Interstitial lung disease associated with collagen vascular disorders: disease quantification using a computer-aided diagnosis tool

The purpose of this study was to evaluate a computer-aided diagnosis (CAD) tool compared to human observers in quantification of interstitial lung disease (ILD) in patients with collagen-vascular disorders. A total of 52 patients with rheumatoid arthritis (n = 24), scleroderma (n = 14) and systemic lupus erythematosus (n = 14) underwent thin-section CT. Two independent observers assessed the extent of ILD (EoILD), reticulation (EoRet) and ground-glass opacity (EoGGO). CAD assessed EoILD twice. Pulmonary function tests were obtained. Statistical evaluation used 95% limits of agreement and linear regression analysis. CAD correlated well with diffusing capacity (DL_CO) (R = −0.531, P < 0.0001) and moderately with forced vital capacity (FVC) (R = −0.483, P = 0.0008). There was close correlation between CAD and the readers (EoILD vs. CAD: R = 0.716, P < 0.0001; EoRet vs. CAD: R = 0.69, P < 0.0001). Subgroup analysis including patients with minimal EoGGO (<15%) strengthened the correlations between CAD and the readers, readers and PFT, and CAD and PFT. EoILD by readers correlated strongly with DL_CO (R = −0.705, P < 0.0001) and moderately with FVC (R = −0.559, P = 0.0002). EoRet correlated closely with DL_CO and moderately with FVC (DL_CO: R = −0.663; FVC: R = −0.436; P ≤ 0.005). The CAD system is a promising tool for ILD quantification, showing close correlation with human observers and physiologic impairment.