A comparison of hyperelastic warping of PET images with tagged MRI for the analysis of cardiac deformation

The objectives of the following research were to evaluate the utility of a deformable image registration technique known as hyperelastic warping for the measurement of local strains in the left ventricle through the analysis of clinical, gated PET image datasets. Two normal human male subjects were sequentially imaged with PET and tagged MRI imaging. Strain predictions were made for systolic contraction using warping analyses of the PET images and HARP based strain analyses of the MRI images. Coefficient of determination R2 values were computed for the comparison of circumferential and radial strain predictions produced by each methodology. There was good correspondence between the methodologies, with R2 values of 0.78 for the radial strains of both hearts and from an R2 = 0.81 and R2 = 0.83 for the circumferential strains. The strain predictions were not statistically different (P ≤ 0.01). A series of sensitivity results indicated that the methodology was relatively insensitive to alterations in image intensity, random image noise, and alterations in fiber structure. This study demonstrated that warping was able to provide strain predictions of systolic contraction of the LV consistent with those provided by tagged MRIWarping.