Related Experiment Video
Updated: May 12, 2025
08:05
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
14.0K
nnU-Net-based high-resolution CT features quantification for interstitial lung diseases
Qiuxi Lin1, Ziyi Zhang2, Xirui Xiong3
1Department of Radiology, The First Affiliated Hospital of Guangzhou Medical University, Guangzhou, China.
European Radiology
|May 9, 2025
Summary
A new AI tool, CVILDES, accurately quantifies interstitial lung disease (ILD) on CT scans, matching expert visual assessments. This computer vision system offers a reliable method for evaluating ILD progression and treatment efficacy.
Area of Science:
- Radiology
- Artificial Intelligence
- Pulmonary Medicine
Background:
- Visual assessment of interstitial lung diseases (ILDs) on high-resolution computed tomography (HRCT) is time-consuming and suffers from poor inter-observer agreement.
- Accurate quantification of ILD abnormalities is crucial for evaluating disease progression and therapeutic efficacy.
Purpose of the Study:
- To develop a novel high-resolution CT (HRCT) quantification tool, CVILDES, for interstitial lung diseases (ILDs) utilizing the nnU-Net framework.
- To validate the clinical reliability and precision of CVILDES-derived quantitative parameters against expert visual evaluation.
Main Methods:
- A deep learning model based on nnU-Net was developed using supervised learning on HRCT scans from 83 ILD cases and 20 other diffuse lung diseases.
- Clinical validation involved quantitative evaluation of CT parenchymal patterns in 51 interstitial pneumonia with autoimmune features (IPAF) and 14 idiopathic pulmonary fibrosis (IPF) cases using CVILDES and visual assessment.
- Correlations between CVILDES and visual evaluations for ILD features, and between these methods and pulmonary function parameters (DLCO%, FVC%, FEV%), were analyzed.
Main Results:
- CVILDES successfully quantified all CT data, including total ILD extent, ground-glass opacity, consolidation, reticular pattern, and honeycombing.
- CVILDES-quantified results showed strong correlations with visual evaluation (r=0.64-0.89, p<0.0001), particularly for fibrosis extent (r=0.82, p<0.0001).
- CVILDES quantification demonstrated comparable or superior correlation with pulmonary function parameters compared to visual evaluation.
Conclusions:
- The nnU-Net-based CVILDES tool provides a reliable and accurate quantification of ILD abnormalities on HRCT.
- CVILDES offers a valuable potential application for quantitative assessment of ILDs in clinical settings, addressing limitations of visual assessment.
