[Detection of pulmonary nodules with breath-hold magnetic resonance imaging in comparison with computed tomography]

Evaluation of sensitivity and false positive findings of two fast MRI sequences for the detection of pulmonary nodules in comparison with spiral CT by two independent observers.

All 30 enrolled patients had a spiral CT or MSCT as base line study. MRI was performed with a 1.5 T MR scanner (Sonata, Siemens) using a transverse 3D gradient echo sequence (3D-GRE: TR/TE/flip = 2.9 ms/1.1 ms/5 degrees ) and a half-Fourier single-shot fast spin-echo sequence (HASTE: TR/TE/flip = 800/25/150 degrees ) acquired in three planes. A separate analysis for both sequences was carried out prospectively by two independent readers (A and B) with different experience regarding pulmonary MRI. Additionally, a retrospective reading with knowledge of the CT scans was done. Results were calculated for all lesions and for lesions larger than 4 mm.

The sensitivities were 73 %, 70 % and 84 % for the 3D-GRE sequence (reader A, reader B, retrospective reading) and 65 %, 68 % and 81 % for the HASTE sequence. For lesions larger than 4 mm, the sensitivities were 93 %, 89 %, 96 % for the 3D-GRE sequence and 85 %, 85 %, 96 % for the HASTE sequence. The rate of false positive findings depended on the reader's experience, but was generally lower for the 3D-GRE sequence with 2 and 16 (reader A and B) false positive nodules compared to 4 and 40 false positive findings for the HASTE sequence. The 3D-GRE sequence was more accurate for both readers (reader A: p = 0.08, reader B: p = 0.00003).

The sensitivity of MRI for the detection of lung nodules was only acceptable for lesions larger than 4 mm. The 3D-GRE sequence is superior to the HASTE sequence due to the reduced amount of false positive findings with comparable sensitivity.