Proton MRI in the evaluation of pulmonary sarcoidosis: comparison to chest CT.

PURPOSE

The purpose of this study was to determine the feasibility of proton MRI of the lung in sarcoidosis patients and the agreement between the imaging appearance of pulmonary sarcoidosis on MRI and CT.

MATERIALS AND METHODS

Chest CT scans and dedicated pulmonary MRI scans (including HASTE, VIBE, and TrueFISP sequences) were performed within 90 days of each other in 29 patients. The scans were scored for gross parenchymal opacification, reticulation, nodules, and masses using a 3-point lobar scale. Total and subset scores for corresponding MRI and CT scans were compared using the Spearman correlation test, Bland-Altman plots, and Cohen's quadratic-weighted kappa analysis. MRI scores were compared to CT by lobe and disease category, using percentage agreement, Spearman rank correlation, and Cohen's quadratic-weighted kappa.

RESULTS

The mean (± s.d.) time between MRI and CT scans was 33 ± 32 days. There was substantial correlation and agreement between total disease scoring on MRI and CT with a Spearman correlation coefficient of 0.774 (p<0.0001) and a Cohen's weighted kappa score of 0.646. Correlation and agreement were highest for gross parenchymal opacification (0.695, 0.528) and reticulation (0.609, 0.445), and lowest in the setting of nodules (0.501, 0.305). Agreement testing was not performed for mass scores due to low prevalence. Upper lobe scoring on MRI and CT demonstrated greater agreement compared to the lower lobes (average difference in Cohen's weighted kappa score of 0.112).

CONCLUSION

There is substantial correlation and agreement between MRI and CT in the scoring of pulmonary sarcoidosis, though MRI evaluation in the upper lobes may be more accurate than in the lower lobes.