Room: Exhibit Hall | Forum 8
Purpose: To create a publicly available tool for evaluating effects of different sampling schemes in breast DCE-MRI and to produce digital phantoms to test and compare k-space sampling schemes over a variety of realistic pathologies.
Methods: A simulated scanner was developed to test image acquisition schemes in breast DCE-MRI. Rather than approximate k-space acquisition via FFT, this scanner samples a time-varying phantom at each k-space acquisition point.Breast DCE-MRI digital phantoms were built from patient scans from an ultrafast spoiled gradient echo dynamic contrast enhanced MRI sequence with temporal resolution 3.5-7.2s. To simulate continuous-time change in the scanner without invoking a particular signal model, patient images were interpolated with splines in the temporal dimension.Simple phantoms using a spoiled gradient echo signal model were also constructed with parameters estimated from patient data (e.g., native T1/T2*) and tested on the virtual scanner. In addition, we plan to produce model-driven phantoms with more complete quantitative signal models.Some sample k-space acquisition schemes were constructed and tested: 2D radial and spiral acquisitions, sliding-window slice acquisition, and standard and random-order Cartesian sampling in 3D. Acquisition type, voxel size, and phantom physical dimensions specify sampling paths.
Results: The tool produced high-quality reconstructed images of phantoms, and applying various sampling and reconstruction schemes led to reconstructed images that showed characteristic artifacts. High spatiotemporal-resolution phantoms have been successfully produced, both directly from patient data (splines) and data-estimated physical models.Some sampling schemes were tested: a comparison between a sliding window and 2D-slice Cartesian acquisition (respective temporal resolutions: 0.15s/3.8s) showed periodic bias in the reconstructed sliding window signal and an overestimation of initial enhancement time in the Cartesian scheme.
Conclusion: Our publicly available simulation tool will be capable of testing the viability of a variety of image acquisition and reconstruction algorithms on digital phantoms with realistic pathology.