Improved Microbubble Tracking for Super-Resolution Ultrasound Localization Microscopy using a Bi-Directional Long...

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Fluid Dynamics

Background:

• Ultrasound Localization Microscopy (ULM) offers high-accuracy microvessel flow measurements beyond conventional ultrasound limits.
• Microbubbles (MBs) serve as crucial point targets for ULM, but their robust tracking is vital for image reconstruction.
• Current MB tracking methods face limitations in high-density distributions, fast flow, and complex dynamics.

Purpose of the Study:

• To develop and validate a deep learning-based method for robust microbubble pairing and tracking in ULM.
• To improve the accuracy and reliability of MB tracking, especially in challenging imaging scenarios.

Main Methods:

• A bi-directional long short-term memory (LSTM) neural network was employed for MB pairing and tracking.
• The method integrates multiparametric MB characteristics to enhance tracking performance.
• Validation was performed using simulated data, a flow phantom, and in vivo studies on mouse and rat brains.

Main Results:

• The deep learning approach demonstrated robust and accurate MB pairing and tracking.
• The method effectively handled challenging conditions like high MB density and complex flow dynamics.
• Successful validation across simulation, phantom, and in vivo experiments confirmed the method's efficacy.

Conclusions:

• The proposed deep learning-based MB tracking method significantly enhances ULM performance.
• This advancement facilitates more reliable and high-quality microvessel flow imaging.
• The method holds potential for improved diagnostic capabilities in various clinical applications.