A Convolutional Neural Network for Beamforming and Image Reconstruction in Passive Cavitation Imaging.

Convolutional neural networks (CNNs), initially developed for image processing applications, have recently received significant attention within the field of medical ultrasound imaging. In this study, passive cavitation imaging/mapping (PCI/PAM), which is used to map cavitation sources based on the correlation of signals across an array of receivers, is evaluated. Traditional reconstruction techniques in PCI, such as delay-and-sum, yield high spatial resolution at the cost of a substantial computational time. This results from the resource-intensive process of determining sensor weights for individual pixels in these methodologies. Consequently, the use of conventional algorithms for image reconstruction does not meet the speed requirements that are essential for real-time monitoring. Here, we show that a three-dimensional (3D) convolutional network can learn the image reconstruction algorithm for a 16×16 element matrix probe with a receive frequency ranging from 256 kHz up to 1.0 MHz. The network was trained and evaluated using simulated data representing point sources, resulting in the successful reconstruction of volumetric images with high sensitivity, especially for single isolated sources (100% in the test set). As the number of simultaneous sources increased, the network's ability to detect weaker intensity sources diminished, although it always correctly identified the main lobe. Notably, however, network inference was remarkably fast, completing the task in approximately 178 s for a dataset comprising 650 frames of 413 volume images with signal duration of 20µs. This processing speed is roughly thirty times faster than a parallelized implementation of the traditional time exposure acoustics algorithm on the same GPU device. This would open a new door for PCI application in the real-time monitoring of ultrasound ablation.

Acoustic radiation force induced accumulation and dynamics of microbubbles on compliant surfaces.

Ultrasound stimulated microbubbles have been shown to be capable of breaking up blood clots through micro-scale interactions occurring near the clot surface. However, only a small fraction of bubbles circulating in the bloodstream will be in close proximity to such boundaries, where they must be to elicit therapeutic effects. Here, the accumulation and subsequent behavior of microbubbles displaced from an overlying flow channel to a boundary under radiation forces were examined. Experimental data were acquired using a novel high speed microscopy configuration and simulations were conducted to provide insight into the accumulation process. There was broad agreement between experiments and simulations, both indicating that the size distribution and number of bubbles arriving at the boundary depended on channel flow rate, applied pressure, and bubble concentration. For example, higher flow rates and lower pressures favored the accumulation of larger bubbles relative to the native agent distribution. Moreover, bubble dynamics were dependent on the surface type, exhibiting rapid translation along agarose gel surfaces whereas on fibrin surfaces, they accumulated in localized regions inducing repetitive strain cycles. The results indicate that the process of bringing bubbles from within a vessel to a boundary is complex and should be an important consideration in the development of therapeutic applications such as sonothrombolysis.

Translational dynamics of individual microbubbles with millisecond scale ultrasound pulses.

It is established that radiation forces can be used to transport ultrasound contrast agents, particularly for molecular imaging applications. However, the ability to model and control this process in the context of therapeutic ultrasound is limited by a paucity of data on the translational dynamics of encapsulated microbubbles under the influence of longer pulses. In this work, the translation of individual microbubbles, isolated with optical tweezers, was experimentally investigated over a range of diameters (1.8-8.8 µm, n = 187) and pressures (25, 50, 100, 150, and 200 kPa) with millisecond pulses. Data were compared with theoretical predictions of the translational dynamics, assessing the role of shell and history force effects. A pronounced feature of the displacement curves was an effective threshold size, below which there was only minimal translation. At higher pressures (≥150 kPa) a noticeable structure emerged where multiple local maxima occurred as a function of bubble size. The ability to accurately capture these salient features depended on the encapsulation model employed. In low Reynolds number conditions (i.e., low pressures, or high pressures, off-resonance) the inclusion of history force more accurately fit the data. After pulse cessation, bubbles exhibited substantial displacements consistent with the influence of history effects.

Receiver array design for sonothrombolysis treatment monitoring in deep vein thrombosis.

High intensity focused ultrasound (HIFU) can disintegrate blood clots through the generation and stimulation of bubble clouds within thrombi. This work examined the design of a device to image bubble clouds for monitoring cavitation-based HIFU treatments of deep vein thrombosis (DVT). Acoustic propagation simulations were carried out on multi-layered models of the human thigh using two patient data sets from the Visible Human Project. The design considerations included the number of receivers (32, 64, 128, 256, and 512), their spatial positioning, and the effective angular array aperture (100° and 180° about geometric focus). Imaging array performance was evaluated for source frequencies of 250, 750, and 1500 kHz. Receiver sizes were fixed relative to the wavelength (pistons, diameter = λ/2) and noise was added at levels that scaled with receiver area. With a 100° angular aperture the long axis size of the -3 dB main lobe was ~1.2λ-i.e. on the order of the vessel diameter at 250 kHz (~7 mm). Increasing the array aperture to span 180° about the geometric focus reduced the long axis by a factor of ~2. The smaller main lobe sizes achieved by imaging at higher frequencies came at the cost of increased levels of sensitivity to phase aberrations induced during acoustic propagation through the intervening soft tissue layers. With noise added to receiver signals, images could be reconstructed with peak sidelobe ratios < -3 dB using single-cycle integration times for source frequencies of 250 and 750 kHz (NRx ⩾ 128). At 1500 kHz, longer integration times and/or higher element counts were required to achieve similar peak sidelobe ratios. Our results suggest that a modest number of receivers(i.e. NRx = 128) arranged on a semi-cylindrical shell may be sufficient to enable passive acoustic imaging with single-cycle integration times (i.e. volumetric rates up to 0.75 MHz) for monitoring cavitation-based HIFU treatments of DVT.

Megahertz rate, volumetric imaging of bubble clouds in sonothrombolysis using a sparse hemispherical receiver array.

It is well established that high intensity focused ultrasound can be used to disintegrate clots. This approach has the potential to rapidly and noninvasively resolve clot causing occlusions in cardiovascular diseases such as deep vein thrombosis (DVT). However, lack of an appropriate treatment monitoring tool is currently a limiting factor in its widespread adoption. Here we conduct cavitation imaging with a large aperture, sparse hemispherical receiver array during sonothrombolysis with multi-cycle burst exposures (0.1 or 1 ms burst lengths) at 1.51 MHz. It was found that bubble cloud generation on imaging correlated with the locations of clot degradation, as identified with high frequency (30 MHz) ultrasound following exposures. 3D images could be formed at integration times as short as 1 µs, revealing the initiation and rapid development of cavitation clouds. Equating to megahertz frame rates, this is an order of magnitude faster than any other imaging technique available for in vivo application. Collectively, these results suggest that the development of a device to perform DVT therapy procedures would benefit greatly from the integration of receivers tailored to bubble activity imaging.

The microscale evolution of the erosion front of blood clots exposed to ultrasound stimulated microbubbles.

Serial two-photon microscopy of blood clots with fluorescently tagged fibrin networks was conducted during microbubble-mediated sonothrombolysis to examine the microscale evolution of the resulting erosion front. The development of a complex zonal erosion pattern was observed, comprised of a cell depleted layer of fibrin network overlying intact clot which then underwent progressive recession. The fibrin zone architecture was dependent on exposure conditions with 0.1 MPa causing no erosion, 0.39 MPa resulting in homogenous structure, and combination 0.39/0.96 MPa pulses forming large-scale tunnels. High speed imaging and Coulter counter data indicated the fibrin zone formation process involves the ejection of intact erythrocytes.