arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2004.08775physics.med-phphysics.comp-ph
收藏

基于神经网络声学模拟的二维超声成像技术

Two-dimensional Ultrasound Imaging Technique based on Neural Network using Acoustic Simulation

Yoshiki Nagatani, Shigeaki Okumura, Shuqiong Wu, Tetsuya Matsuda

摘要:

二维(2D)超声成像被广泛应用于多个领域。在这项研究中,基于机器学习的二维超声成像技术的潜力。分别将线发射器放置在介质的一侧,并且将接收器阵列放置在介质的相对侧。在接收器阵列处接收的信号被认为是2D图像,并且使用卷积神经网络(CNN)将其转换为目标介质的2D几何形状。使用多达40000个模拟数据的数值模拟研究的结果表明,估计和真实的几何形状之间的良好协议。

英文摘要:

The two-dimensional (2D) ultrasound imaging is widely used in several fields. In this study, the potential of two-dimensional ultrasound imaging technique based on machine learning is shown. A line transmitter was placed at one side and a receiver array was placed at the opposite side of the medium, respectively. The received signals at the receiver array was considered as a 2D image and it was converted to a 2D geometry of the target medium using convolutional neural network (CNN). The results of the numerical simulation studies which use up to 40000 simulation data showed the good agreement between the estimated and true geometry.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑