Localization Free Super-Resolution Microbubble Velocimetry Using a Long Short-Term Memory Neural Network.

Area of Science:

• Biomedical Imaging
• Ultrasound Technology
• Super-resolution Microscopy

Background:

• Ultrasound localization microscopy offers high resolution and deep penetration using microbubbles.
• Conventional methods struggle with high microbubble concentrations, limiting vascular imaging.
• Existing deep learning and sparsity methods do not provide blood flow velocity maps.

Purpose of the Study:

• To introduce Deep-SMV, a novel technique for super-resolution microbubble velocimetry.
• To enable accurate blood flow velocity mapping at high microbubble concentrations.
• To achieve high imaging speed and robustness for functional vascular imaging.

Main Methods:

• Developed Deep-SMV, a localization-free technique using a long short-term memory neural network.
• Trained the model using microbubble flow simulations and in vivo vascular data.
• Demonstrated real-time velocity map reconstruction.

Main Results:

• Deep-SMV achieves super-resolution blood velocity measurements.
• The technique is robust to high microbubble concentrations and offers high imaging speed.
• Successfully applied to flow phantoms, chicken embryos, and mouse brains.

Conclusions:

• Deep-SMV provides a powerful tool for super-resolution microvascular velocimetry.
• Enables real-time functional vascular imaging and pulsatility mapping.
• The open-source implementation facilitates further research and application.