| Summary: | In contrast enhanced ultrasound images (CEUS) that use microbubbles, nonlinear propagation of ultrasound creates artefacts which significantly impact the qualitative and quantitative assessments of tissue perfusion. Such artefact originates from tissue reflecting/scattering nonlinearly propagated ultrasound pulse. Consequently such tissue is misclassified as microbubbles which also generate nonlinear signals. This thesis reports the development and evaluation of an algorithm to reduce the nonlinear propagation artefact in CEUS. The method was evaluated in simulations, and on in vitro and in vivo data at both high and low ultrasound frequencies. Ways to further improve the performance of the method were also investigated. Firstly, the artefact correction algorithm was developed. The algorithm makes use of two independent datasets that are acquired simultaneously during CEUS; the Bmode image, which is dominated by tissue information, and the contrast specific image, which contains information on blood (signal) due to microbubbles, but confounded with some amount of tissue signals (artefact). The unwanted tissue contribution of the contrast specific image is reconstructed by estimating the two components that make up this contribution, namely, the underlying tissue distribution and the nonlinear point spread function (PSF) of the imaging system. To initially evaluate the algorithm, a simulation platform was developed to study artefact generation at various Mechanical Indices (MI), microbubble concentrations and frequencies. The algorithm was then evaluated using the simulation data. The results show that the algorithm is able to reduce the nonlinear propagation artefact at different MI, concentration and frequency under both ideal and noisy conditions. Next, artefact correction was applied to carotid artery imaging. The performance of the algorithm was evaluated using flow phantoms with large and small vessels containing microbubbles of various concentrations at different acoustic pressures. The algorithm significantly reduces nonlinear artefacts while maintaining the contrast signal from bubbles to increase the contrast to tissue ratio (CTR) by up to 11 dB. Contrast signal from a small vessel of 600 µm in diameter buried in tissue artefacts prior to correction is recovered after the correction. The algorithm was then evaluated using in vivo CEUS data acquired on patients’ carotid arteries. The algorithm is able to increase the CTR at the far-wall by up to 7.4 dB in vivo. Artefact correction was then improved by taking the spatial variance of the ultrasound field into account and improving the nonlinear PSF estimation. The new version of the algorithm was tested on in vitro and in vivo data and the improvements verified. The new version of the algorithm provides an additional increase in CTR by up to 5.4 dB in the far field, 4.3 dB at focus and 3.2 dB in the near field for the in vitro data over previous results. The additional increase in CTR for the in vivo data is up to 4 dB more in the near field and 5 dB more in the far field over the previous results. Nonlinear propagation correction was also applied to deep tissue imaging where lower ultrasound frequency than carotid imaging was used. The algorithm could suppress tissue signal more than 6 dB. However, due to the strong presence of the microbubbles in the B-mode image at low frequencies, the algorithm reduces microbubble signal by up to 2 dB. The resulting increase in CTR is up to 4 dB under specific imaging conditions. However, depending on the imaging geometry and acquisition settings used, it could fail to produce an increase in CTR. A possible future direction is to combine the algorithm with an attenuation correction method to improve perfusion quantification. The clinical efficacy of the combined nonlinear propagation and attenuation correction could be evaluated. Given that the method is purely post-processing, it is easier to implement it on current commercial scanners than some other existing techniques. The implementation of the algorithm using GPUs could be investigated and could possibly instigate translation into clinics.
|
|---|
