Related Experiment Video
Updated: Oct 11, 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 for Lung Cancer Detection on Screening CT Scans: Results of a Large-Scale Public Competition and an
Colin Jacobs1, Arnaud A A Setio1, Ernst T Scholten1
1Department of Radiology, Nuclear Medicine and Anatomy, Radboud University Medical Center, Geert Grooteplein 10, 6525 GA, Nijmegen, the Netherlands (C.J., A.A.A.S., E.T.S., P.K.G., H.B., M.B., B.G., S.S., B.v.G.); Department of Digital Technology & Innovation, Siemens Healthineers, Erlangen, Germany (A.A.A.S.); Department of Radiology, University Medical Center Utrecht, Utrecht, the Netherlands (F.A.M.H., P.A.d.J.); ETZ (Elisabeth-TweeSteden Ziekenhuis), Tilburg, the Netherlands (E.R.); Section of Radiology, Department of Medicine and Surgery (DiMeC), University of Parma, Parma, Italy (M.S.); Department of Radiology, Meander Medical Center, Amersfoort, the Netherlands (K.C., S.S.); Department of Radiology, AZ Zeno, Knokke-Heist, Belgium (J.M.); Department of Imaging, Royal Brompton Hospital, London, England (A.D.); Division of Cancer Prevention (P.F.P.) and Center for Biomedical Informatics & Information Technology (K.F.), National Cancer Institute, National Institutes of Health, Bethesda, Md; British Columbia Cancer Agency and the University of British Columbia, Vancouver, Canada (S.C.L.); and Fraunhofer MEVIS, Bremen, Germany (B.v.G.).
Deep learning algorithms show performance comparable to radiologists in detecting lung cancer on low-dose CT scans. These AI tools, developed in a public competition, offer promising results for thoracic oncology screening.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Lung cancer screening with low-dose computed tomography (CT) is crucial for early detection.
- Deep learning (DL) algorithms have shown potential in medical image analysis.
- Assessing the performance of DL algorithms against human experts is essential for clinical adoption.
Purpose of the Study:
- To evaluate if deep learning algorithms from a public competition can detect lung cancer on low-dose CT scans.
- To compare the diagnostic performance of these algorithms with that of experienced radiologists.
Main Methods:
- Retrospective analysis of 300 low-dose CT scans (150 competition, 150 independent).
- Utilized three top-performing deep learning algorithms (grt123, JWDH, Aidence) from the Kaggle Data Science Bowl 2017.
- Compared algorithm performance against 11 radiologists using receiver operating characteristic (ROC) analysis.
Main Results:
- Deep learning algorithms achieved areas under the ROC curve (AUC) between 0.877 and 0.902.
- Radiologists achieved an average AUC of 0.917.
- No statistically significant difference in performance was found between radiologists and two of the deep learning algorithms (JWDH and Aidence).
Conclusions:
- Deep learning algorithms demonstrate performance levels similar to radiologists for lung cancer identification on low-dose CT.
- These findings suggest the potential of AI in augmenting lung cancer screening programs.
- Further validation and integration studies are warranted for clinical implementation.
