and segmentation, radiomics, end-to-end deep learning, longitudinal modeling, and multimodal integration across incidentally detected, screening-detected, subsolid, and multiple-nodule scenarios. Particular attention is given to real-world failure modes and their causes, including data and label bias, scanner- and protocol-related distribution shift, spectrum bias and overfitting in subsolid nodules, missed atypical lesions, automation bias, and management of discordance between AI outputs and guideline-based recommendations. Current evidence suggests that AI may improve detection efficiency and risk reclassification in selected settings; however, evidence regarding cross-population calibration, patient-relevant outcomes, cost-effectiveness, and post-deployment monitoring remains limited. AI should therefore be positioned as a human-in-the-loop adjunct within guideline-governed pathways, supported by scenario-specific validation, transparent reporting, interpretability and accountability, continuous performance auditing, and multidisciplinary decision-making.