Contrast-Free Super-Resolution Power Doppler (CS-PD) Based on Deep Neural Networks

Area of Science:

• Medical Imaging
• Ultrasound Technology
• Biomedical Engineering

Background:

• Ultrasound localization microscopy (ULM) provides super-resolution microvessel imaging deep within tissues.
• Current ULM methods require contrast agents, lengthy data acquisition, and intensive postprocessing.

Purpose of the Study:

• To introduce a contrast-free super-resolution power Doppler (CS-PD) technique for rapid microvessel imaging.
• To evaluate the performance and generalizability of CS-PD across different tissues and imaging conditions.

Main Methods:

• Developed a deep network-based CS-PD technique utilizing ultrafast ultrasound signals.
• Trained the network on in vivo mouse brain data and tested on mouse brain, chicken embryo CAM, and human subjects.

Main Results:

• CS-PD achieved approximately twofold spatial resolution improvement over conventional power Doppler in mouse models.
• Generated microvascular images showed high agreement with ULM, with a structural similarity index of 0.7837 and PSNR of 25.52.
• Preserved temporal blood flow characteristics and demonstrated generalizability across diverse datasets.

Conclusions:

• CS-PD offers a practical, fast, and robust contrast-free solution for microvessel imaging.
• The technique's rapid inference enables real-time imaging applications.
• CS-PD holds significant potential for preclinical and clinical Doppler ultrasound applications.