ABSTRACT
Quantitative evaluation using image biomarkers calculated from threshold-segmented low-attenuation areas on chest computed tomography (CT) images for diagnosing chronic obstructive pulmonary diseases (COPD) has been widely investigated. However, the segmentation results depend on the applied threshold and slice thickness of the CT images because of the partial volume effect (PVE). In this study, the air volume fraction (AV/TV) of lungs was calculated from CT images using a two-compartment model (TCM) for COPD diagnosis. A relative air volume histogram (RAVH) was constructed using the AV/TV values to describe the air content characteristics of lungs. In phantom studies, the TCM accurately calculated total cavity volumes and foam masses with percent errors of less than 8% and ±4%, respectively. In patient studies, the relative volumes of normal and damaged lung tissues and the damaged-to-normal RV ratio were defined and calculated from the RAVHs as image biomarkers, which correctly differentiated COPD patients from controls in 2.5- and 5-mm-thick images with areas under receiver operating characteristic curves of >0.94. The AV/TV calculated using the TCM can prevent the effect of slice thickness, and the image biomarkers calculated from the RAVH are reliable for diagnosing COPD.
