Diagnostic accuracy of unenhanced, contrast-enhanced perfusion and angiographic MRI sequences for pulmonary embolism diagnosis: results of independent sequence readings.

Objectives: To independently evaluate unenhanced, contrast-enhanced perfusion and angiographic MR sequences for pulmonary embolism (PE) diagnosis.

Methods: Prospective investigation, including 274 patients who underwent perfusion, unenhanced 2D steady-state-free-precession (SSFP) and contrast-enhanced 3D angiographic MR sequences on a 1.5-T unit, in addition to CTA (CT angiography). Two independent readers evaluated each sequence independently in random order. Sensitivity, specificity, predictive values and inter-reader agreement were calculated for each sequence, excluding sequences judged inconclusive. Sensitivity was also calculated according to PE location.

Results: Contrast-enhanced angiographic sequences showed the highest sensitivity (82.9 and 89.7 %, reader 1 and reader 2, respectively), specificity (98.5 and 100 %) and agreement (kappa value 0.77). Unenhanced angiographic sequences, although less sensitive overall (68.7 and 76.4 %), were sensitive for the detection of proximal PE (92.7 and 100 %) and showed high specificity (96.1 and 99.1 %) and good agreement (kappa value 0.62). Perfusion sequences showed lower sensitivity (75.0 and 79.3 %), specificity (84.8 and 89.7 %) and agreement (kappa value 0.51), and a negative predictive value of 84.8 % at best.

Conclusions: Compared with contrast-enhanced angiographic sequences, unenhanced sequences demonstrate lower sensitivity, except for proximal PE, but high specificity and agreement. The negative predictive value of perfusion sequences was insufficient to safely rule out PE.

Key Points: • Unenhanced angiographic MR sequences are very specific and can identify proximal PE. • Contrast-enhanced MR angiographic sequences show high sensitivity for PE diagnosis. • A normal MR perfusion result does not exclude PE. • Inter-reader agreement is better for angiographic than perfusion MR sequences.