Ultrasound imaging is a cornerstone of medical diagnostics, offering high-resolution, real-time visualization of anatomical structures. However, its application to molecular and cellular imaging has been limited by the lack of nanoscale contrast agents. Gas vesicles (GVs), air-filled protein nanostructures evolved for buoyancy in microorganisms, offer transformative potential in this domain. This thesis explores the engineering and application of GVs as biomolecular ultrasound contrast agents, emphasizing their use as biosensors for molecular imaging. It builds on their unique genetic encodability, tunable acoustic properties, and nanoscale dimensions to address key limitations in ultrasound imaging.

In chapter 1 and chapter 2 we introduce the potential of GVs as genetically encoded ultrasound contrast agents, highlighting their advantages over traditional agents like microbubbles. These chapters provide an overview of ultrasound imaging’s evolution, focusing on the emerging field of biomolecular ultrasound imaging, which leverages GVs to bridge the gap between molecular processes and ultrasound modalities. Applications such as neuroscience imaging and functional imaging of dynamic biological processes are discussed, emphasizing GVs’ nanoscale properties and unique acoustic behavior.

The contents of chapter 3 focus on the cryo-electron microscopy (cryo-EM) structural analysis of GVs. This study provides an atomic-level model of the GV shell, particularly in the absence of the reinforcement protein GvpC. By combining structural insights with a sequence analysis of GvpC, the chapter proposes a hypothetical binding mechanism that informs mutagenesis experiments in later work. This structural foundation is critical for the subsequent engineering of GVs for biosensor applications.

In chapter 4 we present the development and validation of pHonon, the first GV-based pH biosensor. By engineering pH-sensitive histidine residues into GvpC, the biosensor’s acoustic properties were tuned to detect pH variations. Validation experiments, conducted both in vitro and in vivo, demonstrated pHonon’s efficacy for real-time, non-invasive pH imaging. This work highlights the versatility of GVs as platforms for biosensor engineering and their potential for applications in both basic research and clinical diagnostics.

Then, chapter 5 explores alternative approaches to enhancing GV functionality through aggregation. By inducing GV clustering via methods like biotin-streptavidin interactions and depletion interactions, significant improvements in ultrasound contrast were achieved. This chapter shows that aggregation enhances both linear and non-linear acoustic responses, providing a complementary strategy to genetic engineering for optimizing GV performance. These findings open new avenues for improving the signal strength and utility of GVs in various imaging applications.

The thesis concludes by summarizing the key findings and their implications for the field of biomolecular ultrasound imaging. It emphasizes the breakthroughs achieved, such as the high-resolution GV structural model, the development of pHonon, and the exploration of aggregation-based contrast enhancement. These contributions advance the field significantly, offering innovative solutions to challenges in molecular imaging. This thesis lays the basis for future research on broadening the array of biomarkers identifiable by GVs, improving genetic engineering methods, and investigating additional imaging techniques to enhance the effectiveness of biomolecular ultrasound imaging.

Ultrasound imaging is a widespread clinical tool, known best for prenatal examinations of developing human embryos. Recently, a technological breakthrough has revolutionized the field of ultrasound imaging by enabling imaging at thousands of frames per second. This increase in temporal resolution has opened the door to many new applications, such as monitoring the subtle motion of heart walls, measuring electromechanical waves in muscles and detecting the stiffness of organs. Furthermore, the fast frame rates have significantly improved ultrasound sensitivity to small vessels and enables monitoring of local changes in blood flow. This has let to the development of functional ultrasound imaging (fUS) in 2011.

Functional ultrasound is a new neuroimaging modality that allows imaging of brain function at high spatial and temporal resolution. fUS measures variations in cerebral blood flow that occur in response to neuronal activation, a phenomenon known as neurovascular coupling, and therefore provides an indirect measure of brain activity. The underlying principle is similar as what is measured in functional magnetic resonance imaging (fMRI), the current clinical standard for brain imaging. Compared to fMRI, fUS offers several advantages, it is portable, cost-effective, higher temporal resolution, and higher sensitivity to cerebral blood flow. This makes it a promising tool for both preclinical neuroscience, and clinical application. However, there remain significant challenges to overcome before fUS can be widely adopted in clinical settings.

First, the brain is protected by the skull, which poses a barrier for ultrasound waves. The skull bone distorts and attenuates ultrasound signals, leading to decreased transcranial image quality. Therefore, most studies to date are restricted to animal models, where the skull can be surgically removed or thinned. In humans, fUS has been applied during intraoperative procedures, where the skull is removed, and the brain is exposed.
Second, fUS generates enormous amounts of data, which complicates its use in real-time applications.

This thesis addresses both challenges. It focuses on enhancing transcranial image quality, bringing us closer to fully noninvasive, high resolution brain imaging. In addition, it introduces methods for reconfigurable functional imaging, aimed at reducing data rates to enable real time decoding of brain activity into actionable outputs. Together, these advances increase the translational potential of fUS and lower the barrier for clinical and neuroscience adoption.

The field of aberration correction consist in improving image quality by compensating for distortions caused by the medium through which the ultrasound waves travel. In this thesis, we apply aberration correction to restore transcranial image quality.
Aberration correction starts with knowing the exact properties of the ultrasound probe. Chapter 2 introduces a simple method to estimate the speed and thickness of the silicone lens on 1D transducers. This calibration is essential to accurately estimate the speed of sound in a medium, independent of imaging depth and transmission parameters. Using optimal lens parameters, and the estimated sound speed, we demonstrated an improvement in image resolution and contrast.

In chapter 3, attention shifts to the challenge of restoring transcranial image quality. An adaptive aberration correction approach is presented, using ray tracing through four tissue layers: transducer lens, skin, skull, and brain. This model estimates wave speeds in each layer, then reconstructs images based on the actual (refracted) wave paths. Applied to Doppler imaging in rats, the method improves both resolution and sensitivity, especially in cortical areas where skull induced aberrations are strongest.

Chapter 4 takes on the problem of high data rates in 3D imaging. Since functional activation in the brain is typically sparse, volumetric imaging often captures unnecessary data. Here, a new technique called selective-plane fUS is introduced. It combines focused wave transmission with a Row-Column Addressed (RCA) transducer to target only the brain regions of interest. This significantly reduces the computational and data transfer load and potentially paves the way for lightweight, portable brain-machine interfaces based on fUS.

In chapter 5, the thesis explores imaging of cellular activity and capillary flow using ultrasound contrast agents. We introduce a technique called Nonlinear Sound-sheet Microscopy (NSSM), that enables high resolution imaging of contrast agents within thin planes. This approach captures both vascular and gene expression data in living tissue and extends ultrasound imaging toward cellular resolution in opaque organs.

Together, these chapters lay the technical foundation for next-generation functional and biomolecular ultrasound: systems that are more accurate, less invasive, and better suited for high resolution brain imaging in real time. By addressing both the physical challenges of wave distortion and the computational load of volumetric data, this thesis brings fUS a step closer to clinical and translational neuroscience applications.

For over 50 years, medical ultrasound has been a pioneering force in healthcare, seamlessly blending diagnostic and therapeutic applications. In diagnostics, ultrasound reveals intricate internal structures based on the generation of acoustic pressure waves in the human body and capturing reflections from tissue and body structures, enabling precise anomaly identification. Simultaneously, in therapy, ultrasound utilizes its acoustic prowess for targeted interventions and submillimeter localization in unconventional cancer treatments like proton therapy. The field has shifted significantly with the advent of nonlinear acoustics, exploring wave propagation phenomena. Contrast agents marked a revolutionary leap, enhancing the specificity, sensitivity, and efficiency of diagnostic and therapeutic ultrasound. Microbubbles, as the most conventionally used agents, exhibit strong nonlinear scattering under ultrasonic excitation, making them suitable for CEUS applications.

Medical ultrasound is crucial in modern healthcare, with high-frame-rate imaging expanding its applications to both high blood velocities over large fields-of-view and slow-moving blood. Accurate flow measurements require high temporal and spatial resolution, necessitating the use of matrix transducers, which face several challenges depending on the application. For monitoring the brains of preterm babies, challenges include designing ASIC-integrated matrices that meet clinical needs and minimizing cables to ensure unrestricted mother-child contact. For atherosclerosis screening, challenges involve sensor-ASIC integration, developing high-frame-rate data acquisition, and addressing high side-lobe levels in sparse matrices. The objective of this thesis is to address these challenges and develop the necessary technology to enable the use of ultrasound matrix transducers for quantifying blood flow and detecting physiologically abnormal flow patterns and velocities in the two mentioned applications.

This thesis describes the design, prototyping and evaluation of matrix-based clamp-on ultrasonic flow meters. Several new measurement techniques are presented as well as an Application-Specific Integrated Circuit (ASIC) designed for accurate measurement of flow velocity with matrix transducers.

The influence of circuit topologies on the zero-flow performance of ultrasonic flow meters has been analyzed and an algorithm is presented to reduce the offset. With a linear transducer array, flow measurements have been performed via two different acoustic paths, demonstrating the ability to accurately measure flow with array transducers through a stainless-steel pipe wall. In order to improve signal quality, an ASIC has been designed that is able to drive and read-out 96 piezo transducer elements. The ASIC has been characterized electrically and flow measurements have been performed in combination with the linear transducer arrays.

Several new techniques, enabled using transducer arrays, have also been explored. By tapering the amplitude of the transmit signals, spurious waves can be suppressed. An auto-calibration technique has been developed that uses additional acoustic measurements to estimate the diameter of the pipe and the speed of sound in the pipe wall and liquid. Finally, a simulation study has been performed to explore the possibility of exploiting the beam-steering capabilities of transducer arrays to measure flow velocity profiles by using measurements obtained via multiple acoustic paths.

Ultrasonic flow meters are widely applied to measure flow in a variety of applications. The vast majority of ultrasonic flow meters are based on the measurement of the transit time of an acoustic pulse through the fluid. This can either be done in-line, by inserting a spool piece with ultrasonic transducers into the pipe carrying the fluid, or by clamping the transducers on an existing pipe. Clamp-on meters are attractive as they can be installed without cutting the pipe or shutting down the flow, but their stability is limited, and they are unable to measure flow profiles (in contrast with expensive multi-path in-line meters), which limits their linearity at low flow speeds. Moreover, their installation requires complex manual alignment of the transducers and input of a variety of setup parameters (e.g. pipe dimensions and material properties, speed of sound in the fluid) by the user. In this thesis, clamp-on meters based on matrix ultrasonic transducers are developed to address these drawbacks. These matrix transducers consist of a two-dimensional array of 100+ elements that enables beam steering in two directions by programming the timing of the electrical pulses applied to the elements. This allows to develop three innovative measurement techniques: (1) automatic beam alignment by adjusting the steering angles so as to optimize the signal-to-noise ratio and the path of the received pulse, thus simplifying installation and improving stability; (2) multi-path measurement by steering the beam at different angles, realizing the measurement of multiple paths through the fluid with a single pair of matrix transducers, and thus providing information about the flow profile; (3) self-calibration by using pulse-echo measurements between the elements of the matrix transducer to characterize the pipe wall and fluid, thus reducing the dependence on a-priori knowledge of their properties. The most significant steps to realize these kind of sensors were taken in this thesis. The mentioned measurement techniques were elaborated, and the relevant wave-propagation phenomena, beamforming schemes and transducer design were performed. Based on this, a prototype sensor was fabricated and successfully tested. Moreover, application-specific integrated circuits (ASICs) were developed with dedicated transmit and receive electronics to realize a cost-effective and accurate implementation of the beam forming and transit-time measurement.

Ultrasound imaging is a widely available and highly portable real-time, ionizing radiation-free imaging modality used as first-line in multiple areas. Conventionally, its use is focused on providing 2D imaging using ultrasound probes. However, with the introduction of matrix probes, the acquisition of 3D scans became possible enabling the sonographer to accurately interpret the scanned anatomy without needing to build a possibly error prone 3D mental map and thus, decreasing operator-dependency. Although with this 3D imaging, volume estimation accuracy increases and surveillance becomes less difficult, matrix probes are expensive, have a limited field-of-view and produce a low contrast image. This necessitates the sonographer to still build a mental map whenever a portion of an anatomy is not fully visible in a single volumetric scan, leading to more laborious surveillance. Ultrasound stitching is investigated as a potential method to overcome these limitations by tracking multiple ultrasound scans and merging them using their position and orientation coordinates. To this end, this work uses a robotic arm as the selected tracking system for the stitching of 2D xplane ultrasound images. First, the initial clinical application for the system was derived using a combination of literature study and interviews with medical experts which led to the decision of utilizing the robotic arm for liver imaging and biopsy needle guidance. The suitability of this application was validated with the evaluation of the three required stages namely anatomy reconstruction, needle path planning and needle insertion. Two phantoms mimicking lesions and surrounding vessels were fabricated for this purpose. They were firstly reconstructed by manually segmenting the elements of interest from multiple scans, placing these segmented scans in their respective 3D positions and finally applying an alpha-shapes reconstruction algorithm. Once that the centroid of the lesions was calculated and together with the position information about the surface of the phantom obtained with the robotic arm, it was possible to evaluate all the paths that connect the surface with the centroid and subsequently select only the ones that did not pass next to a vessel within a specific distance. Finally and as an extension of using the robotic arm to track the ultrasound probe, the combined tracking of the probe with a needle holder was investigated for the needle insertion stage. The needle holder was used to maintain the needle always in plane with the ultrasound beam and it was placed at a preset distance with respect to the probe. The holder also provided insertion depth information that facilitated the representation of the needle via a line in the visualization environment whose end indicated the position of the needle tip. This position was updated in real-time with the information received from the needle holder while the needle was being inserted. The position of the needle tip was verified by overlapping real-time ultrasound images with the visualized line. Ultimately, to further illustrate the suitability of the robotic arm in a clinical setting, an ex-vivo and a clinical experiment for vessel tracking and reconstruction was performed.

During the past decade, time-lapse seismic technology has been widely applied in hydrocarbon reservoir management. It is a very powerful method to obtain information on reservoir changes in the inter-well regions. This information helps to identify bypassed hydrocarbons and extend the economic life of a field. In a typical scenario, one baseline survey and subsequent monitoring surveys are acquired over time. The survey geometry is usually exactly repeated and well-sampled to mitigate acquisition effects on the next steps in the process. By processing and comparing all the datasets, some physical changes, e.g. reflection amplitude and travel-time changes, can be estimated. These time-lapse changes are then used to calculate interpretable parameter changes in dynamic reservoir rock and fluid properties, e.g. pore pressure and fluid saturation.

In a conventional time-lapse processing workflow, all the multiples are first removed from the data, then independent imaging process is employed to each dataset, given the same propagation velocity model. Later on, to compensate the ignored velocity variations between different surveys, a time-shift map (travel-time differences) is estimated from the calculated images and then applied back to them, yielding the final reflection amplitude differences. However, this conventional processing strategy is usually sensitive to the success of multiple removal and survey repeatability, and also requires well-sampled surveys providing proper illumination. Moreover, artifacts are often generated in addition to the actual time-lapse changes due to the non-repeatable uncertainties during the independent processing steps. Regarding the time-shift-map tool, the relative velocity changes derived from the time-shift map are not the actual velocity changes due to its local 1D subsurface assumption that is embedded.

In order to relax these rigid requirements and have a better velocity change indicator, we propose Simultaneous Joint Migration Inversion (S-JMI) as an effective time-lapse tool for reservoir monitoring, which combines a simultaneous time-lapse data processing strategy with the Joint Migration Inversion (JMI) method. JMI is a full wavefield inversion method that explains the measured reflection data using a parameterization in terms of reflectivities and propagation velocities. JMI is able to make use of multiples and at the same time take velocity variations between surveys into account. The simultaneous strategy, which means fitting all the datasets simultaneously, allows the baseline and monitor parameters to communicate and compensate with each other dynamically during inversion via L2-norm constraints, thus, reducing the non-repeatable uncertainties during the time-lapse processing workflow. As a result, more accurate time-lapse differences can be achieved by S-JMI, compared to inverting each dataset independently. Moreover, in order to get more localized time-lapse velocity differences, we further extend the regular S-JMI to a robust high-resolution S-JMI (HR-S-JMI) process by making a link between the reflectivity/reflectivity-difference and velocity/velocity-difference during inversion. With a complex synthetic example based on the Marmousi model, we demonstrate the performance of the time-shift-map-based method, sequential JMI, the regular S-JMI and HR-S-JMI is improving in this particular order.

Next, we further demonstrate the effectiveness of the proposed method in more real-life cases with a highly realistic synthetic model based on the Grane field, offshore Norway, and a time-lapse field dataset from the Troll Field. Moreover, in order to investigate the feasibility of HR-S-JMI in practice, several numerical experiments based on the realistic Grane model are conducted, regarding the following aspects: noise, including random noise and coherent noise caused by the acoustic assumption; the quality of time-lapse surveys, including sparse surveys, non-repeated surveys, and Ocean Bottom Node (OBN) vs streamer (different types of monitoring surveys); non-repeated sources, including source positioning errors and non-repeated source wavelets; spatial weighting operators in the L2-norm constraints; and sensitivity to weak time-lapse effects. These experiments show that HR-S-JMI is very robust to random noise, coherent noise, survey sparsity, survey non-repeatability, source positioning errors and source wavelet discrepancies. Furthermore, HR-S-JMI remains effective when the spatial weighting operators in the L2-norm constraints are largely relaxed and HR-S-JMI is capable of detecting weak time-lapse changes (e.g. velocity changes down to +/- 35 m/s). These features make it a suitable time-lapse processing solution for cost-effective (semi-)continuous monitoring, termed i4D survey technology, in which inexpensive localized and sparse surveys are employed between the conventional full-field surveys. The simultaneous strategy of S-JMI allows the full-field survey information to compensate the poor illumination of the in-between sparse surveys during process. Furthermore, calender-time constraints are proposed and applied to the parameter differences between the baseline and monitors along the calender-time axis by taking advantage of the feature that time-lapse effects usually develop gradually over time. With a complex synthetic example based on the Marmousi model, we demonstrate that S-JMI is a promising tool to process datasets acquired from (semi-)continuous monitoring, like an i4D survey.

In conclusion, we propose high-resolution simultaneous JMI (HR-S-JMI) as an effective time-lapse processing tool for the following main reasons:
• HR-S-JMI is able to make use of multiples to extend the illumination of the subsurface, instead of removing them;
• HR-S-JMI is an extended imaging process, including automatic velocity updating. Therefore, it takes velocity variations between surveys directly into account;
• HR-S-JMI is a good indicator of velocity changes, it can invert for high-resolution accurate time-lapse velocity changes;
• HR-S-JMI is robust to the uncertainties existing in the monitoring surveys, e.g. noise, sparsity, non-repeatability, source positioning errors, source wavelet discrepancy, etc;
• HR-S-JMI has the ability to detect weak time-lapse changes (velocity changes down to +/- 35 m/s)

This thesis describes the analysis, design and evaluation of front-end application-specific integrated circuits (ASICs) for 3-D medical ultrasound imaging, with the focus on the receive electronics. They are specifically designed for next-generation miniature 3-D ultrasound devices, such as transesophageal echocardiography (TEE), intracardiac echocardiography (ICE) and intravascular ultrasound (IVUS) probes. These probes, equipped with 2-D array transducers and thus the capability of volumetric visualization, are crucial for both accurate diagnosis and therapy guidance of cardiovascular diseases. However, their stringent size constraints, as well as the limited power budget, increase the difficulty in integrating in-probe electronics. The mismatch between the increasing number of transducer elements and the limited cable count that can be accommodated, also makes it challenging to acquire data from these probes. Front-end ASICs that are optimized in both system architecture and circuit-level implementation are proposed in this thesis to tackle these problems.
The techniques described in this thesis have been applied in several prototype realizations, including one LNA test chip, one PVDF readout IC, two analog beamforming ASICs and one ASIC with on-chip digitization and datalinks. All prototypes have been evaluated both electrically and acoustically. The LNA test chip achieved a noise-efficiency factor (NEF) that is 2.5 × better than the state-of-the-art. One of the analog beamforming ASIC achieved a 0.27 mW/element power efficiency with a compact layout matched to a 150 µm element pitch. This is the highest power-efficiency and smallest pitch to date, in comparison with state-of-the-art ultrasound front-end ASICs. The ASIC with integrated beamforming ADC consumed only 0.91 mW/element within the same element area. A comparison with previous digitization solutions for 3-D ultrasound shows that this work achieved a 10 × improvement in power-efficiency, as well as a 3.3 × improvement in integration density.

Currently a lot of effort is put into developing matrix arrays which allow for volumetric imaging and new applications. There are however multiple problems. Compared to arrays currently in use which have in the order of 128 elements, matrix arrays can easily contain 1000 to 10000 elements. If all elements would be connected independently, the cable would become very thick. Therefore, beamforming methods are required that can operate with fewer transmit and receive channels. Furthermore, the room for electronics on the chip is limited. So, the required electronics for the beamforming methods should be kept simple.

In this thesis we will propose beamforming methods that are able to operate with fewer channels. This will be done separately for the transmit and receive part, but they do in no way exclude each other. To be able to focus pulsed waves in transmit we propose a method based on Fresnel zone plates which are used in optics to focus continuous wave light. Our method only requires a single continuous-wave excitation signal to be present, which is connected and disconnected on demand to each element. We have evaluated our method with measurements and simulations. As compared to the conventional focusing method, the spatial resolution is not affected by our method, but the Contrast-to-Noise ratio is 5\% lower for shallow depths and up to 20\% lower deeper into the medium. Overall though, the differences were relatively small and so it is clear that our new focusing method works very well. If needed, better results can be obtained by trading in frame rate. In this case the results are almost indistinguishable from the conventional focusing method.

To solve the problem with image formation, we have developed a frequency domain two stage beamforming method for use with matrix arrays, which does not require all element data to be present. This has been done for two matrix types. For the first method we have confirmed with simulations that it performs similar to the respective results obtained with two 2D frequency domain two stage beamforming method that have already been experimentally verified. For the second method we have evaluated the performance with simulations and measurements. Our method was able to obtain a 25\% better spatial resolution as compared to Dynamic Receive Focusing(DRF), without additional artefacts. As an alternative to the last method, we have also developed a frequency domain beamforming method that does require all element data, but only requires a single insonification by a spherical wave. This method did perform worse than the method discussed before in both simulations and measurements, but it does outperform DRF applied to spherical wave data.