Optimization of tagged MRI for quantification of liver stiffness using computer simulated data.
Bottom Line: In the absence of noise, the estimation accuracy of the displacement fields increased as tag spacings decreased.EIs for each of the four displacement fields were lower at 0° and the local minima of the EI were found to correspond to multiples of pixel size.The accuracy of the estimation decreased for increasing levels of added noise; as the level increased, the improved estimation caused by decreasing the tag spacing tended to zero.
Affiliation: IRCCS SDN Foundation, Naples, Italy.
The heartbeat has been proposed as an intrinsic source of motion that can be used in combination with tagged Magnetic Resonance Imaging (MRI) to measure displacements induced in the liver as an index of liver stiffness. Optimizing a tagged MRI acquisition protocol in terms of sensitivity to these displacements, which are in the order of pixel size, is necessary to develop the method as a quantification tool for staging fibrosis. We reproduced a study of cardiac-induced strain in the liver at 3T and simulated tagged MR images with different grid tag patterns to evaluate the performance of the Harmonic Phase (HARP) image analysis method and its dependence on the parameters of tag spacing and grid angle. The Partial Volume Effect (PVE), T1 relaxation, and different levels of noise were taken into account. Four displacement fields of increasing intensity were created and applied to the tagged MR images of the liver. These fields simulated the deformation at different liver stiffnesses. An Error Index (EI) was calculated to evaluate the estimation accuracy for various parameter values. In the absence of noise, the estimation accuracy of the displacement fields increased as tag spacings decreased. EIs for each of the four displacement fields were lower at 0° and the local minima of the EI were found to correspond to multiples of pixel size. The accuracy of the estimation decreased for increasing levels of added noise; as the level increased, the improved estimation caused by decreasing the tag spacing tended to zero. The optimal tag spacing turned out to be a compromise between the smallest tag period that is a multiple of the pixel size and is achievable in a real acquisition and the tag spacing that guarantees an accurate liver displacement measure in the presence of realistic levels of noise.
Related in: MedlinePlus
Mentions: When noise was added, the smoothness of the displacement measurement was affected (Figures 2(d)–(e)). Plots of EIs calculated for each Dgi and the three levels of noise as a function of tag spacing are shown in Figure 6. Fitting the new calculated EIs with linear functions (Table 2), it can be observed that the slope of the linear fit decreased for each increase in the level of noise, negating almost completely the advantage of using the smallest tag spacing. This was particularly true for Dg1 at the maximum level of noise (the EI remained over 25%, even at tag spacing of 4).