In vivo investigation of filter order influence in eigen-based clutter filtering for color flow imaging

Abstract

Eigen-based adaptive filters have shown potential for providing a superior attenuation of clutter in color flow imaging. Critical for the success of this technique is the correct selection of filter order. In this work we review and compare filter order selection schemes for eigen-based filters in an in vivo context. Data was acquired from a thyroid tumor (PRF = 250 Hz, ensemble size = 12), where substantial tissue movement was present due to carotid artery pulsations, respiratory movements, and probe navigation. Eigen-filtering performance was evaluated for 1) an eigenvalue spectrum threshold, 2) a threshold on the ratio of successive eigenvalues, and 3) a threshold on eigenvector mean frequency estimated by the autocorrelation approach. Based on the observed eigenvalue and eigen-frequency distributions in analytical and in vivo examples, all filter order algorithms investigated suffered from potential pitfalls in specific Doppler scenarios. In the in vivo examples, the fixed order eigenfilter gave a sufficient suppression of clutter, but also removed substantial blood signal. Thresholding the ratio of eigenvalues better retained signal from blood, but also spurious artifacts was observed. The most consistent results were achieved by thresholding the mean frequency of the eigenvectors. The results demonstrate that given a suitable filter order algorithm, robust filtering can be achieved with the eigen-based approach. © 2007 IEEE.