Meeting Banner
Abstract #1456

Progressive Dual-Kriging for 2D and 3D QMRI Data Interpolation

Delphine Perie1, 2, Mohamed Aissiou1, 2, Julien Gervais1, 2, Francois Trocchu1, Gilles Beaudoin3, Guillaume Gilbert4

1Mechanical Engineering, Ecole Polytechnique de Montreal, Montreal, Quebec, Canada; 2Research center, CHU Sainte-Justine, Montreal, Quebec, Canada; 3Physics and Biomedical Engineering, CHUM Hopital Saint-Luc, Montreal, quebec, Canada; 4MR Clinical Science, Philips Healthcare, Cleveland, OH, United States

The aim of this study is to validate the reconstruction of 2D and 3D images from low to high spatial resolution. We performed two quantitative MR acquisitions (low and high resolution) on bovine discs. We adopted the Dual-Kriging formulation and we introduced the technique of progressive Kriging. For each MR sequence, Kriging was compared to other interpolation techniques including the zero padding, nearest neighbour, bilinear and bicubic techniques. Progressive Dual-Kriging is a flexible technique that can be optimized to interpolate 2D and 3D data based on the signal distribution. The computation is relatively fast and the interpolation is better fitted to the signal distribution as opposed to other techniques.

Abstract PDF Poster PDF

Keywords

accuracy acquiring acquisition acquisitions adapt adapted adjacent adopted advantage allow anisotropy apparent applications artifact augmented become better bilinear calibration chum clinical close coil collinear comfort common commonly comparing components computation computers considerable consists corrected cost decent defined density determined diffusion directional discs distribution dual easy elements encoding engineering enhancement enough error errors expressions extracted fast fitted fitting flexibility flexible formulation free gilbert good gradient graph head hyper implemented imply improve in vivo intense interpolate interpolated interpolating interpolation interpretation introduced journal lack linear magnetization many maps matrix mechanical minimize much multidimensional nearest noise nucleus objectives open opposed optimized padded padding partial patient performance permit physics pixel poor post prior progressive qualitatively quantitative quantitatively real reconstruction recovery reduces regressions relatively required resolution resolutions ringing robust saint saturation seems segment sense several shape slices slightly spatial spin stable step successfully suffer suffers system tail tend tensor tensors term transfer treatment turbo validate vertebrae visual volume zero