Click here to


Are you sure ?

Yes, do it No, cancel

Time-Tagged Reconstruction: A Robust Approach to Image Reconstruction in Breast DCE-MRI

T Easley*, F Pineda , G Karczmar , B Kim , R Barber , University of Chicago, Chicago, IL


(Sunday, 7/14/2019) 1:00 PM - 2:00 PM

Room: 221CD

Purpose: To facilitate the reconstruction of pharmacokinetic data in breast DCE-MRI, the proposed method reconstructs highly under-sampled breast DCE-MRI data to enhance pharmacokinetic analysis. We use this method to reconstruct high temporal resolution images with data under-sampled from standard, random, uniformly under-sampled Cartesian acquisitions at 14x acceleration. We assess features and performance of both the reconstruction method and tested acquisition schemes.

Methods: The reconstruction partitions k-space measurements by the temporal resolution of the desired reconstruction and imposes a voxel-wise temporal smoothness penalty to constrain the underdetermined reconstruction.Once the acquisition path and k-space data have been computed, the data are partitioned into intervals determined by the temporal resolution of the reconstruction. After requiring that each reconstructed image match the k-space points in the partition corresponding to its interval, we minimize over a weighted and regularized objective function that penalizes the per-voxel discretized second time derivative in the image-domain data. Minimization is carried out via conjugate gradient descent.Each sequence completed a Nyquist-complete k-space sample every 3.5s and was reconstructed at 0.25s temporal resolution.

Results: Time-tagged reconstructions were accurate to the phantom and nearly identical to one another: mean difference did not exceed 0.15% between reconstructions from any of the three sampling patterns, and 99% of reconstructed voxels show less than 2% error (Fig. 3). For each acquisition path, the time-tagged and standard IFFT reconstructions were computed (Fig. 4). Standard IFFT reconstructions are computed at 3.5s temporal resolution: resulting images come from a fully-sampled k-space dataset.

Conclusion: Three different Cartesian acquisition schemes (random-order, uniform undersampling and standard) were simulated and reconstructed. All were recovered well by the time-tagged reconstruction method, though different acquisition schemes produced varied results under IFFT reconstruction (Fig. 4). The temporal resolution obtained by this allows precise arterial bolus tracking and measurements of tumor blood flow.


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email