Placenta-mediated diseases, such as preeclampsia (PE) and small-for-gestational-age (SGA) neonates, are associated with structural and functional changes in the placenta. While ultrasound is routinely performed during pregnancy, its potential for detecting such diseases using deep learning remains underexplored. This study presents a deep transfer learning approach using Convolutional Neural Networks (CNNs) to detect placenta-mediated diseases from ultrasound B-mode images of the human placenta ex vivo. This study is a secondary analysis of data from a prospective case-control study, where placentas were collected from women who delivered at BC Women's Hospital, Vancouver, Canada, from April 2018 to April 2020. Ultrasound data obtained from ex vivo human placentas were used to perform transfer learning using three deep learning architectures: DenseNet-121, ResNet-50, and InceptionV3, each pretrained on one of two datasets: ImageNet or RadImageNet, to detect SGA and/or PE outcome. The dataset consisted of 46 placentas obtained from 40 participants, of which 25 placentas were complicated by SGA and/or PE (including 12 SGA-only placentas, 2 PE placentas, and 11 placentas affected by both SGA and PE). Experimental results demonstrate that DenseNet-121, when pretrained with ImageNet, achieved the most promising performance, with accuracy of 0.76 ± 0.12, F1-score of 0.78 ± 0.09, and area under the receiver operating characteristic curve (AUC) of 0.86 ± 0.10. Among the RadImageNet pretrained models, ResNet-50 yielded an accuracy 0.75 ± 0.10, F1-score 0.77 ± 0.07, and AUC of 0.84 ± 0.09. Our proof-of-concept results demonstrate that standard B-mode ultrasound images of the placenta ex vivo contain sufficient information to distinguish between normal pregnancies and pregnancies affected by SGA and PE.

This study investigates the effects of post-fabrication annealing on polymer-based capacitive micromachined ultrasonic transducers (polyCMUTs). These devices comprise microscopic diaphragms produced via photolithographic patterning of polymer layers. Critical point drying, required to release the diaphragms, can cause significant plastic deformation, thereby reducing electromechanical coupling. Post-fabrication annealing, carried out in incremental steps up to 190 °C, led to an effective increase in coupling by a factor of 5.4. Atomic Force Microscopy showed that the initial upward deflection of 162.7 nm decreased to 6.2 nm after annealing at 190 °C, while also improving surface uniformity. In parallel, the transducer’s resonance frequency rose from 2.33 MHz (unannealed) to 2.60 MHz, and the input impedance phase angle at resonance increased from −68.1° to −4.3°. Together, these changes indicate a significant improvement in resonator behavior and, consequently, device performance. Thus, post-fabrication annealing is an effective measure to achieve the designed performance while enhancing manufacturing yield, thereby increasing the applicability of polyCMUTs.

Wearable ultrasound enables continuous monitoring of physiological processes such as muscle dynamics, bladder volume, and cardiovascular activity. Existing fully wearable ultra-low-power platforms are limited to shallow, low-channel A-mode sensing, while larger multi-mode systems are too bulky and power-hungry for true wearability. We present WULPUS PRO, a runtime-programmable wearable ultrasound acquisition platform measuring 39×21×6 mm and weighing 5 g. It integrates 30 V excitation, 16 time-multiplexed channels, a low-noise receive front-end with up to 70 dB gain, 9.9 MHz bandwidth, time-gain compensation, and 32 dB SNR. The platform supports deep-tissue echo acquisition up to 2.2 MHz in RF-sampling mode and 8 MHz in envelope-detection mode. We demonstrate B-mode imaging in a 16-channel ultra-low-power wearable with sub-millimeter axial and millimeter-scale lateral resolution in phantom experiments, while consuming 40 mW at 50 Hz PRF and under 60 mW at 300 Hz PRF. WULPUS PRO supports both piezoelectric and capacitive micromachined ultrasonic transducers, enabling integration with skin-conformal polymer-based CMUT arrays. As a host-agnostic acquisition front-end, it exposes standard data and power interfaces for BLE- and Wi-Fi-based wearable hosts. We demonstrate wireless transmission with external BLE and Wi-Fi modules and project 1-2 days of BLE operation at 50 Hz PRF and over 3 h of Wi-Fi streaming at 300 Hz PRF using a 300 mAh, 6.4 g Li-Po cell. WULPUS PRO establishes a new class of fully programmable, B-mode-enabled, ultra-low-power wearable ultrasound platforms.

Thermoplastic composite structures enable lightweight, recyclable, and high-throughput aerospace manufacturing, but reliable quality assurance of advanced joining processes remains a key challenge. This work presents a compact, low-cost, and wireless ultrasonic non-destructive testing system for real-time, inline monitoring of continuous ultrasonic welding of thermoplastic carbon fiber composites. The system integrates custom-fabricated polymer-based capacitive micromachined ultrasonic transducers (polyCMUTs) with the ultra-low-power WULPUS platform, enabling operation in the harsh, high-interference welding environment. An eight-element linear polyCMUT array operating at a center frequency of approximately 3.6 MHz is designed, fabricated, packaged, and integrated into an industrial welding setup. Inline measurements are performed during welding of carbon fiber laminates with intentionally introduced defects. Process-synchronous ultrasonic data reveal consistent depth-of-echo shifts at defect locations, in strong agreement with X-ray computed tomography ground truth. Across 21 welds, all induced defects are detected without false negatives and with limited false positives. The results demonstrate that polymer-based CMUT technology enables robust, scalable, and manufacturing-compatible ultrasonic sensing, representing a decisive step toward intelligent process monitoring and quality assurance for next-generation thermoplastic composite welding.

Most imaging solutions that utilize ultrasound wave propagation require the projection of temporal signals received by transducer elements into spatial maps. This process, known as beamforming, requires knowledge of the speed-of-sound (SoS) in the medium. Incorrect SoS assumptions lead to aberration artifacts, reducing image quality and limiting clinical usability. SoS is also a novel imaging biomarker for assessing tissue characteristics. In this study, we propose a spatial SoS distribution estimation method using conventional hand-held ultrasound transducers and simple laser-diode based photoacoustics. By identifying the time of flight between photoacoustic point sources and transducer elements and relating these to SoS values along corresponding propagation paths, the underlying SoS maps are reconstructed by solving an inverse problem. We validate our method through numerical simulations and ex vivo experiments. Numerical phantoms are successfully reconstructed under various noise levels and numbers of photoacoustic sources. In ex vivo experiments with chicken breast, the estimated SoS value is consistent with results reported in the literature. The proposed approach offers a low-cost, compact solution for photoacoustics-based SoS estimation in various clinical applications, such as breast and intra-operative prostate imaging, both for diagnosis and for improving image quality of acoustic-based modalities, including photoacoustic imaging.

Quantitative ultrasound (QUS) enables tissue characterization and has shown promise in placental assessment. However, accurate QUS estimation requires system calibration, typically matched to imaging depth. This study investigates the impact of calibration depth offsets using RF data from a tissue-mimicking phantom and two in vivo placenta scans. QUS parameters—attenuation coefficient estimate (ACE) and backscatter coefficient (BSC)—were computed at imaging depths from 8 to 18 cm with calibration offsets up to 45 mm. Phantom and patient results showed very low errors (<5%) up to 10 mm offsets, with the lower attenuation of placental tissue suggesting greater tolerance to calibration mismatch.

Acoustic Emission (AE) detecting is gaining increased interest in Structural Health Monitoring (SHM) and Non-Destructive Testing (NDT), driving the demand for cost-effective sensors with small and flexible form factors. This paper presents the first front-end amplified polymer-based Capacitive Micromachined Ultrasonic Transducers (polyCMUTs) for AE detection. PolyCMUTs are polymer-based microscopic structures with membranes that vibrate in response to incoming acoustic waves. These sensors offer low cost, broad bandwidth, small footprint and the potential for flexible designs. Following the face-to-face characterization procedures specified in ISO 24543:2022, the front-end amplified polyCMUT sensors measure a sensitivity of 10 dB V/nm across a frequency range of 100 kHz to 2 MHz, comparable to industry -leading piezoelectric sensors. Cost analysis shows that polyCMUT sensors are producible under US $15, a small fraction of the cost of conventional AE sensors. These compact, low-cost and high-performance front-end amplified polyCMUTs demonstrate great commercial potential for AE applications. Future work will focus on improving the sensitivity and noise rejection in stopbands.

Continuous ultrasound monitoring for non-destructive testing (NDT) and wearables requires systems that are miniature, energy efficient, and tightly integrated with low-profile sensor arrays for unobtrusive, multiday operation. Existing solutions struggle to balance size, power, and functionality. This paper presents a compact, low-cost, fully integrated wireless ultrasound patch for long-term single-channel pulse-echo monitoring. The system is derived from WULPUS, an open-source platform for wearable low-power ultrasound, and is optimized for size and bill-of-materials while supporting polymer-based capacitive micromachined ultrasonic transducers (polyCMUTs). PolyCMUTs enable thin (<1 mm) form factors, broad bandwidth, rapid prototyping, and low fabrication cost. The electronics, built entirely from off-the-shelf components, integrate a Bluetooth transceiver for connectivity, a microcontroller with a high-speed analog-to-digital converter, and an ultrasonic front end capable of 30 V unipolar excitation and a 30 V bias tailored to polyCMUT operation. A flexible interposer PCB allows direct wire bonding of the transducer and mates to a main PCB with a rechargeable 400 mAh cell sandwiched between boards; data can be streamed wirelessly to a host for real-time processing and visualization. The assembled device measures 28 mm x 26 mm x 12 mm, weighs 13.3 g, with estimated production costs approximately $25 USD. Active-mode power is <33 mW, supporting >40 h of continuous operation at a measurement rate of 50 Hz. Pulse-echo tests in deionized water demonstrated high signal strength, with receiver amplitudes up to 1.1 V-pp at 13.8 dB gain.

Quantitative Ultrasound in Placenta (QUS-P) measures the acoustic properties of the underlying tissue and therefore can be used to detect structural changes associated with placenta-mediated diseases in utero. To develop a real-time, non-invasive clinical tool using QUS-P, we first conducted an ex-vivo study on post-delivery human placentas (n = 47), of which 25 were from pregnancies affected by pre-eclampsia (PE) and/or small-for-gestational age (SGA). Ultrasound radio-frequency data were analyzed to estimate QUS-P parameters: attenuation coefficient, backscatter coefficient and effective scatterer diameter. A logistic regression model developed using these QUS-P parameters achieved high discrimination (AUROC: 0.89 (95% CI: 0.78–0.98)) and calibration (slope: 0.93 and intercept 0.003) for PE/SGA detection. Building on these findings, we conducted STIMULUS, an in utero study involving pregnant participants (34–36 weeks gestation). Preliminary analysis (n = 184) identified 29 (15.8%) neonates with perinatal hypoxia, 5 of which showed maternal/fetal vascular malperfusion. Logistic regression using QUS-P parameters predicted hypoxia with an AUROC of 0.70 (95% CI: 0.58–0.81) and the sub-category of placenta-mediated hypoxia with an AUROC of 0.98 (95% CI: 0.88–1.0). The findings demonstrate that QUS-P holds potential for integration into prenatal care as an early, non-invasive tool for predicting placenta-mediated diseases before clinical onset.

Quantitative ultrasound (QUS) holds promise for non-invasive placental tissue characterization and disease detection, yet its clinical application is hindered by the effort required for placental segmentation to identify a region-of-interest (ROI) for calculation. Manual segmentation is time-consuming and prone to variability due to the irregular boundaries of the placental tissue. This study aims to develop an automatic placental identification and segmentation model to facilitate QUS integration into clinical practice, alleviate clinician workload, and enable real-time feedback on QUS quality. We employed Mask R-CNN within the Detectron2 framework to automate placental segmentation using a dataset of 149 B-mode ultrasound images annotated by a medical image specialist (M.D.) covering various trimesters. Inference on a validation subset (30 images) yielded an average Dice Similarity Coefficient (DSC) of 0.863. Our model gave quality predictions for a majority of the test images with 57% of segmentations achieving DSC >= 0.85 and 40% with DSC >= 0.90. The model's low inference time of approximately 55 ms per iteration supports near real-time QUS processing. Future work will focus on expanding the dataset, optimizing the loss function, and addressing edge effects on QUS to further enhance segmentation quality.

Conventional B-mode ultrasound is widely used in clinical settings and can be supplemented by quantitative analysis of backscattered signals. Quantitative ultrasound (QUS) has seen growing interest in providing supplemental and quantitative insights into tissue characteristics and disease diagnostics. The accuracy of QUS, however, depends on the acquisition and format of the backscattered signals. This paper investigates the impact of several data acquisition techniques on QUS performance. The study demonstrates that unprocessed channel data, from plane wave imaging, delivers the most accurate Attenuation Coefficient Estimation (ACE) measurement, with an average error of about 5% for ACE across different independent experiments.

Carbon fibre-reinforced polymer (CFRP) materials are used for various applications due to their desirable me-chanical properties. Defects, such as delaminations, can occur during production or operation of CFRP components and can compromise their structural integrity. Ultrasonic imaging is a well-established and cost-effective technique for non-destructive testing (NDT) of CFRP components. We evaluate the use of supervised training of artificial neural networks (ANNs) for the automated detection of delaminations from ultrasound B-scans. After hyperparameter optimization, convolutional neural networks (CNNs) performed better than multilayer perceptrons (MLPs). Using a test specimen, we achieved a recall of 100 % and a precision of 94.3 % on unseen test images.

Introduction: Placenta-mediated diseases are associated with structural changes in the placenta. Quantitative Ultrasound (QUS) imaging measures the acoustic properties of the tissue, which are correlated to the underlying tissue structure. We aimed to develop and validate a diagnostic prediction model using QUS measurements for pre-eclampsia (PE) and small-for-gestational-age (SGA) fetuses/neonates. Methods: For this prospective case-control study, placentas were collected from a group of women who delivered via cesarean section at BC Women's Hospital, Vancouver, Canada. Ultrasound data were collected and processed to compute three QUS parameters, namely, attenuation coefficient estimate (ACE), integrated backscatter coefficient (IBC), and effective scatterer diameter (ESD) from the placentas. We developed a logistic regression model using QUS parameters as predictors. The primary outcome was the occurrence of SGA and PE. Results: The dataset consisted of 47 placentas, of which 25 placentas were complicated by SGA/PE. The final placenta-QUS model included quadratic and interaction terms of ACE, IBC, and ESD parameters. The placentaQUS model was well-calibrated, with a calibration slope of 0.99 (0.57-1.05) and a calibration intercept of 0.003 (-0.02- 0.22). The model predicted the SGA/PE complicated pregnancies with an apparent Area Under the Receive Operating Characteristic Curve (AUROC) of 0.89 (95 % CI: 0.78-0.98). The optimism-adjusted AUROC was 0.88 (95 % CI: 0.78-0.98). Discussion: A model for SGA and PE has been developed using QUS measures from the placenta ex vivo. The model showed promising performance in detecting SGA/PE. Future studies will be performed to assess the model measures in utero.

This paper introduces a hybrid modeling approach to accurately predict the performance of polymer-based Capacitive Micromachined Ultrasonic Transducers (polyCMUTs) by coupling finite element analysis (FEA) with analytical methods. The coupled FEA and analytical (CFA) model integrates characteristics from a single-cell FEA into a multi-cell equivalent circuit. Acoustic cross-coupling between cells is considered using analytical methods, and the acoustic far-field is computed via the Rayleigh integral. We validated the model on rectangular designs with 11x11 cells and varying cell-to-cell pitches. CFA results showed in average less than 7% deviation from full FEA in terms of center frequency, fractional bandwidth, and peak sensitivity, while requiring less than 1% of the computation time. We also observed good agreements with measurements, with a deviation of 17% for the rectangular designs and less than 4% for a larger linear array element (428 cells) we recently produced. This makes the CFA model a powerful tool for fast design exploration and optimization of CMUTs.

The objective of this work was to investigate changes in the acoustic characteristics of micromachined transducers caused by acoustic cross-coupling between cells. We used hexagonal, polymer-based capacitive micromachined ultrasonic transducers (polyCMUTs) consisting of 127 cells connected in parallel. The distances between the cells were varied, while the cell dimensions and number of cells remained constant. The resulting changes in characteristics were evaluated in terms of peak frequency fpk , fractional bandwidth FBW , peak transmit sensitivity Spk and opening angle Φ t . The study relies on results from an analytic multicell model (MCM) which considers cross-coupling effects between cells through a mutual acoustic impedance matrix. The results are compared with finite element (FE) analyses and measurements on fabricated prototypes. The manufacturing processes used to produce the polyCMUT prototypes are explained in detail. We found significant changes in all acoustic characteristics: as cell spacing increases, fpk and Φ t decrease, while Spk gradually rises to about twice the initial value. The FBW varies due to the change in fpk , peaking at small to intermediate cell-to-cell distances. While both modeling approaches cover the general effects, discrepancies in comparison to the measurements were identified. The FE model provided better fits than the analytic MCM, albeit at significantly higher computational costs. The effects on the acoustic characteristics were found strongest at lower frequencies and if many cells are in close proximity to each other. Hence, rotational symmetric or square transducers operating at lower frequencies are affected most. The results demonstrate that design approaches based on modeling single cells may lead to significant deviations from design goals. Both, analytic and FE models are suitable tools to estimate the effects of acoustic interactions and to predict the performance. This aids in meeting design requirements of micromachined ultrasound transducers consisting of multiple radiators.

Ultrasonic welding (UW) of thermoplastic composites (TCs) is an emerging technology in the field of composite joining techniques in the aerospace sector. Through a mechanical oscillator, ultrasound at a frequency of 20kHz is induced into the material via a welding horn, where microscopic friction and damping effects melt the thermoplastic. Under further pressure the weld area cools down, permanently joining both parts together. Like all joining processes in the aerospace industry the resulting joints need to be tested for their quality and structural integrity. The traditional testing method using water-coupled ultrasound includes extra steps. This process could be considerably improved by assessing the quality of the weld directly after or even during the welding process, allowing for immediate rework or discard of the parts in question. Ultrasound is still the best solution for this quality assessment, being inexpensive, well understood, and able to create B-Mode images, allowing a look into the cross-section of the weld. However, there are several major problems: To increase the system complexity as little as possible it is necessary to attach the ultrasound unit next to the welding equipment, and as close to the welding horn and compactor as possible to save space and keep the end- effector manoeuvrable. This brings problems for classic piezoelectric ultrasonic arrays: The low welding frequency and its resonance modes reach into the lower resonance modes of the piezoelectric sensors leading to immense noise, hiding any potential echo from the welding zone. Classic piezoelectric crystals are also very brittle and can suffer damage from sustained exposure to this violent environment. The authors present a novel solution: a custom-made polymer-based capacitive micromachined ultrasonic transducer array (polyCMUT). polyCMUTs are tiny drums with two electrodes. One on the bottom and the other suspended over a cavity sandwiched between two layers of polymers. By applying a DC-bias an electrical field is created and the membrane is set under tension. If then an AC voltage is applied, the strength of the electric field decreases, allowing the membrane to snap back into its original position. If done at the resonance frequency of the membrane, a strong ultrasonic signal is created. To receive this signal the polyCMUT is charged with a DC-bias, allowing it to receive the echo of the transmitted signal by measuring the changing capacitance. Not only is the polymer robust and inexpensive to fabricate, the general architecture of CMUTs also allows a design where the first mode of resonance is the actual mode the CMUT is operating in. By designing for a resonance frequency over 5 MHz all noise from the initial welding process is ignored, leading to a working pulse echo imaging system. The array is then mounted onto a PEEK block attached to the compactor unit of the welding end-effector. This publication is intended to present initial results, the design process of the custom array and the tests leading there.

Three-dimensional shear wave absolute vibro-elastography (S-WAVE) is a steady-state, volumetric elastography imaging technique similar to magnetic resonance elastography (MRE), with the additional advantage of multifrequency imaging and a significantly shorter examination time. We present a novel ultrasound matrix array implementation of S-WAVE for high-volume refresh rate acquisition. This new imaging setup is equipped with real-time shear wave monitoring for an improved data collection workflow and image quality. The image processing and elasticity reconstruction pipeline is tailored for high body mass index (BMI) subjects. We characterized this system with tissue phantoms and a human study cohort composed of 7 healthy volunteers and 25 patients with nonalcoholic fatty liver disease. The validation results show that S-WAVE can maintain a high agreement with the liver tissue stiffness measurements obtained with both the 2-D and 3-D MRE techniques, with an average cross correlation >93% and an average , which outperforms the conventional transient elasticity technique. Our findings show that the matrix array-based 3-D S-WAVE is a suitable volumetric elastography imaging solution for delivering a similar assessment of liver fibrosis as MRE in a more accessible, flexible, and cost-effective way.

Photoacoustic imaging (PAI) has emerged as a promising technique for various image guidance procedures. While convolutional neural networks (CNNs) trained on simulated radiofrequency (RF) data have been employed for point source reconstruction, their performance on real data remains a challenge. This paper addresses this limitation by introducing a novel deep learning-based method that utilizes a limited amount of experimental laser-diode-based data for the reconstruction of multiple point sources. The proposed approach employs a dual generative adversarial network (Dual-GAN) trained on experimental RF data from a combination of point source images. The Dual-GAN exhibits superior performance compared to the conventional delay-and-sum (DAS) method, demonstrating enhanced image contrast and a reduced full width at half maximum (FWHM). Notably, the axial and lateral localization errors of the Dual-GAN predictions surpass previous studies, measuring 0.028 +/- 0.018 mm and 0.087 +/- 0.096 mm, respectively. Additionally, the model demonstrates generalization capability by successfully reconstructing multiple point sources imaged using a different Nd:YAG laser system. This innovative method marks a significant advancement, offering improved accuracy and versatility in PAI applications involving multiple point sources.

Real-time 3-D photoacoustic (PA) imaging plays a significant role in volumetric imaging applications, such as breast imaging where PA has demonstrated significant potential. Challenges in 3-D PA imaging include long data acquisition time and limited compatibility with commonly used data acquisition systems. This paper introduces a new real-time 3-D PA data acquisition system using a matrix array transducer. Furthermore, we present a 3-D Delay and Glow (DAG) method for source localization that extends upon recently developed 2-D DAG. The experimental results show the functionality of the 3-D PA system. The DAG outperformed the conventional delay and sum (DAS) where axial, lateral, and elevational resolutions, respectively, are 0.06 +/- 0.00, 0.25 +/- 0.15, and 0.24 +/- 0.18mm for DAG and 0.14 +/- 0.06, 3.87 +/- 0.30, and 2.81 +/- 0.62mm for DAS.

This work presents a quantitative evaluation of polymer-based Capacitive Micromachined Ultrasonic Transducers (polyCMUTs) for 3D Ultrasound Computed Tomography (3D USCT). The study was motivated by limitations of the currently used PZT fiber technology in terms of bandwidth and transmit sensitivity. We developed finite element models of polyCMUT elements consisting of 127 cells to predict the acoustic performance. We fabricated prototype transducers using a novel method for microstructuring polymer layers. The produced samples reach a fractional bandwidth of 116%, an opening angle of 44° and increase the transmit sensitivity by 54%, compared to the PZT fiber transducers. The developed models allow for accurate predictions of the acoustic field over a large range of angles and frequencies. More work is required to improve the reliability and reduce sample-to-sample variations. Based on the measured performance and the general properties of the technology, polyCMUTs are very promising for 3D USCT.