• Department of Blood Transfusion, Key Laboratory of Cancer Prevention and Therapy in Tianjin, National Clinical Research Center for Cancer, Tianjin's Clinical Research Center for Cancer, Tianjin Medical University Cancer Institute and Hospital, Tianjin Medical University, Tianjin, China.
• Public Health Science and Engineering College, Tianjin University of Traditional Chinese Medicine, Tianjin, China.
• Department of Epidemiology and Statistics, School of Public Health, Hebei Medical University, Hebei Key Laboratory of Environment and Human Health, Shijiazhuang, China.
• Department of Cancer Epidemiology and Biostatistics, Tianjin Key Laboratory of Molecular Cancer Epidemiology, Clinical Research Center for Cancer, Key Laboratory of Cancer Prevention and Therapy, Tianjin's Clinical Research Center for Cancer, Tianjin Medical University Cancer Institute and Hospital, Tianjin Medical University, Tianjin, China.
• Department of Radiology, National Clinical Research Centre for Cancer, Key Laboratory of Cancer Prevention and Therapy, Tianjin's Clinical Research Center for Cancer, Tianjin Medical University Cancer Institute and Hospital, Tianjin Medical University, Tianjin, China.
• Department of Lung Cancer, Key Laboratory of Cancer Prevention and Therapy, National Clinical Research Center for Cancer, Tianjin's Clinical Research Center for Cancer, Tianjin Medical University Cancer Institute and Hospital, Tianjin Medical University, Tianjin, China.
• Department of Cancer Epidemiology and Biostatistics, Tianjin Key Laboratory of Molecular Cancer Epidemiology, Clinical Research Center for Cancer, Key Laboratory of Cancer Prevention and Therapy, Tianjin's Clinical Research Center for Cancer, Tianjin Medical University Cancer Institute and Hospital, Tianjin Medical University, Tianjin, China. songfengju@163.com.

BACKGROUND: Lung cancer is a health threat, particularly in regions where advanced screening methods like LDCT are limited. In China, chest X-rays (CXRs) are the primary tool for early detection. Integrating AI can enhance CXR diagnostic accuracy, addressing current challenges in early lung cancer detection.

METHODS: We collected 4079 CXRs from 2518 individuals at TMUCIH. These were divided into a training set (1762 patients, 2965 images) and a validation set (756 patients, 1114 images). A deep learning (DL) model, based on the CXR-RANet architecture, was developed and validated using two external cohorts: 24,697 individuals (88,562 images) from the PLCO dataset and 4848 individuals from the ChestDR dataset. The model's performance was compared with mainstream DL algorithms and traditional machine learning (ML) model in feature extraction and classification.

RESULTS: In the TMUCIH dataset, 47.8% of patients had positive CXR results, compared to 3.9% in PLCO and 13.7% in ChestDR. The CXR-RANet model achieved an AUC of 0.933 in the internal validation set and 0.818 in the ChestDR dataset. In the PLCO dataset, it predicted lung cancer occurrence with AUCs of 0.902, 0.897, and 0.793 for 3, 5, and 10 years, respectively. The model outperformed mainstream DL algorithms in feature extraction and most ML algorithms in classification.

CONCLUSION: The CXR-RANet presents a robust, scalable tool for diagnosing pulmonary nodules and lung cancer, enhancing the capabilities of community physicians in early detection and management, independent of expert experience. Its superior performance in feature extraction and classification underscores its value in lung cancer screening.