Ultrasonic Imaging最新文献

To evaluate the ability of two-dimensional ultrasound radiomics, integrated with clinical features, to predict central lymph node metastasis (CLNM) in papillary thyroid carcinoma (PTC). We conducted a retrospective study of PTC patients treated at the Second People's Hospital of Changzhou from January 2018 to February 2023. A total of 725 eligible patients were randomly allocated to training and test cohorts in a 7:3 ratio. Radiomic features were extracted from the PTC primary nodal region region on two-dimensional ultrasound images. Dimensionality reduction was performed using Mann-Whitney U tests, Spearman correlation analysis, and least absolute shrinkage and selection operator regression, yielding a radiomics signature (Rad-score). Seven machine-learning algorithms-logistic regression, support vector machine, k-nearest neighbors, decision tree, random forest, light gradient boosting machine, and gaussian naïve bayes-were compared to identify the optimal classifier. A joint predictive model was then constructed by integrating the Rad-score with clinically significant variables identified by univariate and multivariate logistic regression, and implemented using the optimal machine-learning classifier. Model performance was comprehensively evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis. Among the seven algorithms, gaussian naïve bayes achieved the highest predictive performance. Univariate and multivariate logistic regression revealed that sex, age, and tumor aspect ratio were independent predictors of CLNM. These variables were integrated with the Rad-score to yield a joint model that achieved AUCs of 0.840 (95% CI, 0.806-0.873) and 0.811 (95% CI, 0.746-0.866) in the training and test cohorts, respectively. Calibration curves and decision curve analysis indicated that the joint model was well-calibrated and afforded favorable clinical utility. The joint model integrating two-dimensional ultrasound radiomics with clinical features enables effective preoperative prediction of CLNM in PTC.

Accurate assessment of perfusion in vital organs like the pancreas is crucial for monitoring various pathologies, particularly tumors and their growth. While tumor growth inhibition usually results in decreased vascularization, current techniques for non-invasive and cost-effective perfusion assessment lack sufficient vessel separability for pancreatic applications, hindering optimal treatment selection and monitoring. Ultrasound (US) imaging offers advantages like low cost, rapid acquisition, non-invasiveness, and non-ionizing radiation. However, speckle, patient-related and acquisition-related motion artifacts, and limitations in distinguishing contrast-enhanced blood vessels, particularly in single frames, pose significant challenges. This study presents a novel solution utilizing image analysis of US contrast agents (UCAs) to characterize vascularization. The approach involves data denoising, selection of static frames, spatio-temporal registration, and deconvolution. The post-processed images are analyzed based on temporal intensity changes and normalized to extract trends. Data from 13 patients undergoing chemotherapeutic treatment with FOLFIRINOX or gemcitabine/abraxane were analyzed. Though no direct effects on vascularization are expected, the results suggest a correlation between derived vascularization trends (based on B-mode and CEUS data) and observed clinical treatment outcomes. Four patients exhibited negative slope (related to vascularization regression) aligned with clinical improvement, while six showed positive slope (related to increased vascularization) coinciding with treatment deterioration. Two patients displayed negative slopes without clinical improvement, and one patient displayed positive slope but had clinical improvement. These findings indicate the potential of this method to estimate treatment efficacy and guide personalized therapy, although the sample size is small and further investigation is warranted. Trial Registry Name: Sonoporation and Chemotherapy for the Treatment of Pancreatic Cancer. URL: https://clinicaltrials.gov/study/NCT04821284. Registration Number: NCT04821284.

This study aimed to investigate the diagnostic and differential value of high frequency ultrasound (HFUS), shear wave elastography (SWE), and superb microvascular imaging (SMI) for benign and malignant skin tumors, both individually and in combination. A total of 155 patients diagnosed with skin tumors via surgical treatment or puncture biopsy at the First Affiliated Hospital of Dalian Medical University were included in the study. The findings from HFUS, SWE, and SMI were recorded for each case. Pathological results served as the gold standard for comparing the differential diagnostic value of these parameters in benign and malignant skin tumors. Independent risk factors were further screened to construct a nomogram model, which was evaluated using receiver operating characteristic curves, decision curves, and calibration curves. Among the 155 patients with skin tumors, 107 were benign, and 48 were malignant. HFUS revealed significant differences in maximum diameter, internal echo, basal boundary, morphology, and blood flow grading between benign and malignant skin tumors (p < .05). In SWE, the maximum shear elastic modulus (E-max) of malignant skin tumors was significantly higher than that of benign tumors (p < .05). In SMI, the vascular index was significantly higher in the malignant group compared to the benign group (p < .001). Multivariate logistic regression identified boundary, maximum diameter, vascular index, and age as independent risk factors, leading to the development of a nomogram model. This model demonstrated an AUC of 0.935 (95% CI: 0.893-0.978), with a sensitivity of 85.4% and a specificity of 90.7%, indicating strong diagnostic value. HFUS, SWE, and SMI possess certain differential diagnostic capabilities for benign and malignant skin tumors. The nomogram model enhances the discrimination between benign and malignant tumors, providing precise diagnostic criteria and significant clinical relevance for the diagnosis of skin tumors.

Transient shear waves from push pulses can be used in elastography to estimate shear wave speed and attenuation, as a step towards the viscoelastic characterization of tissue. While many implementations are in use, less attention has been paid to practical issues of the strong influence of the inevitable background motions of tissue and transducer, the limited time and sampling available, and the deleterious effects of these on spectral estimates. To mitigate these issues, we propose several physics-based steps, first to correct for baseline drift and second to eliminate the need for Fourier transforms by completing all estimations on time domain energy. We target the estimation of shear wave attenuation, and preliminary results are shown for two phantoms and two in vivo livers to demonstrate the potential of this approach, which can serve as an alternative pathway towards shear wave attenuation of tissues for clinical assessment of tissue elastography.

In this paper, we describe the complete fabrication process of a 1D piezoelectric Micromachined Ultrasound Transducer (pMUT) array operating at 16 MHz underwater. We demonstrate the applicability of this pMUT Array in medical imaging using photoacoustic imaging (PAI) and ultrasound imaging (USI) experiments. There are 16 individual pMUT devices in the array, the radius of each device is 25 microns with a pitch of 100 microns (center-to-center). A 1-micron thick AlN (Aluminum Nitride) thin film is the piezoelectric material of choice for our pMUT array. This thin film was achieved by improving upon the control parameters in RF magnetron sputtering process. The working of this pMUT was validated by performing optical, electrical, and acoustic characterization. The 1D pMUT array was characterized optically using Laser Doppler Vibrometer (LDV) wherein the pMUT membrane showcased displacement of 6.2 pm/V for the in-air measurements at resonance of 20 MHz, the resonance frequency underwater was 16.2 MHz. Electrical characteristics were obtained through lock-in amplifier measurements, these were in close match to LDV results. Acoustical characteristics of the array was obtained through imaging experiments.

Conventional focused ultrasound imaging typically utilizes focused transmit beams in conjunction with a delay-and-sum (DAS) beamformer for reception, which yields optimal image quality only within the focal zone. To overcome this limitation, synthetic transmit aperture (STA) techniques and advanced non-linear beamforming methods are being explored to enhance the quality of ultrasound images. However, implementing these two approaches demands substantial computational resources. Although ultrasound systems utilizing GPU technology have demonstrated potential for real-time processing, their practical application is still limited. Real-time and affordable systems utilizing STA imaging and advanced non-linear beamforming are still not common in practical applications. Furthermore, the use of curvilinear array transducers is yet largely unexplored in the context of STA imaging. In this study, we present a GPU-based real-time affordable system for curvilinear array transducers that employs diverging beam with synthetic transmit aperture technique (DBSAT) imaging with non-linear beamforming. Experimental RF data were acquired using a tissue-mimicking phantom (CIRS Model 040GSE) with a DBSAT and conventional focused beamforming (CFB) sequence implemented on the programmable Verasonics Vantage 64 system equipped with a C5-2 curvilinear array probe. Beamforming was performed using an NVIDIA GeForce RTX 3060 GPU, implementing both DAS and filtered delay multiply and sum (FDMAS) algorithms. The results suggest that DBSAT-FDMAS using a curvilinear transducer yields improved image quality when the virtual source is positioned closer to the transducer, compared to an infinite virtual source distance. Further, reducing the number of receive elements has a minimal effect on image quality. The estimated axial and lateral resolutions for CFB-FDMAS range from 0.56 to 0.86 mm and 0.39 to 0.96 mm, respectively, whereas for DBSAT32-FDMAS, they range from 0.62 to 0.93 mm and 0.30 to 0.79 mm, respectively. The estimated CNR and gCNR values for CFB-FDMAS are 1.34 and 0.78, respectively, while those for DBSAT32-FDMAS are 1.84 and 0.81, respectively. In summary, DBSAT-FDMAS using 32 active receive elements offers enhanced image quality compared to CFB-FDMAS, while maintaining similar execution times.

Ultrasound simulation has become an essential tool for transducer design, optimizing imaging strategies, and validating image analysis techniques. A simulation method that accommodates tissue-specific scattering would significantly improve realism of insilico phantoms, generating much needed training data with ground truth information (on anatomy, motion, function) available. This study presents a novel framework for constructing 2-D numerical tissue phantoms based on histological microstructure, enabling accurate and realistic ultrasound simulations. Whole-slide histology images of adipose fat, carotid artery, muscle, and skin were segmented to extract collagen and cellular components. Relative acoustic heterogeneity was estimated for all tissues, which was combined with the segmentations to generate spatial maps of density and speed of sound. Ultrasound simulations were performed using a pseudospectral wave solver and validated against ex vivo data. Quantitative analysis using the Jensen-Shannon Divergence and a multi-level texture anisotropy index demonstrated significantly improved realism in speckle patterns compared to baseline isotropic phantoms. The numerical phantoms combined with computed tomography-based patient geometries show promising results for realistic ultrasound dataset generation.

Three-dimensional (3D) ultrasound vascular imaging (UVI) is essential for visualizing complex vascular structures. Row-column addressed (RCA) arrays, widely used for 3D UVI due to their hardware efficiency, suffer from point spread function (PSF) anisotropy, resulting in ramp-shaped noise that degrades image quality. Although existing denoising methods, including deep learning-based approaches, have shown promise, they are often limited by domain shift bias and the need for condition-specific data collection. Moreover, as full-volume 3D training is often impractical, many studies rely on 2D slice-wise training with 3D reconstruction, which can yield inter-slice intensity inconsistencies when slices are normalized independently. To overcome these limitations, we propose Robust Frequency-based Denoising Network (RFDNet), which integrates a Deep Frequency Filtering (DFF) module into a standard denoising model. The DFF module adaptively filters frequency components within the encoder, suppressing ramp-shaped noise while dynamically balancing spectral content to reduce sensitivity to domain shifts and inter-slice intensity inconsistencies. This adaptive filtering preserves vascular details and improves overall imaging consistency. Experiments on Doppler phantom, carotid artery, and abdominal datasets show that RFDNet significantly outperforms conventional methods in peak signal-to-noise ratio (PSNR), structural similarity (SSIM) and root mean squared error (RMSE). Further validation through 2D frequency spectrum analysis confirmed that the DFF module dynamically adjusts frequency components to maintain spectral balance. In addition, spectral KL divergence analysis demonstrated its robustness against inter-slice intensity inconsistencies introduced by slice-wise normalization. This approach improves domain generalization, reduces noise artifacts, and enhances clinical applicability by improving imaging reliability. Future work will explore 3D training and architectural refinements for better computational efficiency.

In recent years, convolutional neural network (CNN)-based optical flow models for motion estimation have been applied to radio-frequency (RF) ultrasound and B-mode (BM) data, demonstrating excellent performance. However, their architectures result in intricate network structures with a large number of parameters, posing challenges for deployment on resource-constrained devices. This paper proposes a novel approach that integrates dynamic pruning, knowledge distillation, and curriculum learning for model compression. The proposed method substantially reduces the complexity of deep learning models (i.e., memory demands and computational costs) while minimizing performance degradation. The teacher network was initially developed based on the Unsupervised Motion Estimation CNN (UMEN-Net). Subsequently, we developed a sub-network to reduce the number of parameters, referred to as DP-Net, and applied the proposed training techniques to obtain the final model, CDP-KDNet. The CDP-KDNet model was evaluated on simulated, phantom, and in vivo ultrasound data. Compared to DP-Net and other lightweight CNNs, CDP-KDNet achieves superior Signal-to-Noise Ratio (SNR) and Contrast-to-Noise Ratio (CNR) for axial strain estimation across all tested datasets. Its performance closely matches that of the teacher network while utilizing only 45.3% of the parameters and 67.8% of the floating-point operations. Additionally, as an unsupervised model, CDP-KDNet does not require ground-truth labels during training, rendering it a promising approach for ultrasound motion estimation.

We evaluated whether image-radiomics features extracted from ultrasound with integrated genomic data of single nucleotide polymorphisms (SNPs) associated with CIN susceptibility and clinical features could improve the differential diagnosis of high-grade intraepithelial disease (HSIL) and stage IA CC. Models were developed from ultrasound-derived radiomic features, SNPs data and clinical variables. After random 7:3 allocation into training and validation sets, clinical and SNPs datasets were each screened by univariable then multivariable logistic regression to build separate predictors. Ultrasound radiomics features were reduced with Max-relevance and min-redundancy (mRMR) and least absolute contraction and selection operator (LASSO) to generate an ultrasound radiomic score, which was subsequently used with clinical and SNPs data to establish the combined model. For the differentiation of HSIL and early CC models, only the ultrasound radiomics model showed higher classification efficiency, which the performance in the validation cohort (AUC: 0.885 [95%CI: 0.751-1.000]) than the method combining ultrasound radiomics score, clinical data and SNPs data with an AUC value of 0.850[95%CI: 0.713-0.987]. The model developed and constructed in this study, based on ultrasound radiomics, demonstrates potential for differentiating HSIL from Stage IA CC and exhibits significant clinical application value.