Julian Hossbach1,2,3, Daniel Nicolas Splitthoff3, Bryan Clifford4, Daniel Polak3, Stephan F. Cauley5, and Andreas Maier1
1Friedrich-Alexander-University Erlangen-Nuremberg, Erlangen, Germany, 2Erlangen Graduate School in Advanced Optical Technologies, Erlangen, Germany, 3Siemens Healthcare GmbH, Erlangen, Germany, 4Siemens Medical Solutions, Boston, MA, United States, 5Department of Radiology, Athinoula A. Martinos Center for Biomedical Imaging, Boston, MA, United States
1Friedrich-Alexander-University Erlangen-Nuremberg, Erlangen, Germany, 2Erlangen Graduate School in Advanced Optical Technologies, Erlangen, Germany, 3Siemens Healthcare GmbH, Erlangen, Germany, 4Siemens Medical Solutions, Boston, MA, United States, 5Department of Radiology, Athinoula A. Martinos Center for Biomedical Imaging, Boston, MA, United States
Using a
small motion-free scout, our method can prospectively detect and assess
patient's motion. For that, a neural network is trained to score the motion
based on a coil mixing error matrix. We show that this can be used to remove or
reacquire the N most affected ET to improve the image quality.
Fig. 5: Reconstruction
of the motion corrupted k-space by removing or replacing the highest ranked ET
by the NN (RMSE: red; SSIM yellow). The curve in the plot below shows the
ground truth motion; the numbers represent the ranking of the motion severity
obtained by the NN. For simplicity only rotation is simulated in this example.
Fig. 3: Structure
of the Neural network with respective output sizes (left) and the visualization
of the MS (right).
