Meeting Banner
Abstract #2846

MRSI reconstruction pipeline in the age of Deep Learning

Stanislav Motyka1, Lukas Hingerl1, Philipp Moser1, Asan Agibetov2, Georg Dorffner2, and Wolfgang Bogner1
1Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Vienna, Austria, 2Section for Artificial Intelligence and Decision Support (CeMSIIS), Medical University of Vienna, Vienna, Austria

Whole-brain MRSI measured with a concentric ring trajectories based FID-MRSI sequence generates large amounts of data, which makes post-processing very time-consuming (up to several hours). To speed-up the reconstruction, deep learning approaches could be applied. AUTOMAP provides an attractive solution to reconstruct data directly from non-Cartesian kSpace data. However, it requires single-channel data. Therefore, the coil combination needs to be performed in the kSpace domain. We showed that this strategy is in principle feasible, but requires future work on stability against noise.

This abstract and the presentation materials are available to members only; a login is required.

Join Here