Super-resolution ultrasound imaging using a lensed row-column array

Ultrasound imaging provides a noninvasive and accessible form of medical imaging, but has been fundamentally limited by both the diffraction limit and 2D images. Recently, super-resolution ultrasound imaging (SRUI) has surpassed the diffraction limit, offering macroscopic resolution. Similarly, volumetric imaging with row-column arrays (RCAs) can be used to make 3D images, and has the potential for clinical application due to the low number of connecting channels in the transducer. Designing an RCA using capacitive micromachined ultrasonic transducer (CMUT) technology offer several advantageous features for SRUI, such as wide bandwidth and narrow inter-element pitch. However, the usable field of view (FOV) of an RCA is limited, and the inclusion of an acoustic lens is therefore useful. It is hypothesized that if combined, these four modern ultrasound techniques has the potential to create a practical tool for imaging microvascular structures in whole organs, with microscopic resolution and macroscopic FOVs. However, several developmental steps remain before this proposal can be tested and applied.
Firstly, the minor sequences used for contrast enhanced ultrasound (CEUS) imaging in SRUI are studied theoretically. It is shown algebraically that pulse inversion (PI) and amplitude modulation (AM) can be used to eliminate linear back-scattering while preserving the nonlinear back-scattering from microbubbles. Furthermore, by applying the Marmottant et al. (2005) model governing nonlinear scattering from microbubbles, examples of the back-scattering from the contrast agent SonoVue are modeled and used to validate the efficiency of the two sequences. This theoretical validation method is explored in order to also theoretically validate the use of CMUTs for the CEUS. When CMUTs emit sound, the harmonic content in the sound is nonlinearly dependent on the applied voltage to the transducer. It is found that CMUTs can be used for CEUS imaging without loss of image quality, if a three-pulse AM sequence is applied. In this thesis, this is validated theoretically. Moreover, measurements of the contrast-to-tissue ratio (CTR) of images of microbubble contrast agents acquired using a CMUT and a comparable lead zirconate titanate (PZT) array show that the average enhancement of contrast, compared with B-mode images, was 37.4 dB for the CMUT and 49.9 dB for the PZT array. The discrepancy is attributed to the CMUT achieving a poorer signal-to-noise ratio.
Secondly, the use of a RCA for 3D SRUI is demonstrated. In preparation, the beamformation of signals created by an RCA is studied and implemented, and examples of the point-spread-function of an RCA emitting a single-element synthetic aperture sequence are simulated. Then, an experimental set-up using a 3D printed microflow phantom is designed and used to collect images of flowing SonoVue. The images are collected using a 62 + 62 element PZT RCA. From these, ultrasound localization microscopy is performed. The microbubble localization precision is found to be 15.4 μm and 16.0 μm. Whether this fulfills the true definition of super-resolution is discussed, but, regardless, the resulting 3D imaging indicates that RCAs can be used for SRUI.
Lastly, the transmission of sound from an RCA through a lens is studied. When the sound is emitted and received, its paths are refracted by the lens, and this makes the prediction of the time of flight (TOF) complicated. The TOF is needed to beamform images from the received signals, and errors in the prediction lead to lower image quality. A simplification of the problem is offered by the thin lens model, which assumes that the lens is infinitely thin and has a single focal point. An RCA with an infinitely thin lens is simulated by making the elements of the transducer curved. The thin lens model is used to predict the TOF of emissions from the transducer, and comparison with simulations of the emitted field indicate that no phase errors are introduced by the model. However, real lenses have a finite thickness and the simulations do not validate that the model is applicable to real lenses, due to the underlying assumptions in the model. Therefore, a ray tracing model is presented, which does not make simplifying assumptions about the geometry of the lens. How accurately the thin lens model and the ray tracing model can predict the TOF of an emitted wave is quantified by comparison of the predictions with the emitted field simulated using finite element method modeling of a lens with finite thickness. The accuracy of the prediction affects the usable FOV, which for four example cases is found to be higher or equal to the FOV when the TOF is predicted by the ray tracing model. For the case with the biggest discrepancy, the ray tracing model increased the FOV by 25.1°. Furthermore, the ray tracing model is validated by comparison of its predicted TOF with measurements of the emitted field from a 128 + 128 element RCA through a concave silicone lens. This resulted in a maximum phase difference of 0.19 λ, despite errors in the alignment of the hydrophone. Lastly, the usable transmitted FOV of the 128 + 128 element RCA is quantified using the ray
tracing model. The FOV is found to be 37.2 cm × 12.1 cm at 9 cm to 10 cm depth, which is large enough to image a whole human kidney. Thus, it is indicated that the lensed array has the capability of making 3D SRUI of whole organs.