Ultrasound Multiple Point Target Detection and Localization using Deep Learning

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

118 Downloads (Pure)

Abstract

Super-resolution imaging (SRI) can achieve subwavelength resolution by detecting and tracking intravenously injected microbubbles (MBs) over time. However, current SRI is limited by long data acquisition times since the MB detection still relies on diffraction-limited conventional ultrasound images. This limits the number of detectable MBs in a fixed time duration. In this work, we propose a deep learning-based method for detecting and localizing high-density multiple point targets from radio frequency (RF) channel data. A Convolutional Neural Network (CNN) was trained to return confidence maps given RF channel data, and the positions of point targets were estimated from the confidence maps. RF channel data for training and evaluation were simulated in Field II by placing point targets randomly in the region of interest and transmitting three steered plane waves. The trained CNN achieved a precision and recall of 0.999 and 0.960 on a simulated test dataset. The localization errors after excluding outliers were within ± 46 µm and ± 27 µm in the lateral and axial directions. A scatterer phantom was 3-D printed and imaged by the Synthetic Aperture Real-time Ultrasound System (SARUS). On measured data, a precision and recall of 0.976 and 0.998 were achieved, and the localization errors after excluding outliers were within ± 101 µm and ± 75 µm in the lateral and axial directions. We expect that this method can be extended to highly concentrated microbubble (MB) detection in order to accelerate SRI.
Original languageEnglish
Title of host publicationProceedings of 2019 IEEE International Ultrasonics Symposium
PublisherIEEE
Publication date2019
Pages1937-1940
ISBN (Electronic)978-1-7281-4596-9
DOIs
Publication statusPublished - 2019
Event2019 IEEE International Ultrasonics Symposium - SEC Glasgow, Glasgow, United Kingdom
Duration: 6 Oct 20199 Oct 2019
http://attend.ieee.org/ius-2019/

Conference

Conference2019 IEEE International Ultrasonics Symposium
LocationSEC Glasgow
CountryUnited Kingdom
CityGlasgow
Period06/10/201909/10/2019
Internet address

Cite this

@inproceedings{bddf699319bc4f61905efd2399ab8887,
title = "Ultrasound Multiple Point Target Detection and Localization using Deep Learning",
abstract = "Super-resolution imaging (SRI) can achieve subwavelength resolution by detecting and tracking intravenously injected microbubbles (MBs) over time. However, current SRI is limited by long data acquisition times since the MB detection still relies on diffraction-limited conventional ultrasound images. This limits the number of detectable MBs in a fixed time duration. In this work, we propose a deep learning-based method for detecting and localizing high-density multiple point targets from radio frequency (RF) channel data. A Convolutional Neural Network (CNN) was trained to return confidence maps given RF channel data, and the positions of point targets were estimated from the confidence maps. RF channel data for training and evaluation were simulated in Field II by placing point targets randomly in the region of interest and transmitting three steered plane waves. The trained CNN achieved a precision and recall of 0.999 and 0.960 on a simulated test dataset. The localization errors after excluding outliers were within ± 46 µm and ± 27 µm in the lateral and axial directions. A scatterer phantom was 3-D printed and imaged by the Synthetic Aperture Real-time Ultrasound System (SARUS). On measured data, a precision and recall of 0.976 and 0.998 were achieved, and the localization errors after excluding outliers were within ± 101 µm and ± 75 µm in the lateral and axial directions. We expect that this method can be extended to highly concentrated microbubble (MB) detection in order to accelerate SRI.",
author = "Jihwan Youn and Ommen, {Martin Lind} and Stuart, {Matthias Bo} and Thomsen, {Erik Vilain} and Larsen, {Niels Bent} and Jensen, {J{\o}rgen Arendt}",
year = "2019",
doi = "10.1109/ultsym.2019.8925914",
language = "English",
pages = "1937--1940",
booktitle = "Proceedings of 2019 IEEE International Ultrasonics Symposium",
publisher = "IEEE",
address = "United States",

}

Youn, J, Ommen, ML, Stuart, MB, Thomsen, EV, Larsen, NB & Jensen, JA 2019, Ultrasound Multiple Point Target Detection and Localization using Deep Learning. in Proceedings of 2019 IEEE International Ultrasonics Symposium. IEEE, pp. 1937-1940, 2019 IEEE International Ultrasonics Symposium, Glasgow, United Kingdom, 06/10/2019. https://doi.org/10.1109/ultsym.2019.8925914

Ultrasound Multiple Point Target Detection and Localization using Deep Learning. / Youn, Jihwan; Ommen, Martin Lind; Stuart, Matthias Bo; Thomsen, Erik Vilain; Larsen, Niels Bent; Jensen, Jørgen Arendt.

Proceedings of 2019 IEEE International Ultrasonics Symposium. IEEE, 2019. p. 1937-1940.

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

TY - GEN

T1 - Ultrasound Multiple Point Target Detection and Localization using Deep Learning

AU - Youn, Jihwan

AU - Ommen, Martin Lind

AU - Stuart, Matthias Bo

AU - Thomsen, Erik Vilain

AU - Larsen, Niels Bent

AU - Jensen, Jørgen Arendt

PY - 2019

Y1 - 2019

N2 - Super-resolution imaging (SRI) can achieve subwavelength resolution by detecting and tracking intravenously injected microbubbles (MBs) over time. However, current SRI is limited by long data acquisition times since the MB detection still relies on diffraction-limited conventional ultrasound images. This limits the number of detectable MBs in a fixed time duration. In this work, we propose a deep learning-based method for detecting and localizing high-density multiple point targets from radio frequency (RF) channel data. A Convolutional Neural Network (CNN) was trained to return confidence maps given RF channel data, and the positions of point targets were estimated from the confidence maps. RF channel data for training and evaluation were simulated in Field II by placing point targets randomly in the region of interest and transmitting three steered plane waves. The trained CNN achieved a precision and recall of 0.999 and 0.960 on a simulated test dataset. The localization errors after excluding outliers were within ± 46 µm and ± 27 µm in the lateral and axial directions. A scatterer phantom was 3-D printed and imaged by the Synthetic Aperture Real-time Ultrasound System (SARUS). On measured data, a precision and recall of 0.976 and 0.998 were achieved, and the localization errors after excluding outliers were within ± 101 µm and ± 75 µm in the lateral and axial directions. We expect that this method can be extended to highly concentrated microbubble (MB) detection in order to accelerate SRI.

AB - Super-resolution imaging (SRI) can achieve subwavelength resolution by detecting and tracking intravenously injected microbubbles (MBs) over time. However, current SRI is limited by long data acquisition times since the MB detection still relies on diffraction-limited conventional ultrasound images. This limits the number of detectable MBs in a fixed time duration. In this work, we propose a deep learning-based method for detecting and localizing high-density multiple point targets from radio frequency (RF) channel data. A Convolutional Neural Network (CNN) was trained to return confidence maps given RF channel data, and the positions of point targets were estimated from the confidence maps. RF channel data for training and evaluation were simulated in Field II by placing point targets randomly in the region of interest and transmitting three steered plane waves. The trained CNN achieved a precision and recall of 0.999 and 0.960 on a simulated test dataset. The localization errors after excluding outliers were within ± 46 µm and ± 27 µm in the lateral and axial directions. A scatterer phantom was 3-D printed and imaged by the Synthetic Aperture Real-time Ultrasound System (SARUS). On measured data, a precision and recall of 0.976 and 0.998 were achieved, and the localization errors after excluding outliers were within ± 101 µm and ± 75 µm in the lateral and axial directions. We expect that this method can be extended to highly concentrated microbubble (MB) detection in order to accelerate SRI.

U2 - 10.1109/ultsym.2019.8925914

DO - 10.1109/ultsym.2019.8925914

M3 - Article in proceedings

SP - 1937

EP - 1940

BT - Proceedings of 2019 IEEE International Ultrasonics Symposium

PB - IEEE

ER -