Niek RF Huttinga1, Tom Bruijnen1, Cornelis AT van den Berg1, and Alessandro Sbrizzi1
1Department of Radiotherapy, Computational Imaging Group for MR therapy & Diagnostics, University Medical Center Utrecht, Utrecht, Netherlands
1Department of Radiotherapy, Computational Imaging Group for MR therapy & Diagnostics, University Medical Center Utrecht, Utrecht, Netherlands
We tighten the gap towards clinical application of MR-MOTUS, by extending to a practical workflow for real-time 3D respiratory-motion tracking on an MR-LINAC. We demonstrate this workflow on five volunteers, showing that consistently realistic motion-fields are reconstructed in 160ms.
Fig. 3: A visualization of the reconstructed 3D motion-fields. For this visualization the motion-fields were upscaled to $$$3\times{3}\times{3}$$$mm, and the reference images were warped with these motion-fields to show moving images. Left to right shows a coronal slice of volunteers 1 to 5. Note that these reconstructions are 3D+t, but that only one slice and one respiratory cycle per volunteer were visualized because of file-size restrictions.
Fig. 5: (A) Comparison between MR-MOTUS reconstructions $$$\boldsymbol{\Psi}$$$ for volunteer 1 and a 1D motion surrogate extracted with PCA on the feet-head spokes,[6] overlayed on projection images obtained from a 1D FFT of the FH-spokes along the readout direction. (B) A quantitative validation by means of the Pearson correlation between the 1D motion surrogate and MR-MOTUS reconstructions. We obtain 0.88$$$\pm$$$0.097 correlation before AR-filtering, and 0.94$$$\pm$$$0.018 after AR-filtering, indicating an almost one-to-one linear correspondence.
