Related Experiment Video
Updated: Jan 9, 2026
08:05
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
14.7K
Automated interstitial lung abnormalities detection at CT: external validation and potential recognition of traction
Yusei Nakamura1,2, Taiki Fukuda3,4, Kota Aoyagi5
1Center for Pulmonary Functional Imaging, Department of Radiology, Brigham and Women's Hospital and Harvard Medical School, 75 Francis Street, Boston, MA, 02115, USA. ynakamura4@bwh.harvard.edu.
Japanese Journal of Radiology
|December 10, 2025
Summary
An artificial intelligence system for detecting interstitial lung abnormalities (ILA) shows robust performance across diverse populations. Its AI scores correlate with the severity of traction bronchiectasis/bronchiolectasis, aiding in diagnosis.
Area of Science:
- Radiology and Medical Imaging
- Artificial Intelligence in Medicine
- Pulmonary Medicine
Background:
- An artificial intelligence (AI) system for detecting interstitial lung abnormalities (ILA) has been developed.
- External validation is crucial to assess the AI system's generalizability and reliability across different populations.
- Understanding the relationship between AI-derived metrics and specific disease features, like traction bronchiectasis/bronchiolectasis, is important for clinical application.
Purpose of the Study:
- To externally validate an AI system for ILA detection across diverse cohorts.
- To assess the robustness of the AI system in identifying interstitial lung abnormalities.
- To investigate the association between AI-derived ILA probability scores and the severity of traction bronchiectasis/bronchiolectasis.
Main Methods:
- Secondary analysis of CT scans from the Rotterdam Study and the AGES-Reykjavik Study.
- AI system calculated ILA probability scores (AI scores) for each CT scan.
- Independent reader evaluation for ILA and consensus-based assessment of traction bronchiectasis/bronchiolectasis severity (TBI).
- Statistical analysis using Receiver Operating Characteristic (ROC) analysis and Kruskal-Wallis test.
Main Results:
- The AI system achieved high diagnostic performance with an area under the ROC curve of 0.841 in the Rotterdam Study and 0.823 in the AGES-Reykjavik Study.
- AI scores demonstrated a strong correlation with expert readers' assessments of ILA presence and certainty.
- Higher AI scores were significantly associated with increased severity of traction bronchiectasis/bronchiolectasis, as indicated by the TBI.
Conclusions:
- The AI system for ILA detection exhibits robust performance and generalizability across different populations.
- AI-derived ILA probability scores are associated with the severity of traction bronchiectasis/bronchiolectasis.
- This validated AI system holds potential for aiding in the assessment of interstitial lung diseases.
Keywords:
Artificial intelligenceCTExternal validationInterstitial lung abnormalitiesMachine learningTraction bronchiectasis/bronchiolectasis
