P728Comparison of different optical flow algorithms for velocity filed characterization; results obtained with simulated Echopiv sequences.


    loading  Checking for direct PDF access through Ovid

Abstract

Echopiv has been introduced as a new ultrasound methodology to measure two dimensional intra-ventricular flow patterns. We have recently validated against optical PIV, our optical flow algorithm approach for dynamic sequences analysis based on: (1) image decomposition and regularization, (2) optical flow movement detection. The objective of this study was to compare an exhaustive panel of optical flow methods according to the state of the art of optical flow estimation using simulated Echopiv sequences.Methods: A dynamic 3D anatomical model of the LV was combined with a realistic inflow velocity profile to build a CFD flow vector. Base on these data motion of injected bubbles was simulated. Two frame rates (113 and 227 i/s) and two different contrast densities were used. Four classical optic flow algorithms were applied to theses simulated EchoPIV sequences (Classical approaches: Cross Correlation, Horn and Schunck, Lucas and Kanade, and the method developed by Sun) and four different version of our algorithm based on Total Variation (TV) as regularizing term (TVL1) and its extension to spatio-temporal aspects (ST+) For comparison against truth velocity field, we used Average angular error (AVE) and average endpoint error(EPE) variables witch integrate the magnitude and the angle error estimation. On table 1, are represented the average end point error obtained with the different algorithms for the entire flow field. Best results are obtained with our method integrating the spatio temporal aspect of the sequence, with less dependency to frame rate.

    loading  Loading Related Articles