Key points are not available for this paper at this time.
PURPOSE: To evaluate the impact of a new, cross-correlation based method for compensation of respiratory induced motion of the heart using an individually adapted three-dimensional (3D) translation or affine transformation approach. MATERIALS AND METHODS: A total of 32 patients underwent a routine cardiac MR examination. In each patient, a calibration scan was performed during free-breathing to register breathing-related motion within a 3D ellipsoid registration kernel covering the entire heart. Three navigators were employed for all three spatial dimensions (feet-head, anterior-posterior, and left-right) and the optimal translatory correction factors for each spatial dimension were determined. In addition, the cross-correlations for different motion models (no compensation, fixed 1D-translation, adapted 3D-translation, and affine transformation) were calculated. RESULTS: The mean correction factor for the feet-head direction was 0.45 +/- 0.13. Though the mean correction factors for the anterior-posterior and left-right direction were nearly zero (-0.01 +/- 0.08 and 0.02 +/- 0.09, respectively), the correction factors exceeded the amount of 0.1 in 12 (19%) and in 19 patients (30%), respectively. All motion compensation models showed significantly higher cross-correlations when compared to "no compensation" (P < 0.05). In particular, the affine transformation algorithm achieved the highest cross-correlation values (88.3 +/- 5.1%) with a significant increase compared to fixed 1D translation (84.7 +/- 6.5%, P < 0.05). CONCLUSION: A considerable number of patients demonstrated relevant breathing-related movement of the heart in the anterior-posterior or left-right direction in addition to the predominant breathing-related movement in the feet-head direction. Thus, it is recommended to compensate for all three spatial dimensions. The affine transformation algorithm combined with three navigators significantly improved breathing-related cardiac motion compensation when compared to the conventionally applied 1D translation with a fixed correction factor.
Jahnke et al. (Wed,) studied this question.