Contrast-free super-resolution ultrasound imaging based on two-stage diffusion probabilistic models

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Ultrasound Technology

Background:

• Super-resolution ultrasound (SR-US) imaging, or ultrasound localization microscopy (ULM), enhances spatial resolution for microvessel visualization.
• Clinical use of SR-US is limited by the need for ultrasound contrast agents (UCAs).
• Existing contrast-free SR-US methods using red blood cells (RBCs) show limited resolution improvements.

Purpose of the Study:

• To develop a contrast-free SR-US imaging framework to overcome resolution limitations.
• To introduce the two-stage diffusion probabilistic model termed 2S-DUS.
• To enable high-fidelity microvascular imaging without UCAs.

Main Methods:

• Implemented a two-stage diffusion probabilistic model (2S-DUS).
• Stage one uses a diffusion model for low-resolution power Doppler (PD) to super-resolution map transformation.
• Stage two employs a physics-guided module for refining vascular continuity and anatomical detail.

Main Results:

• Achieved a spatial resolution of approximately 24.6 μm (FWHM) in vivo on rat brain datasets.
• Demonstrated superior performance in perceptual and resolution metrics compared to existing methods (RS, ULM-GAN, ULM-MbCNRT).
• Ablation studies confirmed the two-stage architecture's effectiveness in balancing realism and fidelity.

Conclusions:

• The proposed 2S-DUS framework effectively realizes contrast-free SR-US imaging.
• This technique offers significant potential for clinical applications in monitoring and diagnosing cerebrovascular diseases.
• 2S-DUS advances the field by providing high-resolution imaging without the need for contrast agents.