Performance Improvement of Single Plane-Wave Imaging Using U-Net and Discrete Wavelet Transform

Single Plane-Wave Imaging (SPWI), which transmits a single plane wave, can acquire ultrasound images at more than 1,000 fps, although it has poor lateral resolution and contrast. Some methods have been proposed to improve the quality of ultrasound images acquired by SPWI using deep learning, however, the quality is lower than that of compounded images, which are composed of multiple SPWI images. In addition, the RF signal is used as an input, which is computationally expensive. In this paper, we propose a method to improve the performance of SPWI using U-Net and the Discrete Wavelet Transform (DWT). The proposed method uses In-phase and Quadrature (IQ) data as the input and output of U-Net and loss functions that take into account the characteristics of the RF signal to improve the quality of images, and also uses IQ data after DWT to reduce the computational complexity and the inference time. Through a set of experiments using our ultrasound image dataset, we demonstrate the effectiveness of the proposed method.