Evaluation of the Reproducibility of the Automated Measurement of the Extent of Interstitial Lung Disease on Chest CT

In interstitial lung disease (ILD), the extent of ILD on chest computed tomography (CT) is recognized as an important prognostic factor. In recent years, several tools based on texture analysis or deep learning methods have been developed to provide rapid and accurate automated quantification of ILD extent.

An advantage of automated scoring methods is that they theoretically offer perfect repeatability of measurement. However, this is only true if the measurement is repeated on the same images. Several parameters can alter the appearance of the lungs on scanner images, such as the degree of inspiration or the reconstruction kernel used to reconstruct the images. There is a lack of data on the reproducibility of the whole process of automated ILD quantification and therefore its accuracy.

Therefore, the purpose of this study is to evaluate the variability of automated ILD quantification on chest CT.

Reproducibility will be assessed by repeating chest CT scans and using different tools to measure the extent of disease in patients with ILD from 2 institutions. Chest CT will be repeated the same day. This will allow assessment of the variability of automated measurement of ILD extent between 2 CT scans performed on the same day and when using different software. It will also allow to assess the variability of lung volume between the 2 CT scans and the effect of disease (idiopathic pulmonary fibrosis or connective tissue disease-related ILD) on the reproducibility of ILD quantification.