Sparisty Performance Of Endoscopic Ultrasound Signal Under Different Sparse Transforms

摘要

This paper introduced the compressed sensing (CS) technique into endoscopic ultrasound (EUS). An In-vitro tissue experiment of swine digestive was designed to obtain high-quality EUS signals. Two kinds of EUS signals were obtained through traditional sinusoidal excitation and chirp coded excitation. Four types of sparse bases - discrete cosine transform (DCT), discrete wavelet transform (DWT), Hadamard-Walsh transform (HW), and wave atom transform (WA) - were used to compare the sparsity performance for these signals. And the reconstruction error under different sparse transform methods with different proportions of data was simulated and analyzed. The results show that for the sparsity performance, WA realized the highest sparsity of the two EUS signals. In terms of reconstruction error, WA also performs best in most cases.