Ultrafast ultrasound for super­resolution imaging

Ultrafast ultrasound for super­resolution imaging: The effect of pulse inversion and harmonic­ or SVD­filtering on microbubble detection

ter Wee, Annemiek (TU Delft Mechanical, Maritime and Materials Engineering)

Deckers, Roel (mentor)
Vos, Rik (graduation committee)

Delft University of Technology

Biomedical Engineering | Medical Physics

2021-03-29

Background/purpose: The mapping of microvasculature requires a high­resolution imaging method. Super­resolution ultrasound­imaging can provide such high resolution by detecting individual microbub­ bles (MBs) in a flow circuit. This method relies on the dynamic appearance of MBs, which appear as point scatterers on the US image, mixed with tissue signal. The detection of these MBs require good spatio­temporal resolution and suppression of the tissue signal. These requirements are met by using high­frame­rate plane or diverging wave imaging combined with tissue suppression methods. In this project, the effect of pulse inversion (PI) and harmonic filtering or singular value decomposition (SVD) filtering on the MB detection in slow flow was evaluated. This effect was measured in a soft­tissue­ mimicking phantom containing a 400 휇m diameter wall­less channel. Methods: The US acquisition was done with a Verasonics V1 system, using a convex (C5­2) probe. Two sequences for coherent compound diverging­wave imaging were designed: the first for imaging in the fundamental (F) mode and the second for PI. The axial and lateral resolution were measured and the localization precision determined. Next, the effect of changing the number of compound angles, voltage, flow velocity, the receive mode (fundamental or harmonic), SVD­filtering and tissue motion on the MB detections was evaluated in four experiments. Results: The results of the system characterization showed an average lateral & axial resolution in F­ mode of 2.5 mm and 1 mm, respectively. In PI­mode the lateral & axial resolution was higher because of imaging at higher frequencies (2.0 mm and 0.8 mm, respectively). The MB localization precision was 20 휇m lateral and 6 휇m axial, from which we deduce that ultrasound localization microscopy improves spatial resolution with an average factor of 75. The results of the MB detection experiments showed a decrease of MB detections when the number of compounding angles increased. Possibly because of less false detections, since compound imaging results in improvement of contrast and a reduction in speckle and artifacts. No direct relation between the increase of transmit voltage and MB detection was found. In the SVD­filtered PI­mode data, the increase of flow velocity from 1 till 2 till 4 mm/s was accompanied by an increase in MB detections and a decrease in the false positive detections. At low flow velocities, the difference in spatial coherence between tissue and MBs was low which could have resulted in poor differentiation. Receiving in the harmonic­mode, resulted in a small spatial shift of the location of the MB detections because of asymmetry of the used radio­frequency filter. Besides, in the harmonic receive­mode the total number of MB detections dropped. A probable explanation is that less false MB detection were made because of suppression of flash artifacts. When tissue mo­ tion was induced, this led to higher MB detections in the non­filtered PI mode than in the stationary case. However, these can also be motion artifacts which are falsely detected as MBs. SVD­filtering or harmonic­filtering of these acquisitions resulted in practically zero MB detections. Conclusion: We conclude that non­filtered PI results in more true­positive MB detections than harmonic­ filtering and SVD­filtering when flow velocities are low. When tissue motion is negligible, PI­imaging results in good SR images, with a possible disturbance of flash artefacts.

Ultrafast ultrasound
super-resolution
micro-bubbles
Filtering

http://resolver.tudelft.nl/uuid:5663ad01-92b2-40a6-9b24-f3b6a5f0216a

52.0866, 5.1799

Student theses

master thesis

© 2021 Annemiek ter Wee