Ultrasonic Imaging最新文献

In medical imaging, segmentation is a critical task for analysis and diagnosis. Deep learning-based segmentation has been actively studied and has shown remarkable performance. Building high-accuracy segmentation models requires a large amount of high-quality labeled data, but the cost of collecting such data is extremely high in medical imaging. In ultrasound imaging, the differences in image features depending on the equipment are significantly greater compared to other medical imaging modalities. Consequently, models need to be trained for each specific device, which entails substantial costs and time, leading to various practical challenges. To address these challenges, we propose a robust and accurate segmentation network that can operate independently of the ultrasound equipment. We integrated the Deep Frequency Filtering (DFF) module into a U-Net-based model. The proposed model retains the U-Net's encoder-decoder structure but applies frequency filtering within the latent space of each encoder layer, enabling adaptive selection of frequency components for breast tumor detection. Moreover, batch normalization was replaced with instance normalization to remove stylish features. We evaluated the model using three public datasets acquired from different scanners, achieving superior performance on unseen testing datasets compared to existing models. Notably, when tested on the unseen BUS-BRA dataset, DAUS-Net achieved a Dice score of 0.76, compared to 0.61 by the conventional U-Net. This improvement is attributed to the synergy between the DFF module and instance normalization. Our results demonstrate that the proposed model consistently detects and segments breast tumors, highlighting its potential for generalized clinical segmentation task. The source code for implementing DAUS-Net is publicly available at https://github.com/shlee8638/DAUS-Net.

Early detection of fetal cardiac diseases can dramatically improve neonatal outcomes by enabling timely intervention and informed clinical management. However, accurate diagnosis remains challenging due to the complexity of fetal heart structures in ultrasound images and the subtlety of congenital anomalies. To address these challenges, this work introduces Fetal Cardiac Disease Detection Using Ultrasound Imaging with Gegenbauer Graph Neural Networks (FCD-UI-GGNN). Ultrasound images are collected from the Fetal Phantom Ultrasound Dataset 23 (FPUS23) and pre-processed using Broad Collaborative Filtering (BCF) to resize images while preserving critical anatomical details. Fast Continual Multi-View Clustering (FCMVC) segments target vessel structures, and Gegenbauer Graph Neural Networks (GGNN) detects cardiac anomalies by modeling both local and global vessel relationships. The network weights are optimized using the Bitterling Fish Optimization Algorithm (BFOA) to improve accuracy. The framework is evaluated utilizing Accuracy, Precision, F1-Score, Recall, and ROC analysis, achieving 98.78% Accuracy, 99.01% Recall, 98.36% Precision, and 99.12% F1-Score. Validation on additional datasets, including FPUS23 and 4D Fetal Cardiac Ultrasound images, confirms robust generalization. These results demonstrate highly reliable and precise detection, supporting early clinical intervention for fetal cardiac anomalies.

The diagnostic problem of grading evaluation of ultrasonic images of Metacarpophalangeal rheumatoid arthritis (RA) is mostly dependent on the skills of sonographers with training. A grading system is used to identify and evaluate the geometric and textural features of bone deterioration and synovium thickening. In this manuscript, utilizing optimized mixed-order relation-aware recurrent neural network for metacarpophalangeal rheumatoid arthritis grading via ultrasound images (MRAG-UI-MORARNN-BWKA) is proposed. First, Tianjin University of Traditional Chinese Medicine's First Teaching Hospital provides the input ultrasound images. The pre-processing step uses confidence partitioning sampling filtering (CPSF) to resize the input images and eliminate background noise. Afterward, the pre-processed images were given to unpaired multi-view graph clustering (UMGC) for segmenting the region of interest (ROI). The holistic dynamic frequency transformer (HDFT) was used for extracting the geometric features like area, thickness, and shape. The Black winged kite algorithm (BWKA) was then employed to optimize the mixed-order relation-aware recurrent neural network (MORARNN) for precise grading of rheumatoid arthritis detection, with grades 0 (no synovium thickening), 1, 2, and 3 (mild, moderate, and severe, respectively). Python is used in the implementation of the proposed MRAG-UI-MORARNN-BWKA method. The proposed strategy achieves significant improvements over existing methods in grading rheumatoid arthritis via ultrasound images. The proposed model attains an accuracy of 97.02%, precision of 97.5% and sensitivity of 97.25%, respectively. These results clearly indicate the better performance and robustness of the proposed method analyzed to existing methods.

In medical ultrasound imaging, achieving high-quality reconstructed images while avoiding a huge computational burden is an important challenge. The Null subtraction imaging (NSI) algorithm results in a high-resolution reconstructed image. However, this method is not successful in recovering the background speckle information. In this paper, a novel algorithm, known as NSI-based generalized coherence factor (GCF)-along with delay-and-sum (DAS), which is abbreviated as NSG-DAS, is developed to overcome this limitation. In the proposed method, by using a hybrid technique, the desired resolution and effective noise suppression of the NSI algorithm, as well as the background speckle information of the conventional DAS beamformer are recovered simultaneously. More precisely, by using the GCF method, a new weighing factor is introduced that enhances the coherent regions of the image and suppresses the off-axis signals. Evaluations prove the favorable performance of the suggested technique; in particular, by using the proposed NSG-DAS method, a resolution comparable to the NSI algorithm is achieved for the geabr0 dataset, which is improved by about 42% compared to DAS. Also, the contrast evaluation parameter of the suggested technique is comparable to the DAS algorithm and is improved by about 63% compared to the NSI method. This indicates the ability of the suggested technique to improve either resolution or contrast simultaneously.

Speckle noise in ultrasound imaging remains a major obstacle to accurate clinical interpretation and reliable anatomical segmentation. Existing enhancement methods often compromise anatomical details while reducing noise, particularly under challenging imaging conditions. To address this, we introduce an innovative hybrid framework combining the Smart Adaptive Framework for Image Enhancement (SAFIE), a denoising engine based on adaptive fractional convolutions and gradient-based refinement, with a segmentation strategy integrating superpixel-based hypergraph modeling and neural ordinary differential equations. This framework enables effective noise suppression and precise segmentation of anatomical structures by capturing both spatial coherence and temporal feature dynamics. The enhanced images reveal improved visibility of anatomical structures and boundaries. Qualitative evaluation by four experienced radiologists confirmed this improvement, with strong inter-observer agreement measured by Fleiss' kappa, highlighting the robustness and clinical relevance of the approach. Quantitative results corroborate these observations, demonstrating performance substantially superior to several state-of-the-art methods. Ablation studies further indicate that each component contributes significantly to overall improvement. These findings suggest that the proposed framework enhances segmentation reliability and provides robust support for diagnostic interpretation in ultrasound imaging.

Ultrasound array probes can transmit diverging wavefronts from virtual source (VS) locations behind the array to obtain ultrafast compounded images with a broad field-of-view, but determining a practical set of diverging-wave VS locations is non-trivial, given the infinite half-plane of possibilities. In this work, we propose VS placement at a constant radial distance r from the array origin, and we compare this to a previous (and less direct) method of VS placement at a constant opening angle β relative to the ends of the array. Each method was implemented in Field II with a 64 element, 2.7 MHz phased-array geometry to simulate point-spread functions (PSFs) at regular 10 mm intervals over the field-of-view; the lateral and axial resolution, peak side-to-main lobe amplitude ratio (PSMR), and maximum amplitude of each PSF were measured. Each method was also implemented on a research scanner with a corresponding probe to acquire images of a tissue-mimicking phantom for comparison. Results from both methods in simulation and phantom experiments showed that the increase in PSF lateral resolution with range was consistent (≈38 µm/mm) and the mean axial resolution agreed within 0.01 mm; mean differences in PSMR and amplitude were <5% and <4%, respectively. Generalized contrast-to-noise ratio (gCNR) was highest for the constant-β₂ method, with differences between methods within ±1%. These results indicate that, relative to the constant-β method, comparable image quality can be achieved with a streamlined constant-r method of VS placement for diverging-wave ultrafast imaging.

Thyroid hormones are significant for controlling metabolism, and two common thyroid disorders, such as hypothyroidism. The hyperthyroidism are directly affect the metabolic rate of the human body. Predicting and diagnosing thyroid disease remain significant challenges in medical research due to the complexity of thyroid hormone regulation and its impact on metabolism. Therefore, this paper proposes a Complex-valued Spatio-Temporal Graph Convolution Neural Network optimized with Giraffe Kicking Optimization Algorithm for Thyroid Nodule Classification in Ultrasound Images (CSGCNN-GKOA-TNC-UI). Here, the ultrasound images are collected through DDTI (Digital Database of Thyroid ultrasound Imageries) dataset. The gathered data is given into the pre-processing stage using Bilinear Double-Order Filter (BDOF) approach to eradicate the noise and increase the input images quality. The pre-processing image is given into the Deep Adaptive Fuzzy Clustering (DAFC) for Region of Interest (RoI) segmentation. The segmented image is fed to the Multi-Objective Matched Synchro Squeezing Chirplet Transform (MMSSCT) for extracting the features, like Geometric features and Morphological features. The extracted features are fed into the CSGCNN, which classifies the Thyroid Nodule into Benign Nodules and Malign Nodules. Finally, Giraffe Kicking Optimization Algorithm (GKOA) is considered to enhance the CSGCNN classifier. The CSGCNN-GKOA-TNC-UI algorithm is implemented in MATLAB. The CSGCNN-GKOA-TNC-UI approach attains 34.9%, 21.5% and 26.8% higher f-score, 18.6%, 29.3 and 19.2% higher accuracy when compared with existing models: Thyroid diagnosis with classification utilizing DNN depending on hybrid meta-heuristic with LSTM method (LSTM-TNC-UI), innovative full-scale connected network for segmenting thyroid nodule in UI (FCG Net-TNC-UI), and Adversarial architecture dependent multi-scale fusion method for segmenting thyroid nodule (AMSeg-TNC-UI) methods respectively. The proposed model enhances thyroid nodule classification accuracy, aiding radiologists and endocrinologists. By reducing misclassification, it minimizes unnecessary biopsies and enables early malignancy detection.

The acoustic nonlinearity parameter (B/A) could enhance the diagnostic capabilities of conventional ultrasonography and quantitative ultrasound in tissues and diseases. Nonlinear acoustic propagation theory of plane waves has been used to develop a dual-energy model of the depletion of the fundamental related to the Gol'dberg number and subsequently to the B/A of media (a reference phantom is used as a baseline). The depletion method, however, needs a priori information of the attenuation coefficient (AC) of the assessed media. For this reason, recently, a work introduced a simultaneous estimator of the B/A and AC based on fitting depletion method measurements to a nonlinear model using the iterative algorithm Gauss-Newton Levenberg-Marquardt (GNLM). However, the GNLM method presented high sensitivity to the initial guess values of the algorithm which limits the robustness of the approach. In the present work, the Gauss-Newton method is combined with a total variation regularization approach (GNTV), which is achievable by expanding the nonlinear model of the GNLM method for joint estimation of the B/A and AC of all pixels of the parametric images instead of a block-wise approach. In addition, the GNTV used compounding data from several tone-burst transmissions at different center frequencies rather than only one narrowband tone-burst. The results suggest that incorporating regularization and increasing the number of frequencies improves the robustness of the GNTV compared to the GNLM method by accurately estimating B/A values in uniform and nonuniform experimental phantoms (mean relative error less than 18%). The best performance of B/A reconstruction was observed when the sample medium exhibited a constant Gol'dberg number.

Ultrasound imaging systems rely on accurate point spread function (PSF) estimation to support advanced image quality enhancement techniques such as deconvolution and speckle reduction. Phase aberration, caused by sound speed inhomogeneity within biological tissue, is inevitable in ultrasound imaging. It distorts the PSF by increasing sidelobe level and introducing asymmetric amplitude, making PSF estimation under phase aberration highly challenging. In this work, we propose a deep learning framework for estimating phase-aberrated PSFs using U-Net and complex U-Net architectures, operating on RF and complex k-space data, respectively, with the latter demonstrating superior performance. Synthetic phase aberration data, generated using the near-field phase screen model, is employed to train the networks. We evaluate various loss functions and find that log-compressed B-mode perceptual loss achieves the best performance, accurately predicting both the mainlobe and near sidelobe regions of the PSF. Simulation results validate the effectiveness of our approach in estimating PSFs under varying levels of phase aberration. Furthermore, we demonstrate that more accurate PSF estimation improves performance in a downstream phase aberration correction task, highlighting the broader utility of the proposed method.

To establish an ultrasound-based radiomics model to differentiate fibro adipose vascular anomaly (FAVA) and intramuscular venous malformation (VM). The clinical data of 65 patients with VM and 31 patients with FAVA who were treated and pathologically confirmed were retrospectively analyzed. Dimensionality reduction was performed on these features using the least absolute shrinkage and selection operator (LASSO). An ultrasound-based radiomics model was established using support vector machine (SVM) and random forest (RF) models. The diagnostic efficiency of this model was evaluated using the receiver operating characteristic. A total of 851 features were obtained by feature extraction, and 311 features were screened out using the t-test and Mann-Whitney U test. The dimensionality reduction was performed on the remaining features using LASSO. Finally, seven features were included to establish the diagnostic prediction model. In the testing group, the AUC, accuracy and specificity of the SVM model were higher than those of the RF model (0.841 [0.815-0.867] vs. 0.791 [0.759-0.824], 96.6% vs. 93.1%, and 100.0% vs. 90.5%, respectively). However, the sensitivity of the SVM model was lower than that of the RF model (88.9% vs. 100.0%). In this study, a prediction model based on ultrasound radiomics was developed to distinguish FAVA from VM. The study achieved high classification accuracy, sensitivity, and specificity. SVM model is superior to RF model and provides a new perspective and tool for clinical diagnosis.