Thresholding-based Wavelet Packet Methods for Doppler Ultrasound Signal Denoising

Abstract

Doppler ultrasound is widely used to diagnose vascular diseases because of its noninvasive advantages. Denoising of Doppler signals is necessary as a pre-processing for a high quality Doppler ultrasound system. Recently, there has been much work on denoising methods based on noise statistics and the spectrum distribution. Wavelet packet decomposing methods not only process low frequency components well, but also perform multi-level decomposition on high frequency contents, which is quite applicable in Doppler ultrasound signals because they have comparatively high frequency components dependent on flow velocity.

This paper presents a threshold-based wavelet packet denoising method, which preserves useful high frequency components and offers higher signal-to-noise ratio (SNR) compared with straightforward wavelet-based denoising methods. We then propose several algorithms to improve the selection of the threshold, and these methods are adaptive in the sense of coefficients obtained from different decomposed levels using the characteristics of the wavelet transform. In computer simulations, we have tested our algorithms and show improved SNR of simulated Doppler I/Q signals and better visualization of displayed Doppler spectrum.