All initial and follow-up MRIs were interpreted by two experienced cardiovascular radiologists (O.V., P.L.), who reached a consensus unaware of the clinical findings. Scores ranging from − 1 to + 1 were assigned to six different myocardial areas (right ventricle, septal wall, and anterior, lateral, inferior and posterior walls) for morphologic T2-weighted and gadolinium-DTPA–enhanced spin-echo images, on the basis of a qualitative evaluation of signal intensity: 0 = normal (medium homogeneous signal intensity of the myocardial wall); + 1 = increased (hyperintense area with signal intensity higher than normal myocardial wall); − 1 = decreased (hypointense area with signal intensity lower than normal myocardial wall). Myocardial wall thickness was evaluated on axial, left ventricular long-axis and short-axis images; scores ranging from − 1 to + 1 were assigned to the same myocardial areas: 0 = normal (3 to 10 mm), + 1 = increased (> 10 mm); − 1 = decreased (< 3 mm). Functional gradient-echo images were analyzed using specific software (MASS; Medis; Lieden, the Netherlands). Regional wall motion was semiquantitatively assessed by cine-loop simulation in both the true long and short left ventricular axes, and the two observers assigned scores ranging from 0 to + 1 after reaching a consensus (0 = normokinetic, + 1 = hypokinetic, akinetic or dyskinetic) for the six myocardial areas: right ventricle, septal wall, and anterior, lateral, inferior and posterior walls. A total score, obtained by adding together the scores obtained for each myocardial area in terms of signal intensity, myocardial thickness, and contractility, was then calculated for each patient.