Clutter Suppression In Ultrasound Color Flow Imaging System

The ultrasound color flow imaging (CFI) technology can display human blood velocity distribution on a two-dimensional plane in real-time. It now has become a major diagnostic method for detecting vascular abnormalities. Unfortunately, along with echoes scattered from red blood cells, high-energy noises reflected from tissues or vessel walls are also contained in signals received by the ultrasound transducer. These noises are called clutters. Therefore, a clutter rejection step is required before any velocity estimation.Four novel clutter rejection methods have been proposed to deal with non-stationary clutter movements. The dynamic region partition (DRP) method can adjust its effective interval size which is reversely proportional to the non-stationarity level of clutter movements. After intervals have been adjusted, the Eigenfilter is applied on each interval one by one. Thus, the contradiction encountered when using a uniform partition approach can be solved. The projection pursuit (PP) adopts a robust projection index instead of the variance to eliminate the influence caused by outliers. Then the clutter subspace is formed by these robust eigenvectors and blood signals are finally obtained by removing the reconstructed clutter components from the original signal. The recursive eigenvalue decomposition (RED) algorithm uses both the current input and its predecessors to update the clutter subspace recursively. The RED not only keeps the temporal adaptiveness inherited from the Eigenfilter, but also achieves the spatial adaptiveness at the same time. The first order perturbation (FOP) method is proposed as a fast version of the RED. It avoids the calculation of the singular value decomposition (SVD) by introducing the first order approximation theory. The FOP can reduce the computational complexity effectively and its clutter subspace updating procedure is similar to the RED.Computer simulations have been carried out using five clutter movement models. The proposed methods are compared with three conventional methods:the high-pass filter, the Eigenfilter and the Hankel-SVD. The estimated blood velocity profiles are shown and the mean square error (MSE) and the cross correlation coefficient r are also listed. Results show that when the clutter movement model is complicated, i.e., the clutter non-stationarity is severe, the proposed methods will outperform the conventional ones. Moreover, when clutter movements are spatially non- continuous, the RED and FOP will obtain even better results. In vivo experiments are also carried out by using human carotid data. The CFI output images are finally shown. Results show that the proposed methods can achieve better outcomes when considering the integrity of the vessel and the suppression of the artificial clutter velocities at the same time. Among all methods, the RED and FOP perform the best.Computational complexities of different clutter rejection algorithms are finally calculated. Results show that the FOP can achieve a lower computational complexity compared to the RED when a similar effectiveness is guaranteed. Hopefully, the FOP method will become a practical clutter rejection appraoch for the ultrasound CFI in the future.