Skip to main navigation Skip to search Skip to main content

Dual-Phase Deep Learning Motion Correction for Ultrasound Localization Microscopy

  • Haoxuan Yao*
  • , Clara Rodrigo Gonzalez
  • , Su Yan
  • , Biao Huang
  • , Jipeng Yan
  • , Joseph Hansen-Shearer
  • , Rifkat Zaydullin
  • , Qingyuan Tan
  • , Cameron A.B. Smith
  • , Mengjie Shi
  • , Thomas Else
  • , Meng Xing Tang
  • *Corresponding author for this work
  • Imperial College London
  • School of Mechatronics Engineering, Harbin Institute of Technology

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Ultrasound Localization Microscopy (ULM) holds great promise for a wide range of clinical applications. However, tissue motion can reduce localization accuracy, highlighting the need for fast and robust motion correction techniques. We propose a dual-phase deep learning motion estimation method and a simulation pipeline designed for carotid artery motion correction. Trained on simulated images and fine-tuned with in-vivo images, initial results suggest that our model outperforms conventional and baseline deep learning methods on simulated data and demonstrates fast correction on in-vivo data, indicating its potential for motion correction in ULM applications.

Original languageEnglish
Title of host publication2025 IEEE International Ultrasonics Symposium, IUS 2025
PublisherIEEE Computer Society
ISBN (Electronic)9798331523329
DOIs
StatePublished - 2025
Externally publishedYes
Event2025 IEEE International Ultrasonics Symposium, IUS 2025 - Utrecht, Netherlands
Duration: 15 Sep 202518 Sep 2025

Publication series

NameIEEE International Ultrasonics Symposium, IUS
ISSN (Print)1948-5719
ISSN (Electronic)1948-5727

Conference

Conference2025 IEEE International Ultrasonics Symposium, IUS 2025
Country/TerritoryNetherlands
CityUtrecht
Period15/09/2518/09/25

Keywords

  • ULM
  • deep learning
  • motion correction
  • simulation

Fingerprint

Dive into the research topics of 'Dual-Phase Deep Learning Motion Correction for Ultrasound Localization Microscopy'. Together they form a unique fingerprint.

Cite this