MR imaging of lung parenchyma at 0.2 T: evaluation of imaging techniques, comparative study with chest radiography and interobserver analysis

The purpose of this study was to evaluate low-field MR imaging of the lung parenchyma in comparison with postero-anterior (PA) and lateral chest radiographs (CR). One hundred one prospectively randomized patients who had received routine CR were additionally examined with magnetic resonance imaging (MRI) at 0.2 T. Utilized sequences were: constructive interference in steady state (CISS), true fast imaging in steady state precession (True-FISP) and T1-weighted spin-echo (T1SE). Consensus reading of two observers was performed for CR. Three other observers analyzed hardcopies of the MRI examinations for each sequence independently. The individual results for the comparisons between the sequences and CR were calculated using kappa coefficients with their corresponding confidence intervals. Additionally, an interobserver analysis was performed.The proportions of agreement for the three sequences compared with CR were high, with 0.93 for CISS, 0.89 for True-FISP and 0.91 for T1SE. The kappa coefficients and the corresponding confidence intervals were 0.81 [0.68; 0.95] for CISS, 0.72 [0.57; 0.88] for True-FISP and 0.78 [0.65; 0.92] for T1SE. Concerning CISS, differences between MRI and CR were mainly related to advantages resulting from cross-sectional imaging. The smallest 95% lower confidence bound of the three kappa measures for comparing the MR readers with each other was 0.97, indicating a high interobserver agreement. Low-field MRI of the lung parenchyma using the CISS sequence is well comparable with chest radiography and demonstrates slight advantages resulting from the cross-sectional imaging technique. (orig.)