3 Tesla proton MRI for the diagnosis of pneumonia/lung infiltrates in neutropenic patients with acute myeloid leukemia: initial results in comparison to HRCT.

PURPOSE: To evaluate the diagnostic accuracy of 3 Tesla proton MRI for the assessment of pneumonia/lung infiltrates in neutropenic patients with acute myeloid leukemia.
MATERIAL AND METHODS: In a prospective study, 3 Tesla MRI was performed in 19 febrile neutropenic patients (5 women, 14 men; mean age 61 years ± 14.2; range 23-77 years). All patients underwent high-resolution CT less than 24h prior to MRI. The MRI protocol (Magnetom Tim Trio, Siemens) included a T2-weighted HASTE sequence (TE/TR: 49 ms/∞, slice thickness 6mm) and a high-resolution 3D VIBE sequence with an ultra-short TE<1 ms (TE/TR 0.8/2.9 ms, slice thickness 2mm). The VIBE sequence was examined before and after intravenous injection of 0.1 mmol/kg gadoterate meglumine (Dotarem, Guerbet). The presence of pulmonary abnormalities, their location within the lung, and lesion type (nodules, consolidations, glass opacity areas) were analyzed by one reader and compared to the findings of HRCT, which was evaluated by a second independent radiologist who served as the reference standard. The findings were compared per lobe in each patient and rated as true positive (TP) findings if all three characteristics (presence, location, and lesion type) listed above were concordant to HRCT.
RESULTS: Pulmonary abnormalities were characterized by 3 Tesla MRI with a sensitivity of 82.3% and a specificity of 78.6%, resulting in an overall accuracy of 88% (NPV/PPV 66.7%/89.5%). In 51 lobes (19 of 19 patients), pulmonary abnormalities visualized by MR were judged to be concordant in their location and in the lesion type identified by both readers. In 22 lobes (11 of 19 patients), no abnormalities were present on either MR or HRCT (true negative). In 6 lobes (5 of 19 patients), ground glass opacity areas were detected on MRI but were not visible on HRCT (false positives). In 11 lobes (7 of 19 patients), MRI failed to detect ground glass opacity areas identified by HRCT. However, since the abnormalities were disseminated in these patients, accurate treatment decisions were possible in every case based on MRI. In one case MRI showed a central area of cavitation, which was not visualized by HRCT.
CONCLUSION: Infectious nodules and consolidations can be detected in neutropenic patients with acute myeloid leukemia with a sufficient diagnostic accuracy by 3 Tesla MRI. Detection of ground glass opacity areas is the main limitation of 3-Tesla MRI when compared to HRCT.