Related Experiment Video
Updated: Jun 18, 2025
06:02
Multi-timescale Microscopy Methods for the Characterization of Fluorescently-labeled Microbubbles for Ultrasound-Triggered Drug Release
Published on: June 12, 2021
3.9K
In vivo ultrasound localization microscopy for high-density microbubbles
Gaobo Zhang1, Xing Hu2, Xuan Ren3
1Department of Biomedical Engineering, School of Information Science and Technology, Fudan University, Shanghai 200438, China.
Ultrasonics
|July 31, 2024
Summary
This study introduces ULM-TransUNet, a deep learning framework enhancing Ultrasound Localization Microscopy (ULM) for microvascular imaging. It improves microbubble localization in dense conditions, overcoming previous limitations.
Area of Science:
- Biomedical imaging
- Medical physics
- Deep learning applications
Background:
- Ultrasound Localization Microscopy (ULM) offers sub-wavelength resolution for microvasculature visualization using microbubbles (MBs).
- High MB density causes overlapping signals, leading to localization errors and limiting ULM to low concentrations.
- A trade-off exists between localization accuracy and MB density in current ULM techniques.
Purpose of the Study:
- To develop a deep learning framework, ULM-TransUNet, for accurate MB localization in dense conditions.
- To overcome the limitations of existing ULM methods in high-concentration microbubble environments.
- To improve the trade-off between localization efficiency and MB density for enhanced microvascular imaging.
Main Methods:
- A novel deep learning framework, ULM-TransUNet, combining Transformer and U-Net architectures was developed.
- The non-linear model was trained to learn complex data patterns of overlapping MBs.
- Performance was evaluated using numerical simulations and in vivo experiments.
Main Results:
- ULM-TransUNet demonstrated significant improvements in detection rate (+21.93%), precision (+17.36%), and sensitivity (+20.53%) compared to ULM-UNet in simulations.
- In vivo experiments yielded a high spatial resolution of 9.4 μm and rapid inference speed (26.04 ms/frame).
- The method effectively detected more small vessels and resolved closely spaced vessels.
Conclusions:
- ULM-TransUNet successfully addresses the challenge of MB localization in dense conditions.
- The framework offers enhanced microvascular imaging capabilities, particularly in high-density MB scenarios.
- This advancement has the potential to significantly improve ULM performance and applications in microvascular research.
