Related Experiment Video
Updated: Dec 14, 2025
Author Spotlight: Advancing 3D Modeling for Enhanced Diagnosis and Treatment of Pulmonary Nodules in Early-Stage Lung Cancer
Published on: October 13, 2023
Deep Learning-based Automatic Detection Algorithm for Reducing Overlooked Lung Cancers on Chest Radiographs
Sowon Jang1, Hwayoung Song1, Yoon Joo Shin1
1From the Department of Radiology, Seoul National University Bundang Hospital, 300 Gumi-dong, Bundang-gu, Seongnam-si, Gyeonggi-do 13620, Korea (S.J., H.S., Junghoon Kim, Jihang Kim, K.W.L., S.S.L., K.H.L.); Department of Radiology, Konkuk University Medical Center, Seoul, Korea (Y.J.S.); Seoul National University College of Medicine, Institute of Radiation Medicine, Seoul National University Medical Research Center, Seoul, Korea (K.W.L.); Department of Public Health Science, Graduate School of Public Health, Seoul National University, Seoul, Korea (W.L.); and Program in Biomedical Radiation Sciences, Department of Transdisciplinary Studies, Graduate School of Convergence Science and Technology, Seoul National University, Seoul, Korea (S.L.).
A deep learning-based automatic detection algorithm (DLAD) improved lung cancer detection on chest radiographs. Observers using the DLAD identified more overlooked cancers and recommended appropriate follow-up CT scans, enhancing diagnostic accuracy.
Area of Science:
- Radiology
- Artificial Intelligence
- Oncology
Background:
- Uncertainty exists regarding the effectiveness of deep learning-based automatic detection algorithms (DLAD) for identifying malignant nodules on chest radiographs in lung cancer diagnosis.
- Assessing the impact of DLAD on observer performance for lung cancer detection is crucial for improving diagnostic outcomes.
Purpose of the Study:
- To evaluate the efficacy of a DLAD in improving observer performance for detecting lung cancers on chest radiographs.
- To compare diagnostic accuracy metrics, including sensitivity and recommendation rates for chest CT, with and without DLAD assistance.
Main Methods:
- Retrospective analysis of 117 lung cancer patients and 234 healthy controls with chest radiographs from 2010-2014.
- Nine observers reviewed radiographs with and without DLAD assistance, assessing lung cancer detection and recommending chest CT follow-up.
- Observer performance was quantified using area under the alternative free-response receiver operating characteristic curve (AUC), sensitivity, and CT recommendation rates.
Main Results:
- The average AUC significantly increased from 0.67 without DLAD to 0.76 with DLAD (P < .001).
- DLAD use improved detection of overlooked lung cancers (sensitivity: 53% with DLAD vs. 40% without; P < .001) and increased appropriate chest CT recommendations (62% with DLAD vs. 47% without; P < .001).
- No significant difference in chest CT recommendation rates was observed in the healthy control group (8-10%; P = .13).
Conclusions:
- DLAD can significantly enhance observer performance in detecting lung cancers on chest radiographs.
- The algorithm aids in reducing overlooked malignancies without a disproportionate increase in follow-up imaging recommendations.
- DLAD shows promise as a tool to improve early lung cancer diagnosis through improved interpretation of chest radiographs.
