Computer-aided analysis in evaluation and grading of interstitial lung diseases in correlation with CT-based visual scoring and pulmonary function tests

Abstract

Abstract Background Interstitial lung diseases (ILDs) represent a large group of more than 200 different entities. High resolution computed tomography (HRCT) is accepted as the gold standard imaging modality in the diagnosis of ILD. The visual-based scoring offers an advantage in finding a specific type of ILD. Computer-aided CT attenuation histogram is another way of characterizing and quantifying diffuse lung disease. The histogram analysis (HIST) consists of calculating skewness, kurtosis, and mean lung density to quantify lung disease and monitor progression. The aim of our study was to investigate the value of computer-aided analysis of HRCT for interstitial lung diseases in correlation with scoring and pulmonary function tests. Results This prospective study included 50 patients with suspected ILD. The mean age of patients was 46.7 years ± 12.5. Mean forced expiratory volume FEV1 was 63.6 ± 20.9. HRCT examination was done for all patients followed by CT-based visual scaling. Most of the studied patients (43.3%) had a CT visual semi-quantitative scoring ranged between 40 and 64. CT-based lung density histograms (LDH) were obtained for all patients using the 3D Slicer Software (Chest Imaging Platform). There was a significant difference between patient’s groups of different (mild, moderate, and severe) grades of ILD according to FEV1 regarding MLD, skewness, and kurtosis of corresponding CT-based density histograms (p values < 0.001). More significant and higher correlation was observed between computerized aided CT quantified mean lung densities (MLD) and (FEV1) (p value < 0.001 and r = − 0.570). The ROC curve analysis demonstrated good performance for CT visual scoring with PFT (AUC = 0.71); a cutoff scoring 15 or higher was associated with best sensitivity (75%) and specificity (100%). Meanwhile, ROC curve analysis for MLD and FEV1 demonstrated an excellent performance for computer-based CT quantification (AUC = 0.85) with a value of − 769 HU which increased sensitivity to 65% and specificity to 100%. Conclusion Visual-based scoring techniques offer an advantage in finding a specific type of ILD. Computer-based quantification system could be a means for accurately monitoring the disease progression or response to therapy.