Automated Platform Converts Super-Resolution Ultrasound into Vascular Biomarkers

Super-resolution ultrasound localization microscopy (ULM) can visualize microvessels several centimeters below the surface, but heterogeneous processing impedes consistent measurement. Standardized, quantitative biomarkers are needed to support diagnosis, monitoring, and trials. To help address this challenge, researchers have developed U‑VBA, a pipeline that standardizes ULM analysis for non-invasive tracking of early microvascular change.

U‑VBA (ULM‑based Vascular Biomarker Automated Analysis) is a ready‑to‑use framework that links super‑resolution ultrasound reconstruction to automated vascular biomarker quantification. Outlined in a study published in BME Frontiers, the platform incorporates motion correction and cascaded denoising to improve signal fidelity. It then computes six complementary biomarkers—vascular density, intervessel distance, diameter, flow velocity, perfusion, and tortuosity—capturing structural and hemodynamic properties needed for consistent microvascular assessment.

The platform was validated across phantoms, animal models, and patient samples. In chicken chorioallantoic membrane testing, it resolved capillaries as small as 16.2 micrometers, demonstrating high spatial precision for superficial microvessels. In a rabbit optic‑nerve‑injury glaucoma model, longitudinal imaging captured dynamic vascular remodeling during intraocular‑pressure elevation and partial recovery, with 13 biomarkers showing significant post‑injury alterations consistent with retinal neural damage seen on histology.

Analytical reliability and image quality were emphasized. U‑VBA showed high reproducibility with an intra‑class correlation coefficient of 0.981 across repeat analyses, indicating robust metric stability. Compared with a benchmark algorithm, it produced cleaner microvascular maps with minimal background noise, supporting clearer interpretation of vessel architecture.

Clinical feasibility was explored in 39 patients with cervical lymph‑node disease spanning benign lesions, lymphoma, and metastatic carcinoma. Four biomarkers differed significantly among groups, and a support‑vector‑machine classifier built from these multiparametric features achieved 85% accuracy in five‑fold cross‑validation. The authors note compatibility with commercial clinical ultrasound systems and alignment with existing workflows, underscoring potential to reduce reliance on invasive biopsies where appropriate.

Study limitations include small animal sample sizes, two‑dimensional imaging, and dependence on contrast agents, with plans for future three‑dimensional and contrast‑free implementations to broaden applicability.