Denoising Ultrasound RF Signals by Wavelet Cycle Spinning Shrinkage

Abstract

A speckle reduction procedure, based on wavelet Cycle Spinning (CS) Shrinkage, has been applied to radio-frequency RF pulse-echo signals. Several sets of 100 synthetic ultrasonic signals were generated using a model of speckle formation which includes frequency dependent attenuation, frequency dependent scattering, and an accurate model for the pulse-echo response of the broad band ultrasonic transducer. The frequency spectra overlapping of ultrasound signal and speckle noise is shown. The mean and standard deviation of the signal to noise ratio (SNR) of sets of 100 RF signals have been taken as denoising performance indexes. Results of CS denoising have been compared with those of Discrete Wavelet Transform (DWT) denoising, using Universal and SURE decomposition level dependent threshold selection rules. CS overcomes the lack of translation invariance of DWT providing on average a better denoising performance. A preliminary evaluation of the influence of different mother wavelets is also presented.