Medical Ultrasound Imaging Research Articles

Combining ultrasound and elastography for the detection of a non-palpable, non-sonographically visualized Peyronie's plaques. Our experience.

B-mode ultrasound (US) medical imaging is very effective in localizing and describing Peyronie's disease (PD). Moreover, elastography is a new technique used to evaluate tissue elasticity to detect penile Peyronie's plaques that are not visible using standard B-mode US. The main objective of this study was to evaluate the diagnostic efficacy of real-time elastography (RTE) in PD patients and to determine whether its combined use with standard US improved diagnostic accuracy. RTE is also known as strain elastography (SE). Additionally, this study aimed to assess whether RTE was useful for monitoring PD patients undergoing conservative treatment. A group of 37 PD patients in the active phase was selected based on US examination showing isoechoic or hypo-isoechoic plaques, with or without associated hyperechoic or calcified plaque areas. All patients underwent traditional US combined with RTE before starting conservative treatment with antioxidants, during treatmentand after treatment. After each examination with RTE, a specific "Strain Ratio"(SR) was used to identify the specific elasticity of the tissue. Using B-mode US with RTE, we detected all 13 non-palpable penile plaques present in the 37 PD patients (100% of cases). Using only B-mode US, we detected only 8 of the 13 non-palpable plaques (61.5% of cases). The DI of the plaque decreased during and after treatment in all cases, indicating that RTE is effective for monitoring conservative PD treatment. A statistically significant correlation was found between the DI and plaque volume in all patients (p=0.002). Our study has shown that the combination of US and RTE methods allowed for a more accurate diagnosis in PD patients.

Reconfigurable ultrasound focusing effect through acoustic barriers

The low transmission efficiency of ultrasonic waves in waveguides of a high acoustic impedance (referred to as dense materials), due to the impedance mismatch between the background media and the dense materials, poses a significant obstacle to practical applications of high-intensity focused ultrasound (HIFU) such as ultrasound therapy or medical imaging. To address this challenge, we present an inverse optimization scheme for fabrication of novel acoustic meta-lenses, enabling strengthened penetration and enhanced focusing of ultrasonic waves when the waves traverse barriers. Both simulation and experiment validate the effectiveness of the developed meta-lenses which are annexed to hemispherical plates, and demonstrate an enhanced transmission of the sound power by an order of magnitude compared to a scenario without the use of the meta-lens. The focal distance is reconfigurable by adjusting the geometric parameters of the meta-lenses. The proposed design philosophy is not restricted by the complexity of the target structures, and it allows the ultrasonic waves to pass through acoustic barriers with a non-uniform thickness yet maintaining efficient wave focusing. This study holds appealing applications in HIFU-enabled ultrasound imaging and therapy.

Wavelet Transform and Thresholding Techniques in Medical Ultrasound Image Speckle Reduction

Wavelet Transform and Thresholding Techniques in Medical Ultrasound Image Speckle Reduction

Improved subarray coherence factor beamformer for ultrasound plane-wave imaging

Abstract Coherent plane-wave compounding (CPWC) has gained significant traction in medical ultrasound imaging. The coherence factor (CF) weighting algorithm, a classical adaptive beamforming strategy, aims to enhance the CPWC imaging lateral resolution, but it also noticeably compromises the speckle patterns, particularly.at a low signal-to-noise ratio (SNR). Delay multiply and sum (DMAS) has been proposed to reduce side lobes. However, the reconstructed image quality with DMAS is limited in terms of resolution and speckle level. In the present study, an improved subarray coherence factor (ISCF) algorithm is introduced by combining CF with DMAS strategies, aiming to improve the contrast and resolution while preserving the speckle level. The performance of the proposed algorithm is evaluated using simulation and phantom datasets. Results indicate that the resolution and the contrast of ISCF can be improved by 41.6% and 79.4%, respectively; compared with delay and sum (DAS) algorithm. Additionally, the resolution and the speckle level of ISCF can be improved by 10.6% and 35.7% compared with CF algorithm, respectively. The proposed ISCF algorithm outperforms CF in terms of resolution, contrast, and speckle preservation.

Modified Layer-based Ultrasound Imaging of Irregularly Cortical Bone

Abstract Cortical bone, known for its multilayered composition with a diverse sound velocity distribution and anisotropy dependent on anatomical position, poses a challenge in conventional medical ultrasound imaging. The limitations arise due to the inability of uniform velocity to appropriately capture the intricate properties of hard tissues, leading to inaccurate reconstruction of the inner cortical layers. To tackle this complexity, a modified layer-based method utilizing a 5-layer model and estimated sound velocity was introduced for irregularly cortical bone imaging. Validated through an ex-vivo study using goat tibia, this method showcased remarkable effectiveness. The experimental outcomes revealed an ultrasound-reconstructed cortical image that closely matched the μCT-measured ground truth model, demonstrating mean cortical thickness errors of 0.31 mm. This method stands as a validated and precise approach for accurately imaging cortical bone using a 5-layer model.

Millimeter-scale soft capsules for sampling liquids in fluid-filled confined spaces.

Sampling liquids in small and confined spaces to retrieve chemicals and microbiomes could enable minimally invasive monitoring human physiological conditions for understanding disease development and allowing early screening. However, existing tools are either invasive or too large for sampling liquids in tortuous and narrow spaces. Here we report a fundamental liquid sampling mechanism that enables millimeter-scale soft capsules for sampling liquids in confined spaces. The miniature capsule is enabled by flexible magnetic valves and superabsorbent polymer, fully wirelessly controlled for on-demand fluid sampling. A group of miniature capsules could navigate in fluid-filled and confined spaces safely using a rolling locomotion. The integration of on-demand triggering, sampling, and sealing mechanism and the agile group locomotion allows us to demonstrate precise control of the soft capsules, navigating and sampling body fluids in a phantom and animal organ ex vivo, guided by ultrasound and x-ray medical imaging. The proposed mechanism and wirelessly controlled devices spur the next-generation technologies for minimally invasive disease diagnosis.

US-Net: U-shaped network with Convolutional Attention Mechanism for ultrasound medical images

Ultrasound imaging, characterized by low contrast, high noise, and interference from surrounding tissues, poses significant challenges in lesion segmentation. To tackle these issues, we introduce an enhanced U-shaped network that incorporates several novel features for precise, automated segmentation. Firstly, our model utilizes a convolution-based self-attention mechanism to establish long-range dependencies in feature maps, crucial for small dataset applications, accompanied by a soft thresholding method for noise reduction. Secondly, we employ multi-sized convolutional kernels to enrich feature processing, coupled with curvature calculations to accentuate edge details via a soft-attention approach. Thirdly, an advanced skip connection strategy is implemented in the UNet architecture, integrating information entropy to assess and utilize texture-rich channels, thereby improving semantic detail in the encoder before merging with decoder outputs. We validated our approach using a newly curated dataset, VPUSI (Vascular Plaques Ultrasound Images), alongside the established datasets, BUSI, TN3K and DDTI. Comparative experiments on these datasets show that our model outperforms existing state-of-the-art techniques in segmentation accuracy.

Cable-Driven Light-Weighting and Portable System for Robotic Medical Ultrasound Imaging

Cable-Driven Light-Weighting and Portable System for Robotic Medical Ultrasound Imaging

A survey on deep learning in medical ultrasound imaging

Ultrasound imaging has a history of several decades. With its non-invasive, low-cost advantages, this technology has been widely used in medicine and there have been many significant breakthroughs in ultrasound imaging. Even so, there are still some drawbacks. Therefore, some novel image reconstruction and image analysis algorithms have been proposed to solve these problems. Although these new solutions have some effects, many of them introduce some other side effects, such as high computational complexity in beamforming. At the same time, the usage requirements of medical ultrasound equipment are relatively high, and it is not very user-friendly for inexperienced beginners. As artificial intelligence technology advances, some researchers have initiated efforts to deploy deep learning to address challenges in ultrasound imaging, such as reducing computational complexity in adaptive beamforming and aiding novices in image acquisition. In this survey, we are about to explore the application of deep learning in medical ultrasound imaging, spanning from image reconstruction to clinical diagnosis.

Multimodal breast cancer hybrid explainable computer-aided diagnosis using medical mammograms and ultrasound Images

Breast cancer is a prevalent global disease where early detection is crucial for effective treatment and reducing mortality rates. To address this challenge, a novel Computer-Aided Diagnosis (CAD) framework leveraging Artificial Intelligence (AI) techniques has been developed. This framework integrates capabilities for the simultaneous detection and classification of breast lesions. The AI-based CAD framework is meticulously structured into two pipelines (Stage 1 and Stage 2). The first pipeline (Stage 1) focuses on detectable cases where lesions are identified during the detection task. The second pipeline (Stage 2) is dedicated to cases where lesions are not initially detected. Various experimental scenarios, including binary (benign vs. malignant) and multi-class classifications based on BI-RADS scores, were conducted for training and evaluation. Additionally, a verification and validation (V&V) scenario was implemented to assess the reliability of the framework using unseen multimodal datasets for both binary and multi-class tasks. For the detection tasks, the recent AI detectors like YOLO (You Only Look Once) variants were fine-tuned and optimized to localize breast lesions. For classification tasks, hybrid AI models incorporating ensemble convolutional neural networks (CNNs) and the attention mechanism of Vision Transformers were proposed to enhance prediction performance. The proposed AI-based CAD framework was trained and evaluated using various multimodal ultrasound datasets (BUSI and US2) and mammogram datasets (MIAS, INbreast, real private mammograms, KAU-BCMD, and CBIS-DDSM), either individually or in merged forms. Visual t-SNE techniques were applied to visually harmonize data distributions across ultrasound and mammogram datasets for effective various datasets merging. To generate visually explainable heatmaps in both pipelines (stages 1 and 2), Grad-CAM was utilized. These heatmaps assisted in finalizing detected boxes, especially in stage 2 when the AI detector failed to automatically detect breast lesions. The highest evaluation metrics achieved for merged dataset (BUSI, INbreast, and MIAS) were 97.73% accuracy and 97.27% mAP50 in the first pipeline. In the second pipeline, the proposed CAD achieved 91.66% accuracy with 95.65% mAP50 on MIAS and 95.65% accuracy with 96.10% mAP50 on the merged dataset (INbreast and MIAS). Meanwhile, exceptional performance was demonstrated using BI-RADS scores, achieving 87.29% accuracy, 91.68% AUC, 86.72% mAP50, and 64.75% mAP50-95 on a combined dataset of INbreast and CBIS-DDSM. These results underscore the practical significance of the proposed CAD framework in automatically annotating suspected lesions for radiologists.

Coherence-based phase aberration correction and beamforming for ring-array ultrasound imaging

Degradation of image quality caused by aberrations and off-axis interferences is common in medical ultrasound imaging. The ring-array echo imaging also suffers from these constraints, where phase aberrations caused by tissue heterogeneity create overlapping of interest target and worsen the interpretability of anatomical structures; while off-axis interferences reduce the spatial resolution and contrast of the reconstructed images. To address them, the signal coherence is available since they can be considered as incoherent interference and noise. The coherence factor (CF) is sensitive to phase aberrations and is thus adopted in this study. Moreover, the convolutional beamforming algorithm (COBA) specifically exploits the autocorrelation operation to improve the signal coherence and effectively reduce noise. Therefore, we first utilize the CF to detect the local coherence and its maximization criterion as a way to estimate the sound speed to correct for phase aberration. Then, CF is used as an adaptive weighting factor to suppress the noise, and based on this, we develop a short-lag COBA to further enhance the image quality, referred to as the CF_SLCOBA method. Simulations and experiments were performed to evaluate the performance of the proposed methods. Results show that, the improvements of the CF_SLCOBA method in resolution is about 26.8% compared with CF; while about 92.0% in contrast ratio (CR) compared with COBA, in the experiments. Meanwhile, the proposed method yields a good performance in phase aberration correction. This demonstrates that the proposed method may provide benefits for rebuilding high-quality images, which also reveals that it offers a certain potential for practical applications.

Deep learning for ultrasound medical images: artificial life variant

Deep learning for ultrasound medical images: artificial life variant

The Instance Segmentation of Medical Ultrasound Images

The Instance Segmentation of Medical Ultrasound Images

Optimisation of the depth range of UIV on cohesive sediment (fluid mud) flows

Ultrasound Image Velocimetry (UIV) is a technique that enables the monitoring of flows in natural high-density cohesive sediments, also known as fluid mud. Using conventional medical ultrasound imaging equipment, the depth range of UIV in mud is however too limited. This manuscript discusses how this depth range can be improved based on the results of experimental UIV measurements in fluid mud. The results show that the depth range limitation is determined by the pixel brightness of the generated speckle pattern images. Since pixel brightness is a measure of the intensity of the backscattered signals, it degrades with increasing depth due to the attenuation of the ultrasound radiation by the mud. The rate at which ultrasound is attenuated by fluid mud is constant and is governed by the density of the mud and the applied ultrasound frequency. A greater depth range can thus be achieved by reducing attenuation and increasing the intensity of the emitted ultrasound. The former was confirmed by the results of experiments with variable mud density and ultrasound frequency. Flexibility of both parameters is however restricted, allowing only a limited increase in depth range. Another experiment confirmed an additional increase in depth range when a higher ultrasound intensity is applied. Patient safety measures of medical ultrasound imaging equipment, however restrict the intensity of the emitted ultrasound. To increase the depth range even further, potential solutions are therefore suggested, such as the use of high-power non-medical ultrasound imaging equipment or a novel procedure for processing the acquired acoustic data.

Value of Original and Generated Ultrasound Data Towards Training Robust Classifiers for Breast Cancer Identification

AbstractBreast cancer represents one of the leading causes of death among women, with 1 in 39 (around 2.5%) of them losing their lives annually, at the global level. According to the American Cancer Society, it is the second most lethal type of cancer in females, preceded only by lung cancer. Early diagnosis is crucial in increasing the chances of survival. In recent years, the incidence rate has increased by 0.5% per year, with 1 in 8 women at increased risk of developing a tumor during their life. Despite technological advances, there are still difficulties in identifying, characterizing, and accurately monitoring malignant tumors. The main focus of this article is on the computerized diagnosis of breast cancer. The main objective is to solve this problem using intelligent algorithms, that are built with artificial neural networks and involve 3 important steps: augmentation, segmentation, and classification. The experiment was made using a publicly available dataset that contains medical ultrasound images, collected from approximately 600 female patients (it is considered a benchmark). The results of the experiment are close to the goal set by our team. The final accuracy obtained is 86%.

S3Net: Semi-self-supervised neural network for visibility enhancement of speckled images

The tremendous growth of imaging sensors, which visually acquire the objects of interest, has attracted substantial interest in different engineering applications, such as underwater sonar imaging, synthetic aperture radar (SAR) imaging, and medical ultrasound imaging, etc. However, due to limited imaging environments, the intelligent vision-empowered visual information often suffers from the low usability of imaging data, e.g., Gamma-distributed speckle noise of unknown level in several imaging systems. It is commonly intractable to implement blind despeckling of visual data since the speckle noise is signal-dependent and multiplicative in nature. To improve the imaging quality, we tended to propose a semi-self-supervised neural network (termed S3Net) that enables intelligent blind despeckling in different practical applications. To be specific, our S3Net is mainly composed of two sub-networks, i.e., self-supervised feature extraction network (FExNet) and supervised feature enhancement network (FEnNet). The robustness and generalization abilities of S3Net can be accordingly guaranteed in different imaging scenarios. It is thus capable of extracting meaningful features and enhancing geometrical features under different receptive fields, leading to image quality improvement. The S3Net denoiser can balance the trade-off between noise suppression and detail preservation. Numerous experiments have been implemented on synthetic and realistic speckle-degraded imaging scenarios. Results have demonstrated that S3Net remarkably outperformed other state-of-the-art methods in terms of both qualitative and quantitative evaluations, with significant boosting on metric and visual performance.

Multi-apodization with cross-correlation combined with generalized sidelobe canceller applied to ultrasound imaging.

Beamforming is vital for medical ultrasound imaging systems. The generalized sidelobe canceller (GSC) beamforming can improve the image quality of lateral resolution, but its performance improvement in contrast and robustness is limited. This paper proposes an improved generalized sidelobe canceller algorithm based on multi-apodization with cross-correlation (MAXB-IGSC), which aims to improve the contrast and robustness of ultrasound imaging while maintaining the high image resolution and background speckle quality of GSC. The proposed MAXB-IGSC uses multiple pairs of complementary received apodization functions to process the echo data individually to obtain multiple pairs of data sets. The average of their normalized cross-correlation coefficients is then calculated and utilized to determine the adaptive subarray length of the GSC covariance matrix and weights the output of the improved GSC. The MAXB-IGSC improves the contrast ratio (CR) by 171.18% in anechoic cyst simulation and by 91.23%/130.97%/171.76% in geabr_0 (a dataset from the University of Michigan) experiment compared with GSC, respectively. Furthermore, MAXB-IGSC exhibits significant noise immunity, which greatly improves the robustness of the imaging. The technology also maintains the brightness and uniformity of the background speckle. The proposed MAXB-IGSC has potential for obtaining high-quality ultrasound images in clinical applications.

PSFHS: Intrapartum ultrasound image dataset for AI-based segmentation of pubic symphysis and fetal head

During the process of labor, the intrapartum transperineal ultrasound examination serves as a valuable tool, allowing direct observation of the relative positional relationship between the pubic symphysis and fetal head (PSFH). Accurate assessment of fetal head descent and the prediction of the most suitable mode of delivery heavily rely on this relationship. However, achieving an objective and quantitative interpretation of the ultrasound images necessitates precise PSFH segmentation (PSFHS), a task that is both time-consuming and demanding. Integrating the potential of artificial intelligence (AI) in the field of medical ultrasound image segmentation, the development and evaluation of AI-based models rely significantly on access to comprehensive and meticulously annotated datasets. Unfortunately, publicly accessible datasets tailored for PSFHS are notably scarce. Bridging this critical gap, we introduce a PSFHS dataset comprising 1358 images, meticulously annotated at the pixel level. The annotation process adhered to standardized protocols and involved collaboration among medical experts. Remarkably, this dataset stands as the most expansive and comprehensive resource for PSFHS to date.

Ultrasound image denoising autoencoder model based on lightweight attention mechanism.

The presence of noise in medical ultrasound images significantly degrades image quality and affects the accuracy of disease diagnosis. The convolutional neural network-denoising autoencoder (CNN-DAE) model extracts feature information by stacking regularly sized kernels. This results in the loss of texture detail, the over-smoothing of the image, and a lack of generalizability for speckle noise. A lightweight attention denoise-convolutional neural network (LAD-CNN) is proposed in the present study. Two different lightweight attention blocks (i.e., the lightweight channel attention (LCA) block and the lightweight large-kernel attention (LLA) block are concatenated into the downsampling stage and the upsampling stage, respectively. A skip connection is included before the upsampling layer to alleviate the problem of gradient vanishing during backpropagation. The effectiveness of our model was evaluated using both subjective visual effects and objective evaluation metrics. With the highest peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) values at all noise levels, the proposed model outperformed the other models. In the test of brachial plexus ultrasound images, the average PSNR of our model was 0.15 higher at low noise levels and 0.33 higher at high noise levels than the suboptimal model. In the test of fetal ultrasound images, the average PSNR of our model was 0.23 higher at low noise levels and 0.20 higher at high noise levels than the suboptimal model. The statistical analysis showed that the p values were less than 0.05, which indicated a statistically significant difference between our model and the other models. The results of this study suggest that the proposed LAD-CNN model is more efficient in denoising and preserving image details than both conventional denoising algorithms and existing deep-learning algorithms.

Enhanced high-frequency piezoelectric ultrasonic transducer based on integrated AlN honeycomb structure

Ultrasonic transducers are vital components for transmitting and receiving sound waves, which are customized to specific applications based on their operating frequencies.Small-sized high-frequency ultrasonic transducers have gained considerable attention and adoption in medical ultrasound imaging and small pipe flow testing. However, the significant impact of attenuation on high-frequency ultrasound restricts detection distance and degrades the signal-to-noise ratio. Additionally, thin-film high-frequency ultrasound transducers exhibit low output sound pressure, due to their low resonance displacement. In this study, we present an innovative approach to address these challenges through the design of an integrated small-size honeycomb-type top, two-electrode aluminum nitride (AIN)-based high-frequency ultrasonic transducer. A comprehensive investigation of the relationship between transducer size, sensitivity, and transmitted sound pressure levels is conducted. Leveraging microelectromechanical systems (MEMS) micro-nano-machining technology, we fabricate a 14 × 13 array with a unit diameter of 50 µm, and a matching circuit is employed to enhance the transducer's amplitude. Experimental results demonstrate remarkable transmit sound pressure levels (SPL) exceeding 210 dB (re:1 µPa/V) at a distance of 2 cm, with a maximum transmit SPL of 233 dB, and the receive sensitivity surpasses − 175 dB (re:1 V/µPa) and, at the 5 MHz resonance, the sensitivity reaches − 143 dB (re:1 V/µPa). Moreover, the transmit and receive linearity values demonstrate 99% and 98%, respectively. Our findings highlight the significant potential of the proposed high-frequency ultrasonic transducer for a diverse range of applications.