GraphULM: A Multi-Resolution CNN and GCN Framework for Ultrasound Localization Microscopy

Mohammad Sabih, Mohamed Khaled Almekkawy
Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, PMLR 315:3931-3946, 2026.

Abstract

Ultrasound Localization Microscopy (ULM) is a prominent technique in medical imaging, widely applied to enhance super-resolution, particularly in in-vivo settings. The process of localization, followed by tracking of microbubble (MB), poses a significant challenge in ULM due to its intricacy and complexity. High MB densities intensify these challenges, thereby diminishing the performance of traditional methods and certain deep learning algorithms in achieving precise localization. We present GraphULM, a novel and computationally efficient architecture that combines a Multi-Resolution Convolutional Neural Network (MRCNN) with a Graph Convolutional Network (GCN) to enhance localization efficacy in ULM. To develop an optimal training dataset, synthetically generated data is pre-combined with in-vivo b-mode samples, which improves feature diversity and generalization. Experimental evaluations in in-vivo demonstrate the model’s high performance, reporting a localization precision of 21.9 m, and a Jaccard index of 0.75, at a MB density of 2 MB/mm2, underscoring the model’s robustness. Additionally, our Frequency Ring Correlation (FRC) analysis reveals a remarkable resolution of 5.62 m. The model operates at three times the speed of traditional pipelines, establishing its suitability for rapid ULM applications.

Cite this Paper

BibTeX
@InProceedings{pmlr-v315-sabih26a, title = {GraphULM: A Multi-Resolution CNN and GCN Framework for Ultrasound Localization Microscopy}, author = {Sabih, Mohammad and Almekkawy, Mohamed Khaled}, booktitle = {Proceedings of The 9th International Conference on Medical Imaging with Deep Learning}, pages = {3931--3946}, year = {2026}, editor = {Huo, Yuankai and Gao, Mingchen and Kuo, Chang-Fu and Jin, Yueming and Deng, Ruining}, volume = {315}, series = {Proceedings of Machine Learning Research}, month = {08--10 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v315/main/assets/sabih26a/sabih26a.pdf}, url = {https://proceedings.mlr.press/v315/sabih26a.html}, abstract = {Ultrasound Localization Microscopy (ULM) is a prominent technique in medical imaging, widely applied to enhance super-resolution, particularly in in-vivo settings. The process of localization, followed by tracking of microbubble (MB), poses a significant challenge in ULM due to its intricacy and complexity. High MB densities intensify these challenges, thereby diminishing the performance of traditional methods and certain deep learning algorithms in achieving precise localization. We present GraphULM, a novel and computationally efficient architecture that combines a Multi-Resolution Convolutional Neural Network (MRCNN) with a Graph Convolutional Network (GCN) to enhance localization efficacy in ULM. To develop an optimal training dataset, synthetically generated data is pre-combined with in-vivo b-mode samples, which improves feature diversity and generalization. Experimental evaluations in in-vivo demonstrate the model’s high performance, reporting a localization precision of 21.9 m, and a Jaccard index of 0.75, at a MB density of 2 MB/mm2, underscoring the model’s robustness. Additionally, our Frequency Ring Correlation (FRC) analysis reveals a remarkable resolution of 5.62 m. The model operates at three times the speed of traditional pipelines, establishing its suitability for rapid ULM applications.} }
Endnote
%0 Conference Paper %T GraphULM: A Multi-Resolution CNN and GCN Framework for Ultrasound Localization Microscopy %A Mohammad Sabih %A Mohamed Khaled Almekkawy %B Proceedings of The 9th International Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2026 %E Yuankai Huo %E Mingchen Gao %E Chang-Fu Kuo %E Yueming Jin %E Ruining Deng %F pmlr-v315-sabih26a %I PMLR %P 3931--3946 %U https://proceedings.mlr.press/v315/sabih26a.html %V 315 %X Ultrasound Localization Microscopy (ULM) is a prominent technique in medical imaging, widely applied to enhance super-resolution, particularly in in-vivo settings. The process of localization, followed by tracking of microbubble (MB), poses a significant challenge in ULM due to its intricacy and complexity. High MB densities intensify these challenges, thereby diminishing the performance of traditional methods and certain deep learning algorithms in achieving precise localization. We present GraphULM, a novel and computationally efficient architecture that combines a Multi-Resolution Convolutional Neural Network (MRCNN) with a Graph Convolutional Network (GCN) to enhance localization efficacy in ULM. To develop an optimal training dataset, synthetically generated data is pre-combined with in-vivo b-mode samples, which improves feature diversity and generalization. Experimental evaluations in in-vivo demonstrate the model’s high performance, reporting a localization precision of 21.9 m, and a Jaccard index of 0.75, at a MB density of 2 MB/mm2, underscoring the model’s robustness. Additionally, our Frequency Ring Correlation (FRC) analysis reveals a remarkable resolution of 5.62 m. The model operates at three times the speed of traditional pipelines, establishing its suitability for rapid ULM applications.
APA
Sabih, M. & Almekkawy, M.K.. (2026). GraphULM: A Multi-Resolution CNN and GCN Framework for Ultrasound Localization Microscopy. Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 315:3931-3946 Available from https://proceedings.mlr.press/v315/sabih26a.html.

Related Material