Deep learning for fast super-resolution ultrasound microvessel imaging

• Published in: Physics in medicine & biology 2023-12, Vol.68 (24), p.245023
• Main Authors: Luan, Shunyao, Yu, Xiangyang, Lei, Shuang, Ma, Chi, Wang, Xiao, Xue, Xudong, Ding, Yi, Ma, Teng, Zhu, Benpeng
• Format: Article
• Language: English
• Subjects: deep learning; microbubble localization; super-resolution ultrasound imaging; ultrasound localization microscopy
• ISSN: 0031-9155; 1361-6560
• DOI: 10.1088/1361-6560/ad0a5a

. Ultrasound localization microscopy (ULM) enables microvascular reconstruction by localizing microbubbles (MBs). Although ULM can obtain microvascular images that are beyond the ultimate resolution of the ultrasound (US) diffraction limit, it requires long data processing time, and the imaging accuracy is susceptible to the density of MBs. Deep learning (DL)-based ULM is proposed to alleviate these limitations, which simulated MBs at low-resolution and mapped them to coordinates at high-resolution by centroid localization. However, traditional DL-based ULMs are imprecise and computationally complex. Also, the performance of DL is highly dependent on the training datasets, which are difficult to realistically simulate. . A novel architecture called adaptive matching network (AM-Net) and a dataset generation method named multi-mapping (MMP) was proposed to overcome the above challenges. The imaging performance and processing time of the AM-Net have been assessed by simulation and experiments. . Simulation results show that at high density (20 MBs/frame), when compared to other DL-based ULM, AM-Net achieves higher localization accuracy in the lateral/axial direction. experiment results show that the AM-Net can reconstruct ∼24.3 m diameter micro-vessels and separate two ∼28.3 m diameter micro-vessels. Furthermore, when processing a 128 × 128 pixels image in simulation experiments and an 896 × 1280 pixels image experiment, the processing time of AM-Net is ∼13 s and ∼33 s, respectively, which are 0.3-0.4 orders of magnitude faster than other DL-based ULM. . We proposes a promising solution for ULM with low computing costs and high imaging performance.