A lasso based ensemble empirical mode decomposition approach to designing adaptive clutter suppression filters

Accurate estimation of the blood flow velocity in ultrasound imaging is an important tool for medical diagnostics. In this paper, we adopt an improved empirical mode decomposition (EMD) framework called ensemble EMD (EEMD). To reduce the errors caused by the outliers in data when using a uniform weight in conventional EEMD, a regularized LASSO EEMD algorithm is proposed to solve for the multiple regression weights. An adaptive clutter rejection filter can then be designed to remove the clutter components. According to our simulation study, the proposed LASSO EEMD approach performs better than the state-of-the-art eigen-based and EMD method in estimating the blood flow velocity. Although the LASSO EEMD derived filter only achieves slightly better results than the cubic regression derived filters at most part of the simulated blood flow center frequencies, the proposed LASSO EEMD algorithm achieves much improved performance over cubic regression at extreme cases when the blood flow center frequency is close to or much higher than that of the clutter.