An efficient adaptive method based on empirical wavelet transform for ultrasound tissue harmonic imaging

Abstract

In this study, an efficient adaptive method based on the empirical wavelet transform (EA_EWT) was developed to separate harmonics from ultrasonic radio frequency (RF) echo signals for B-mode imaging. The spectrum for each RF scan line was calculated, and a meaningful partition based on the scale-space representation was performed on the preprocessed spectrum, from which the harmonic boundaries were automatically determined. Thereafter, the harmonic components were obtained using the empirical scaling function and empirical wavelets. Finally, harmonic signals were obtained after separating all RF scan lines in a frame for imaging. In the experiments, the EA_EWT method was used to separate the ultrasonic RF signals simulated by CREANUIS and obtained from the human carotid artery. The results were compared with separation based on bandpass filtering (BPF), pulse inversion (PI), and complete ensemble empirical mode decomposition with adaptive noise (S_CEEMDAN) methods. The EA_EWT-based simulation results were superior to the PI-based simulation results with shifted scatterer models and comparable to the BPF- and S_CEEMDAN-based results. In the in vivo measurements, the EA_EWT-based harmonic images exhibited a mean contrast, contrast-to-noise ratio, and tissue-to-clutter ratio of 0.896, 3.973, and 8.306, respectively. The processing times of the BPF, PI, and EA_EWT methods were 0.06 s, 0.01 s, and 0.214 s, respectively, which were significantly shorter than that of the S_CEEMDAN method (113.26 s). The EA_EWT approach may be used in practice to generate more precise diagnostic data since it effectively executes a better separation of harmonic components for B-mode imaging.