High resolution 3D localization of ultrasound sparse sources using a SAGE algorithm
Abstract
Reliable localization and characterization of cavitation activity are central to many ultrasound imaging and therapy monitoring applications. Passive acoustic mapping (PAM) can estimate cavitation distributions without prior assumptions but remains limited by the system point spread function. While advanced PAM methods improve spatial resolution, their performance may degrade for transient emissions with uncertain emission times. In this work, we propose a parametric source localization framework based on the Space-Alternating Generalized Expectation-Maximization (SAGE) algorithm to estimate source positions and emitted waveforms directly from array channel data. A greedy source extraction and localization procedure is first used to initialize the model order and source parameters, after which SAGE iteratively refines each source estimate. The proposed method was validated on 49 simulated acquisitions containing up to seven simultaneously emitting sources with emission time uncertainties of 12 microseconds, achieving a source localization precision of 0.24 mm (standard deviation of the total localization error). Experimental evaluation on proton-induced nanodroplet vaporization data acquired with an ultrasound matrix array demonstrated improved spatial resolution compared to a beamforming-based approach and sub-millimeter agreement between vaporization maps and the proton distribution in all three spatial dimensions. These results highlight the potential of parametric source localization for high-resolution proton therapy monitoring and provide a general framework for the localization of simultaneously emitting sparse ultrasound sources.
Related
- Sparse defect localization - A Periodic Empirical Operator Framework for Inverse Scattering and Structural Health Monitoring
- Reconstructing fluid velocity fields from sparse sensors using a variational quantum algorithm
- Spatiotemporal Adaptive Eigenspectrum Remapping in Transcranial Ultrasound Localization Microscopy
- Azimuthal mode decomposition Particle in Cell algorithm for cylindrical plasma sources
- Solving Sparse SDPs in Sublinear Time: A Classical Algorithm Inspired by the Quantum OR Lemma
BibTeX
@misc{heymans2026high,
title = {{High resolution 3D localization of ultrasound sparse sources using a SAGE algorithm}},
author = {Sophie V. Heymans and François Rottenberg and Tatiana Gelvez-Barrera and Adrian Basarab and Barbara Nicolas},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.07966},
eprint = {2611.07966},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.07966},
}
RIS
TY - UNPB
TI - High resolution 3D localization of ultrasound sparse sources using a SAGE algorithm
AU - Sophie V. Heymans
AU - François Rottenberg
AU - Tatiana Gelvez-Barrera
AU - Adrian Basarab
AU - Barbara Nicolas
PY - 2026
DA - 2026/11/05
PB - arXiv
DO - 10.48550/arxiv.2611.07966
UR - https://arxiv.org/abs/2611.07966
AB - Reliable localization and characterization of cavitation activity are central to many ultrasound imaging and therapy monitoring applications. Passive acoustic mapping (PAM) can estimate cavitation distributions without prior assumptions but remains limited by the system point spread function. While advanced PAM methods improve spatial resolution, their performance may degrade for transient emissions with uncertain emission times. In this work, we propose a parametric source localization framework based on the Space-Alternating Generalized Expectation-Maximization (SAGE) algorithm to estimate source positions and emitted waveforms directly from array channel data. A greedy source extraction and localization procedure is first used to initialize the model order and source parameters, after which SAGE iteratively refines each source estimate. The proposed method was validated on 49 simulated acquisitions containing up to seven simultaneously emitting sources with emission time uncertainties of 12 microseconds, achieving a source localization precision of 0.24 mm (standard deviation of the total localization error). Experimental evaluation on proton-induced nanodroplet vaporization data acquired with an ultrasound matrix array demonstrated improved spatial resolution compared to a beamforming-based approach and sub-millimeter agreement between vaporization maps and the proton distribution in all three spatial dimensions. These results highlight the potential of parametric source localization for high-resolution proton therapy monitoring and provide a general framework for the localization of simultaneously emitting sparse ultrasound sources.
ER -