Back to all papers

Data expansion and local tuning: A strategy for effective clinical deployment of AI in pulmonary nodule screening.

November 20, 2026pubmed logopapers

Authors

Su C,Ding Q,Wang S,Xia H,Fu Y

Affiliations (4)

  • Editorial Department of the School Newspaper, Shandong Second Medical University, Weifang, Shandong, China.
  • Department of Endocrinology, Yantai Traditional Chinese Medicine Hospital, Yantai, Shandong, China.
  • Department of Thoracic Surgery, Weifang People's Hospital, Shandong Second Medical University, Weifang, Shandong, China.
  • Faculty of Information Science and Engineering, Ocean University of China, Qingdao, Shandong, China.

Abstract

BackgroundThe clinical performance of pulmonary nodule screening systems may be influenced by training data size and differences in imaging characteristics across institutions. Improving model performance and cross-center adaptability is important for clinical deployment.ObjectiveTo investigate the effects of data expansion and localized fine-tuning on a pulmonary nodule screening system (PNSS) and evaluate its clinical utility for pulmonary nodule detection and classification.MethodsLung CT scans from Weifang People's Hospital and Shandong Provincial Hospital were retrospectively analyzed to assess PNSS performance with different training dataset sizes. Localized fine-tuning was further applied to improve cross-center adaptability. A prospective study at Weifang People's Hospital compared PNSS with experienced radiologists in routine clinical practice.ResultsData expansion increased detection sensitivity from 0.7703 to 0.9428 and classification accuracy from 0.7029 to 0.9486, with performance gains gradually approaching saturation. Localized fine-tuning improved cross-center diagnostic accuracy from 53.57% to 85.36% and sensitivity from 0.5495 to 0.8736. In the prospective evaluation, PNSS achieved a sensitivity of 0.9302 for malignant nodules, 13.95 percentage points higher than radiologists. Overall diagnostic accuracy was 0.92 compared with 0.85 for radiologists. PNSS required only 34 s per case, nearly eight times faster than radiologist review.ConclusionData expansion and localized fine-tuning significantly improved PNSS performance, adaptability, and efficiency, supporting its clinical application in pulmonary nodule screening and early lung cancer detection.

Topics

Journal Article

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAISlice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.