Quantitative CT Measurements of Interstitial Lung Disease: Same-Day Variability Between Two Vendors-A Prospective...

BACKGROUND. Clinical application of quantitative CT (QCT) measurements of interstitial lung disease (ILD) for longitudinal monitoring of disease progression requires an understanding of how such measurements vary across vendors. OBJECTIVE. The purpose of this study was to evaluate same-day variability between vendors for QCT measurements of fibrosis extent in patients with ILD. METHODS. This prospective study recruited participants with ILD diagnosed from January 2023 to November 2023. Participants underwent two serial unenhanced chest CT examinations on the same day using scanners from two different vendors (Discovery CT750 HD [GE HealthCare] and Somatom Definition AS+ [Siemens Healthineers]), with examinations separated by approximately 15 minutes. Examinations were reconstructed using three kernels (lung, chest, and bone) from GE Healthcare and three kernels (B60f, B50f, and B30f) from Siemens Healthineers. The lung and B60f kernels represented each vendor's primary high-frequency kernel for ILD evaluation. Commercially available deep learning-based software was used for QCT analysis. The fibrosis score was calculated as the sum of reticular opacity and honeycombing volume percentages. Intervendor variability was assessed using the Bland-Altman method, and agreement was assessed using the concordance correlation coefficient (CCC). RESULTS. The study included 48 participants (43 men and five women; mean age, 67.8 ± 5.2 [SD] years). The mean fibrosis score for the lung and B60f kernels was 12.64% and 11.84%, respectively. For the fibrosis score, the mean raw difference ranged from a minimum for the bone-B50f kernel pair (0.11%; 95% limits of agreement [LOA], -1.3% to 1.5%) to a maximum for the lung-B30f kernel pair (-2.52%; 95% LOA, -6.5% to 1.4%). The lung-B60f kernel pair showed a mean raw difference of -0.80% (95% LOA, -3.4% to 1.8%). The 95% LOA was narrowest for the bone-B50f kernel pair and widest for the lung-B30f kernel pair. The fibrosis score demonstrated substantial to almost perfect agreement across all intervendor kernel combinations (CCC, 0.953-0.998), including almost perfect agreement for the lung-B60f kernel pair (CCC, 0.989). CONCLUSION. QCT-derived fibrosis scores showed small variability (mean raw difference, -0.80%) for same-day CT examinations performed using systems from two different vendors when reconstructed with each vendor's primary high-frequency kernel for lung evaluation. CLINICAL IMPACT. QCT-derived fibrosis scores across different vendors may be informative for longitudinal assessment of ILD when interpreted in the context of expected measurement variability.