Fast 3D Ultrasound Localization Microscopy via Projection-based Processing Framework
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866909926950961152 |
|---|---|
| author | Zhang, Jingke Yin, Jingyi Lok, U-Wai Huang, Lijie DeRuiter, Ryan M. Wu, Tao Ji, Kaipeng Zhao, Yanzhe Krier, James D. Zhu, Xiang-yang Lerman, Lilach O. Huang, Chengwu Chen, Shigao |
| author_facet | Zhang, Jingke Yin, Jingyi Lok, U-Wai Huang, Lijie DeRuiter, Ryan M. Wu, Tao Ji, Kaipeng Zhao, Yanzhe Krier, James D. Zhu, Xiang-yang Lerman, Lilach O. Huang, Chengwu Chen, Shigao |
| contents | Three-dimensional ultrasound localization microscopy (ULM) enables comprehensive visualization of the vasculature, thereby improving diagnostic reliability. Nevertheless, its clinical translation remains challenging, as the exponential growth in voxel count for full 3D reconstruction imposes heavy computational demands and extensive post-processing time. In this row-column array (RCA)-based 3D in vivo pig kidney ULM study, we reformulate each step of the full 3D ULM pipeline, including beamforming, clutter filtering, motion estimation, microbubble separation and localization into a series of computational-efficient 2D operations, substantially reducing the number of voxels to be processed while maintaining comparable accuracy. The proposed framework reconstructs each 0.75-s ensemble acquired at frame rate of 400 Hz, covering a 25*27.4*27.4 mm3 volume, in 0.52 s (70% of the acquisition time) on a single RTX A6000 Ada GPU, while maintaining ULM image quality comparable to conventional 3D processing. Quantitatively, it achieves a structural similarity index (SSIM) of 0.93 between density maps and a voxel-wise velocity agreement with slope of 0.93 and R2 = 0.88, closely matching conventional 3D results, and for the first time, demonstrating potential for real-time feedback during scanning, which could improve robustness, reduce operator dependence and accelerate clinical workflows. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_21647 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Fast 3D Ultrasound Localization Microscopy via Projection-based Processing Framework Zhang, Jingke Yin, Jingyi Lok, U-Wai Huang, Lijie DeRuiter, Ryan M. Wu, Tao Ji, Kaipeng Zhao, Yanzhe Krier, James D. Zhu, Xiang-yang Lerman, Lilach O. Huang, Chengwu Chen, Shigao Medical Physics Three-dimensional ultrasound localization microscopy (ULM) enables comprehensive visualization of the vasculature, thereby improving diagnostic reliability. Nevertheless, its clinical translation remains challenging, as the exponential growth in voxel count for full 3D reconstruction imposes heavy computational demands and extensive post-processing time. In this row-column array (RCA)-based 3D in vivo pig kidney ULM study, we reformulate each step of the full 3D ULM pipeline, including beamforming, clutter filtering, motion estimation, microbubble separation and localization into a series of computational-efficient 2D operations, substantially reducing the number of voxels to be processed while maintaining comparable accuracy. The proposed framework reconstructs each 0.75-s ensemble acquired at frame rate of 400 Hz, covering a 25*27.4*27.4 mm3 volume, in 0.52 s (70% of the acquisition time) on a single RTX A6000 Ada GPU, while maintaining ULM image quality comparable to conventional 3D processing. Quantitatively, it achieves a structural similarity index (SSIM) of 0.93 between density maps and a voxel-wise velocity agreement with slope of 0.93 and R2 = 0.88, closely matching conventional 3D results, and for the first time, demonstrating potential for real-time feedback during scanning, which could improve robustness, reduce operator dependence and accelerate clinical workflows. |
| title | Fast 3D Ultrasound Localization Microscopy via Projection-based Processing Framework |
| topic | Medical Physics |
| url | https://arxiv.org/abs/2511.21647 |