SMOS: A streamlined model-based optimization approach for super-resolution ultrasound localization microscopy
Résumé
Ultrasound Localization Microscopy (ULM) represents a significant advancement over conventional ultrasound (US) imaging, for reconstructing microvascular structures with micrometer-scale precision. However, existing model-based ULM approaches often rely on a complex five-step processing framework, which complicates parameter tuning and tissue signal suppression. Additionally, many of these methods either insufficiently account for the point spread function (PSF) - thereby compromising the accuracy of the final reconstruction - or are computationally intensive. While end-to-end data-driven methods show promise, they often require substantial training data and reliable ground-truth annotations. In this study, we introduce a novel approach named Streamlined Model-Based Optimization Approach for Super-Resolution Ultrasound Localization Microscopy (SMOS) which addresses these limitations. Our method improved the existing ULM process by replacing the first three steps with a single unified model. Experimental results on both in silico and in vivo datasets demonstrate the enhanced performance of our approach compared to existing benchmarks.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |