Related Experiment Video
Updated: May 29, 2026
10:44
Three-Dimensional Phase Resolved Functional Lung Magnetic Resonance Imaging
Published on: June 21, 2024
Fully Automated Quantification of Functional Small Airway Disease at Inspiratory and Expiratory Chest CT Using Deep
Stefan Gherca1, Shan Yang1, Jens Bremerich1
1Department of Radiology, Clinic of Radiology and Nuclear Medicine, University Hospital Basel, University of Basel, Petersgraben 4, CH-4031 Basel, Switzerland.
Radiology. Cardiothoracic Imaging
|May 28, 2026
Summary
A deep learning (DL) tool for automated functional small airway disease (fSAD) quantification shows excellent agreement and high accuracy compared to manual methods. This machine learning approach offers an efficient and reliable alternative for pulmonary diagnostics.
Area of Science:
- Radiology and Medical Imaging
- Artificial Intelligence in Medicine
- Pulmonary Medicine
Background:
- Functional small airway disease (fSAD) assessment at chest CT traditionally relies on time-consuming semimanual methods.
- Accurate and efficient quantification of fSAD is crucial for diagnosing and managing pulmonary conditions.
- The integration of deep learning (DL) offers potential for automating complex medical image analysis tasks.
Purpose of the Study:
- To evaluate the accuracy and time efficiency of a DL-based tool for automated fSAD quantification.
- To compare the DL tool's performance against a conventional semimanual assessment method.
- To assess the diagnostic classification accuracy for clinically relevant fSAD using the automated tool.
Main Methods:
- Retrospective analysis of 249 paired inspiratory and expiratory chest CT examinations.
- Comparison between a novel DL-based automated fSAD assessment (fSADauto) and a commercial semimanual method (fSADman).
- Evaluation metrics included Spearman correlation, Bland-Altman analysis, sensitivity, specificity, and accuracy; time efficiency was also measured.
Main Results:
- Excellent agreement was observed between fSADauto and fSADman (Spearman r = 0.93, P < .01), with minimal bias (-1.7%).
- The DL tool achieved high diagnostic accuracy for clinically relevant fSAD (≥28% lung involvement): 100% sensitivity, 96.8% specificity, and 97.2% accuracy.
- Automated analysis significantly reduced quantification time compared to the semimanual method.
Conclusions:
- DL-based automated quantification of fSAD demonstrates excellent agreement and high diagnostic accuracy compared to semimanual assessment.
- The automated tool is a time-efficient and reliable alternative, potentially improving workflow optimization in pulmonary diagnostics.
- This machine learning approach shows promise for enhancing the analysis of functional small airway disease on chest CT.
