SU-F-I-21: Unfolding of Aliased Dynamic Acquisitions for the Acceleration of Breast Dynamic-Contrast Enhanced MRI (DCE-MRI).

PURPOSE: To determine the feasibility of increasing temporal resolution in breast DCE-MRI by reducing the field-of-view (FOV) and unfolding the aliased images using un-aliased subtraction images as guides.

METHODS: Experience with an 'ultrafast' protocol for breast DCE-MRI (6-10s acquisitions) showed that the number of significantly enhancing voxels was very low (<7% of the FOV) in the first 30-45s after contrast injection. The sparse enhancement in these images suggests that if the FOV is reduced in the phase-encoding direction (reducing acquisition time) and aliasing occurs, the likelihood of two significantly-enhancing voxels overlapping in the aliased image is relatively low. Aliased images were simulated from the first 30s of fully-sampled ultrafast acquisitions. Five cases with known enhancing lesions and moderate or marked parenchymal enhancement were selected (i.e. more enhancing voxels). FOV's were reduced by 40%. Subtractions were generated for all images, and filtered to determine significant enhancement. The locations that each aliased enhancing voxel could have come from were determined. If only one of the possible locations was enhanced at the later, fully sampled, time-point, the aliased voxel was copied to that location. If both were enhanced overlap likely occurred and the voxel was copied to both locations in the 'unfolded' image.

RESULTS: On average 2.2% (range 0% - 9.81%) of the significantly enhancing voxels, above the chest wall, overlapped in the aliased images. This relatively low number means that it's possible to generate unfolded images that are visually similar to the full-FOV images.

CONCLUSION: These simulations show that it is possible to mostly recover early post-contrast breast DCE-MRI images from aliased acquisitions. While this method cannot separate the signal from each overlapping voxel in the aliased acquisitions, in-progress work will address this by changing the aliasing factor for each time-point, giving each aliased voxel a distinct 'trajectory'.