Comparison of diagnostic quality of 3D ultrashort-echo-time techniques for pulmonary magnetic resonance imaging in free-breathing

Background Ultrashort-echo-time (UTE) sequences have been developed to overcome technical limitations of pulmonary magnetic resonance imaging (MRI). Recently, it has been shown that UTE sequences with breath-hold allow rapid image acquisition with sufficient image quality. However, patients with impaired respiration require alternative acquisition strategies while breathing freely. Purpose To compare the diagnostic performance of free-breathing three-dimensional (3D)-UTE sequences with different trajectories based on pulmonary imaging of immunocompromised patients. Material and Methods In a prospective study setting, two 3D-UTE sequences performed in free-breathing and exploiting non-Cartesian trajectories-one using a stack-of-spirals and the other exploiting a radial trajectory-were acquired at 3 T in patients undergoing hematopoietic stem cell transplantation. Two radiologists assessed the images regarding presence of pleural effusions and pulmonary infiltrations. Computed tomography (CT) was used as reference. Results A total of 28 datasets, each consisting of free-breathing 3D-UTE MRI with the two sequence techniques and a reference CT scan, were acquired in 20 patients. Interrater agreement was substantial for pulmonary infiltrations using both sequence techniques (kappa = 0.77 - 0.78). Regarding pleural effusions, agreement was almost perfect in the stack-of-spirals (kappa = 0.81) and moderate in the radial sequence (kappa = 0.59). No significant differences in detectability of the assessed pulmonary pathologies were observed between both 3D-UTE sequence techniques (P > 0.05), and their level of agreement was substantial throughout (kappa = 0.62-0.81). Both techniques provided high sensitivities and specificities (79%-100%) for the detection of pulmonary infiltrations and pleural effusions compared to reference CT. Conclusion The diagnostic performance of the assessed 3D-UTE MRI sequences was similar. Both sequences enable the detection of typical inflammatory lung pathologies.