Projects per year
Abstract
Super-resolution ultrasound imaging (SRI) can achieve a resolution beyond the spatial resolution imposed by the conventional diffraction-limited imaging systems. SRI combines many image and signal processing methods to detect and track contrast agents, i.e., microbubbles (MBs), down to the very smallest vessels. The technique can resolve tiny vessels that are impossible to image with conventional ultrasound imaging. Therefore, SRI has a big potential in medical imaging with clinical applications in diagnosing and monitoring vascular diseases, cancer, and diabetes. However, various trade-offs make SRI challenging in many scenarios. Spatial vs. temporal resolution, acquisition time vs. MB concentration, MB concentration vs. localization precision, and tissue motion vs. image resolution are only some of the known trade-offs. This Ph.D. project aimed to develop an improved processing pipeline that can cope with tissue motion and MB tracking in a complex vascular tree such as the rodent renal vasculature.
All the data for the projects were acquired with a commercial BK5000 scanner with a frame rate of 50 Hz and a linear array probe or their equivalent simulation models. An SRI processing pipeline, adapted to the BK5000 scanner, was outlined prior to this Ph.D. project, and my work involved making the pipeline more robust.
The first improvement was attained by adding a motion correction algorithm to the SRI processing pipeline. The timing between the CEUS and B-mode images was used to bind motion field and MB positions together, providing motion-corrected MB positions. The in-plane motion was estimated with a mean precision below 10 μm in the rat kidneys, resulting in an enhancement in the resolution from 90 μm before the motion correction to 55 μm after the motion correction.
Next, the SR images were improved by employing more advanced tracking methods. It was shown that the tracking performance can be improved by using Kalman and hierarchical Kalman trackers. This part of the project has been continued by upgrading the Kalman-based trackers with a forward-backward approach.
Lastly, the quantification of SR images was investigated. A primary classification and segmentation method showed the possibilities for quantifying the morphology and dynamics of the attained SR images of the rat renal vasculature. The statistically significantly different features between vein and artery tracks in the SR images were investigated to explore possibilities for automated image segmentation, and the challenging problems with quantifying the track-based vascular structures and dynamics were discussed.
All the data for the projects were acquired with a commercial BK5000 scanner with a frame rate of 50 Hz and a linear array probe or their equivalent simulation models. An SRI processing pipeline, adapted to the BK5000 scanner, was outlined prior to this Ph.D. project, and my work involved making the pipeline more robust.
The first improvement was attained by adding a motion correction algorithm to the SRI processing pipeline. The timing between the CEUS and B-mode images was used to bind motion field and MB positions together, providing motion-corrected MB positions. The in-plane motion was estimated with a mean precision below 10 μm in the rat kidneys, resulting in an enhancement in the resolution from 90 μm before the motion correction to 55 μm after the motion correction.
Next, the SR images were improved by employing more advanced tracking methods. It was shown that the tracking performance can be improved by using Kalman and hierarchical Kalman trackers. This part of the project has been continued by upgrading the Kalman-based trackers with a forward-backward approach.
Lastly, the quantification of SR images was investigated. A primary classification and segmentation method showed the possibilities for quantifying the morphology and dynamics of the attained SR images of the rat renal vasculature. The statistically significantly different features between vein and artery tracks in the SR images were investigated to explore possibilities for automated image segmentation, and the challenging problems with quantifying the track-based vascular structures and dynamics were discussed.
Original language | English |
---|
Publisher | DTU Health Technology |
---|---|
Number of pages | 246 |
Publication status | Published - 2022 |
Fingerprint
Dive into the research topics of 'Super-resolution ultrasound imaging'. Together they form a unique fingerprint.Projects
- 1 Finished
-
Super Resolution Ultrasound Imaging
Taghavi, I. (PhD Student), Cinthio, M. (Examiner), Couture, O. (Examiner), Jensen, J. A. (Main Supervisor) & Stuart, M. B. (Supervisor)
01/03/2019 → 03/08/2022
Project: PhD