Vortex Detection in 4D MRI Data: Using the Proper Orthogonal Decomposition for Improved Noise-Robustness

Abstract

Recent advances in magnetic resonance imaging (MRI) technology enabled the acquisition of time-resolved 3Dblood flow data. Several flow visualization methods have been applied to these data in order to investigate linksbetween cardiovascular diseases and hemodynamic phenomena, such as vortices in the blood flow. In this work,we investigate the use of the proper orthogonal decomposition (POD) for the preprocessing of MRI datasets andstudy its effects with the l2 vortex detection method. By comparing the POD method with the commonly usedGaussian filtering, we show that for comparable filtering strengths, the POD produces qualitatively better results.