Meeting Banner
Abstract #4571

Spline-Based Variational Reconstruction of Variable Density Spiral K-Space Data with Automatic Parameter Adjustment

Benedicte Delattre1, Jean-Nol Hyacinthe1, Jean-Paul Valle1, Dimitri Van De Ville2,3

1Faculty of Medicine, University of Geneva, Geneva, Switzerland; 2Biomedical Imaging Group (BIG), Ecole Polytechnique Fdrale de Lausanne (EPFL), Lausanne, Switzerland; 3Work supported in part by the Center for Biomedical Imaging (CIBM), Geneva and Lausanne, Switzerland

Small-animal cardiac imaging is very challenging because we face with several problems like resolution or flux artifacts. One possible way to assess them is the use of non-Cartesian acquisition scheme like variable density spiral. Regridding reconstruction, which is the most popular alternative, however introduces noticeable artifacts due to k-space interpolation, especially when dealing with undersampled trajectories. We propose a variational approach where the image is described by a spline model and where an automatic adjustment of the regularizing weight is implemented. We evaluate our framework for various degrees of the spline model and different orders of derivation of the regularizer.