Ultrasound Image Reconstruction Research Articles

Nomogram for predicting cervical lymph node metastasis of papillary thyroid carcinoma using deep learning-based super-resolution ultrasound image

ObjectivesTo investigate the feasibility and effectiveness of a deep learning (DL) super-resolution (SR) ultrasound image reconstruction model for predicting cervical lymph node status in patients with papillary thyroid carcinoma(PTC).MethodsIn this retrospective study, researchers recruited 544 patients with PTC and randomly assigned them to training and test sets. SR ultrasound images were acquired using SR technology to improve image resolution, and artificial features and DL features were extracted from the original (OR) and SR images, respectively, to construct a ML, DL model. The best model was selected and aggregated with clinical parameters to construct the nomogram. The performance of the model is evaluated by ROC curves, calibration curves and decision curves.ResultsIn distinguishing the presence or absence of metastatic lymph nodes, the predictive performance of the SR_ResNet 101 and SR_SVM models based on SR outperformed those based on OR. In the test set, SR_SVM AUC was 0.878 (95% CI 0.8203–0.9358), accuracy 0.854, while OR_SVM AUC was 0.822 (95% CI 0.7500–0.8937), accuracy 0.665. SR_ResNet 101 AUC was 0.799 (95% CI 0.7175–0.8806), accuracy 0.793, and OR_ResNet101 AUC was 0.751 (95% CI 0.6620–0.8401), accuracy 0.713. Subsequently, Nomogram_A and Nomogram_B were constructed by integrating the SR_SVM model and SR_ResNet 101 model, respectively, with clinical parameters, while Nomogram_C was constructed solely based on clinical indicators. In the test set, Nomogram_A demonstrated the best performance with an AUC of 0.930 (95% CI 0.8913–0.9682) and accuracy was 0.829. Nomogram_B AUC 0.868 (95% CI 0.8102–0.9261) and accuracy was 0.829, while Nomogram_C AUC 0.880 (95% CI 0.8257–0.9349) and accuracy was 0.787. The DeLong test revealed that the diagnostic performance of Nomogram_A based on SR_SVM was significantly higher than that of Nomogram_B, Nomogram_C, and the level of Radiologist (P < 0.05). The calibration curves and Hosmer–Lemeshow tests confirmed a high degree of fit, and the decision curve analysis demonstrated clinical value and potential patient benefit.ConclusionsThe predictive model constructed using SR reconstructed ultrasound images demonstrated superior performance in predicting preoperative cervical lymph node metastasis in PTC compared to OR images. The nomogram prediction model based on SR images has the potential to enhance the accuracy of predictive models and aid in clinical decision-making.

Non-iterative model-based inversion for low channel-count optical ultrasound imaginga).

Ultrasound image reconstruction is typically performed using the computationally efficient delay-and-sum algorithm. However, this algorithm is suboptimal for systems of low channel counts, where it causes significant image artefacts. These artefacts can be suppressed through model-based inversion approaches; however, their computational costs typically prohibit real-time implementations. In this work, the emerging optical ultrasound (OpUS) modality is considered, where ultrasound waves are both generated and detected using light. With this modality, imaging probes comprise very low channel counts, resulting in significant image artefacts that limit the imaging dynamic range. However, this low channel counts offer an opportunity for non-iterative ("direct") model-based inversion (DMI) on modest computational resources available in a typical workstation. When applied to both synthetic and experimental OpUS data, the presented DMI method achieved substantial reduction in image artefacts and noise, improved recovery of image amplitudes, and-after one-off pre-computation of the system matrices-significantly reduced reconstruction time, even in imaging scenarios exhibiting mild spatial inhomogeneity. Whilst here applied to an OpUS imaging system, DMI can be applied to other low channel-count imaging systems, and is therefore expected to achieve better image quality, reduce system complexity, or both, in a wide range of settings.

A Power Law Reconstruction of Ultrasound Backscatter Images

Ultrasound B-scan images are traditionally formed from the envelope of the received radiofrequency echoes, but the image texture is dominated by granular speckle patterns. Longstanding efforts at speckle reduction and deconvolution have been developed to lessen the detrimental aspects of speckle. However, we now propose an alternative approach to estimation (and image rendering) of the underlying fine grain scattering density of tissues based on power law constraints. The key steps are a whitening of the spectrum of the received signal while conforming to the original envelope shape and statistics, followed by a power law filtering in accordance with the known scattering behavior of tissues. This multiple step approach results in a high-spatial-resolution map of scattering density that is constrained by the most important properties of scattering from tissues. Examples from in vivo liver scans are shown to illustrate the change in image properties from this framework.

Super-resolution reconstruction of ultrasound image using a modified diffusion model

Objective. This study aims to perform super-resolution (SR) reconstruction of ultrasound images using a modified diffusion model, designated as the diffusion model for ultrasound image super-resolution (DMUISR). SR involves converting low-resolution images to high-resolution ones, and the proposed model is designed to enhance the suitability of diffusion models for this task in the context of ultrasound imaging. Approach. DMUISR incorporates a multi-layer self-attention (MLSA) mechanism and a wavelet-transform based low-resolution image (WTLR) encoder to enhance its suitability for ultrasound image SR tasks. The model takes interpolated and magnified images as input and outputs high-quality, detailed SR images. The study utilized 1,334 ultrasound images from the public fetal head-circumference dataset (HC18) for evaluation. Main results. Experiments were conducted at 2× , 4× , and 8× magnification factors. DMUISR outperformed mainstream ultrasound SR methods (Bicubic, VDSR, DECUSR, DRCN, REDNet, SRGAN) across all scales, providing high-quality images with clear structures and rich detailed textures in both hard and soft tissue regions. DMUISR successfully accomplished multiscale SR reconstruction while suppressing over-smoothing and mode collapse problems. Quantitative results showed that DMUISR achieved the best performance in terms of learned perceptual image patch similarity, with a significant decrease of over 50% at all three magnification factors (2× , 4× , and 8× ), as well as improvements in peak signal-to-noise ratio and structural similarity index measure. Ablation experiments validated the effectiveness of the MLSA mechanism and WTLR encoder in improving DMUISR’s SR performance. Furthermore, by reducing the number of diffusion steps, the computational time of DMUISR was shortened to nearly one-tenth of its original while maintaining image quality without significant degradation. Significance. This study demonstrates that the modified diffusion model, DMUISR, provides superior performance for SR reconstruction of ultrasound images and has potential in improving imaging quality in the medical ultrasound field.

Optimization of array encoding for ultrasound imaging

Objective. The transmit encoding model for synthetic aperture imaging is a robust and flexible framework for understanding the effects of acoustic transmission on ultrasound image reconstruction. Our objective is to use machine learning (ML) to construct scanning sequences, parameterized by time delays and apodization weights, that produce high-quality B-mode images. Approach. We use a custom ML model in PyTorch with simulated RF data from Field II to probe the space of possible encoding sequences for those that minimize a loss function that describes image quality. This approach is made computationally feasible by a novel formulation of the derivative for delay-and-sum beamforming. Main results. When trained for a specified experimental setting (imaging domain, hardware restrictions, etc), our ML model produces optimized encoding sequences that, when deployed in the REFoCUS imaging framework, improve a number of standard quality metrics over conventional sequences including resolution, field of view, and contrast. We demonstrate these results experimentally on both wire targets and a tissue-mimicking phantom. Significance. This work demonstrates that the set of commonly used encoding schemes represent only a narrow subset of those available. Additionally, it demonstrates the value for ML tasks in synthetic transmit aperture imaging to consider the beamformer within the model, instead of purely as a post-processing step.

Computer-Aided Diagnosis of Duchenne Muscular Dystrophy Based on Texture Pattern Recognition on Ultrasound Images Using Unsupervised Clustering Algorithms and Deep Learning

ObjectiveThe feasibility of using deep learning in ultrasound imaging to predict the ambulatory status of patients with Duchenne muscular dystrophy (DMD) was previously explored for the first time. The present study further used clustering algorithms for the texture reconstruction of ultrasound images of DMD data sets and analyzed the difference in echo intensity between disease stages. Methodsk-means (Kms) and fuzzy c-means (FCM) clustering algorithms were used to reconstruct the DMD data-set textures. Each image was reconstructed using seven texture-feature categories, six of which were used as the primary analysis items. The task of automatically identifying the ambulatory function and DMD severity was performed by establishing a machine-learning model. ResultsThe experimental results indicated that the Gaussian Naïve Bayes and k-nearest neighbors classification models achieved an accuracy of 86.78% in ambulatory function classification. The decision-tree model achieved an identification accuracy of 83.80% in severity classification. A deep convolutional neural network model was established as the main structure of the deep-learning model while automatic auxiliary interpretation tasks of ambulatory function and severity were performed, and data augmentation was used to improve the recognition performance of the trained model. Both the visual geometry group (VGG)-16 and VGG-19 models achieved 98.53% accuracy in ambulatory-function classification. The VGG-19 model achieved 92.64% accuracy in severity classification. ConclusionRegarding the overall results, the Kms and FCM clustering algorithms were used in this study to reconstruct the characteristic texture of the gastrocnemius muscle group in DMD, which was indeed helpful in quantitatively analyzing the deterioration of the gastrocnemius muscle group in patients with DMD at different stages. Subsequent combination of machine-learning and deep-learning technologies can automatically and accurately assist in identifying DMD symptoms and tracking DMD deterioration for long-term observation.

Efficient methods for iterative ultrasound image reconstruction using L1 and L2 norm regularization

Ultrasound image reconstruction can be posed as an inverse problem, with image pixels as parameters to a model of wave propagation that are estimated from raw ultrasound channel data. Such framework allows one to make assumptions about channel signals and image properties in form of regularization that can be used to improve reconstruction accuracy in the presence of electronic and acoustic noise. A simple model assumes spherical wave propagation from point sources in a homogeneous medium, similar to delay-and-sum (DAS) beamforming; solving such problems numerically can require many iterations and can result in blurred images when traditional L2 norm regularization is used. We use Fast Iterative Shrinkage Thersholding Algorithm (FISTA) to reduce the number of iterations to less than 10, and to apply L1-norm regularization to the data, which improves edge sharpness and image contrast. We demonstrate the concept using FIELD II simulated ultrasound signals from point, speckle, and anechoic targets with controlled levels of electronic and speckle noise. The FISTA-reconstructed point targets show reduced sidelobe levels by 15 dB compared to the traditional DAS image, and the reduction in full-width at half maximum by a factor of 2. We also implement iterative reconstruction in omega-k frequency domain, which allows factoring the forward and adjoint operators at each spatial frequency and paves the way for an intuitive and more memory efficient, multi-threaded implementation of the method.

Improving the Automated Diagnosis of Breast Cancer with Mesh Reconstruction of Ultrasound Images Incorporating 3D Mesh Features and a Graph Attention Network

This study proposes a novel approach for breast tumor classification from ultrasound images into benign and malignant by converting the region of interest (ROI) of a 2D ultrasound image into a 3D representation using the point-e system, allowing for in-depth analysis of underlying characteristics. Instead of relying solely on 2D imaging features, this method extracts 3D mesh features that describe tumor patterns more precisely. Ten informative and medically relevant mesh features are extracted and assessed with two feature selection techniques. Additionally, a feature pattern analysis has been conducted to determine the feature’s significance. A feature table with dimensions of 445 × 12 is generated and a graph is constructed, considering the rows as nodes and the relationships among the nodes as edges. The Spearman correlation coefficient method is employed to identify edges between the strongly connected nodes (with a correlation score greater than or equal to 0.7), resulting in a graph containing 56,054 edges and 445 nodes. A graph attention network (GAT) is proposed for the classification task and the model is optimized with an ablation study, resulting in the highest accuracy of 99.34%. The performance of the proposed model is compared with ten machine learning (ML) models and one-dimensional convolutional neural network where the test accuracy of these models ranges from 73 to 91%. Our novel 3D mesh-based approach, coupled with the GAT, yields promising performance for breast tumor classification, outperforming traditional models, and has the potential to reduce time and effort of radiologists providing a reliable diagnostic system.

Flexible Ultrasound Transducer With Embedded Optical Shape Sensing Fiber for Biomedical Imaging Applications.

Flexible ultrasound transducers (FUTs), capable of conforming to irregular surfaces, have become a research hotspot in the field of medical imaging. With these transducers, high-quality ultrasound images can be obtained only if strict design criteria are fulfilled. Moreover, the relative positions of array elements must be determined, which are important for ultrasound beamforming and image reconstruction. These two major characteristics present great challenges to the design and fabrication of FUTs compared to that for traditional rigid probes. In this study, an optical shape-sensing fiber was embedded into a 128-element flexible linear array transducer to acquire the real-time relative positions of array elements to produce high-quality ultrasound images. Minimum concave and convex bend diameters of approximately 20 and 25 mm, respectively, were achieved. The transducer was flexed 2000 times, and yet no obvious damage was observed. Stable electrical and acoustic responses confirmed its mechanical integrity. The developed FUT exhibited an average center frequency of 6.35 MHz, and average -6-dB bandwidth of 69.2%. The array profile and element positions measured by the optic shape-sensing system were instantly transferred to the imaging system. Phantom experiments for both spatial resolution and contrast-to-noise ratio proved that FUTs can maintain satisfactory imaging capability despite bending to sophisticated geometries. Finally, color Doppler images and Doppler spectra of the peripheral arteries of healthy volunteers were obtained in real time.

Application of a nerve stereoscopic reconstruction technique based on ultrasonic images in the diagnosis of neuralgic amyotrophy.

Objective: To propose a nerve stereoscopic reconstruction technique based on ultrasound imaging for site diagnosis, intuitive reflection of disease severity, and classification of neuralgic amyotrophy (NA). Methods: We enrolled 44 patients with NA who underwent high-frequency ultrasonography examination. Multiple sites on the normal side and the affected side were scanned to calculate the ratio of the cross-section area (CSA) of the affected side to the normal side at each location measured, i.e., the cross-section area swelling ratio (CSASR). The CSASR of 44 patients and 30 normal controls was analyzed to determine their threshold value for the diagnosis of NA. Then, ultrasound images of the cross-section were used to reconstruct the stereoscopic model of the nerve on the affected side and the normal side. Using the CSASR values in each measurement location, a CSASR stereoscopic model was developed. Results: The threshold value of CSASR for ultrasound diagnosis of NA was 1.55. The average diseased segments per patient was 2.49 ± 1.97, with an average overall length of 10.03 ± 7.95cm. Nerve stereoscopic reconstruction could be conducted for swelling, torsion, incomplete constriction, and complete constriction. Conclusion: The ultrasound image reconstruction method proposed in this study can accurately determine the site, range, and type of neuropathies in patients with NA, and simultaneously provide complete and accurate data information and intuitive morphological information.

P07.04.A ELASTIC IMAGE FUSION OF INTRAOPERATIVE ULTRASOUND AND PREOPERATIVE MRI FOR IMPROVING NAVIGATION ACCURACY IN BRAIN TUMOR SURGERY

Abstract BACKGROUND In brain surgery, neuronavigation based on preoperative MRI is limited by various physical and operational factors, such as surgery-induced brain shift. Using navigated intraoperative ultrasound (iUS) imaging and 3D iUS image reconstruction may help improve neuronavigation. However, the registration of MRI and US scans can lead to spatial inconsistencies between the two modalities. This study aims to evaluate the clinical integration of navigated iUS imagining and its potential to improve the accuracy of resection during brain tumor surgery by comparing the performance of rigid and elastic MRI-iUS image fusion algorithms during retrospective evaluation. MATERIAL AND METHODS N = 50 patients with brain tumors were prospectively enrolled in this trial. The neuronavigation system (BrainLab AG) used in the study allowed for the reconstruction and integration of 3D iUS imaging, which was acquired multiple times throughout the surgery using a tracked ultrasound probe (BK Medical). Prototypes of automatic rigid and elastic MRI-iUS image fusion algorithms (Brainlab Elements Image Fusion) were retrospectively applied to improve co-registration accuracy between the MRI and iUS data. To evaluate the image co-registration accuracy, anatomical landmarks were defined in the MRI and iUS scans and used to measure Euclidean distances between each landmark pair. RESULTS Improvements in the image quality over the course of the study indicated that skills in US handling are essential. Contrast behavior and tumor structure influence the visibility of the tumor in the US. Artifacts caused by hemostatic agents challenge the US image quality and the fluid-filled resection cavity requires experience in probe handling. In terms of image fusion, significant mismatches of the iUS and MRI data were reported. For scans acquired before resection, rigid fusion using a novel co-registration algorithm may enhance the registration accuracy with mean target registration errors for two representative patients of 6.5/8.9 mm and 3.7/3.4 mm before and after registration update, respectively. For iUS scans acquired after tumor resection, rigid fusion is inherently limited due to non-linear brain shifts. Therefore, elastic image fusion has been applied to provide deformable MR-iUS image co-registration. CONCLUSION Navigated iUS imaging can be integrated into clinical routine and the use of MR-iUS image co-registration for updating navigation is a promising tool to compensate brainshift during surgical resection. However, further studies are needed to quantify the accuracy and assess whether iUS could potentially replace the use of iMRI in resection control of selected cases.

ARU-GAN: U-shaped GAN based on Attention and Residual connection for super-resolution reconstruction

Plane-wave ultrasound imaging technology offers high-speed imaging but lacks image quality. To improve the image spatial resolution, beam synthesis methods are used, which often compromise the temporal resolution. Herein, we propose ARU-GAN, a super-resolution reconstruction model based on residual connectivity and attention mechanisms, to address this issue. ARU-GAN comprises a Full-scale Skip-connection U-shaped Generator (FSUG) with an attention mechanism and a Residual Attention Patch Discriminator (RAPD). The former captures global and local features of the image by using full-scale skip-connections and attention mechanisms. The latter focuses on changes in the image at different scales to enhance its discriminative ability at the patch level. ARU-GAN was trained using a combined loss function on the Plane-Wave Imaging Challenge in Medical Ultrasound (PICMUS) 2016 dataset, which includes three types of targets: point targets, cyst targets, and in-vivo targets. Compared to Coherent Plane-Wave Compounding (CPWC), ARU-GAN achieved a reduction in Full Width at Half Maximum (FWHM) by 5.78%–20.30% on point targets, improved Contrast (CR) by 7.59–11.29 percentage points, and Contrast to Noise Ratio (CNR) by 30.58%–45.22% on cyst targets. On in-vivo target, ARU-GAN improved the Peak Signal-to-Noise Ratio (PSNR) by 11.94%, the Complex-Wavelet Structural Similarity Index Measurement (CW-SSIM) by 17.11%, and the Normalized Cross Correlation (NCC) by at least 2.17% compared to existing deep learning methods. In conclusion, ARU-GAN is a promising model for the super-resolution reconstruction of plane-wave medical ultrasound images. It provides a novel solution for improving image quality, which is essential for clinical practice.

Three-dimensional ultrasound image reconstruction based on 3D-ResNet in the musculoskeletal system using a 1D probe: ex vivo and in vivo feasibility studies

Objective. Three-dimensional (3D) ultrasound (US) is needed to provide sonographers with a more intuitive panoramic view of the complex anatomical structure, especially the musculoskeletal system. In actual scanning, sonographers may perform fast scanning using a one-dimensional (1D) array probe .at random angles to gain rapid feedback, which leads to a large US image interval and missing regions in the reconstructed volume. Approach. In this study, a 3D residual network (3D-ResNet) modified by a 3D global residual branch (3D-GRB) and two 3D local residual branches (3D-LRBs) was proposed to retain detail and reconstruct high-quality 3D US volumes with high efficiency using only sparse two-dimensional (2D) US images. The feasibility and performance of the proposed algorithm were evaluated on ex vivo and in vivo sets. Main r esults. High-quality 3D US volumes in the fingers, radial and ulnar bones, and metacarpophalangeal joints were obtained by the 3D-ResNet, respectively. Their axial, coronal, and sagittal slices exhibited rich texture and speckle details. Compared with kernel regression, voxel nearest-neighborhood, squared distance weighted methods, and a 3D convolution neural network in the ablation study, the mean peak-signal-to-noise ratio and mean structure similarity of the 3D-ResNet were up to 28.53 ± 1.29 dB and 0.98 ± 0.01, respectively, and the corresponding mean absolute error dropped to 0.023 ± 0.003 with a better resolution gain of 1.22 ± 0.19 and shorter reconstruction time. Significance. These results illustrate that the proposed algorithm can rapidly reconstruct high-quality 3D US volumes in the musculoskeletal system in cases of a large amount of data loss. This suggests that the proposed algorithm has the potential to provide rapid feedback and precise analysis of stereoscopic details in complex and meticulous musculoskeletal system scanning with a less limited scanning speed and pose variations for the 1D array probe.

High-Resolution Ultrasound Characterization of Local Scattering in Cancer Tissue.

Ultrasound (US) has afforded an approach to tissue characterization for more than a decade. The challenge is to reveal hidden patterns in the US data that describe tissue function and pathology that cannot be seen in conventional US images. Our group has developed a high-resolution analysis technique for tissue characterization termed H-scan US, an imaging method used to interpret the relative size of acoustic scatterers. In the present study, the objective was to compare local H-scan US image intensity with registered histologic measurements made directly at the cellular level. Human breast cancer cells (MDA-MB 231, American Type Culture Collection, Manassas, VA, USA) were orthotopically implanted into female mice (N=5). Tumors were allowed to grow for approximately 4 wk before the study started. In vivo imaging of tumor tissue was performed using a US system (Vantage 256, Verasonics Inc., Kirkland, WA, USA) equipped with a broadband capacitive micromachined ultrasonic linear array transducer (Kolo Medical, San Jose, CA, USA). A 15-MHz center frequency was used for plane wave imaging with five angles for spatial compounding. H-scan US image reconstruction involved use of parallel convolution filters to measure the relative strength of backscattered US signals. Color codes were applied to filter outputs to form the final H-scan US image display. For histologic processing, US imaging cross-sections were carefully marked on the tumor surface, and tumors were excised and sliced along the same plane. By use of optical microscopy, whole tumor tissue sections were scanned and digitized after nuclear staining. US images were interpolated to have the same number of pixels as the histology images and then spatially aligned. Each nucleus from the histologic sections was automatically segmented using custom MATLAB software (The MathWorks Inc., Natick, MA, USA). Nuclear size and spacing from the histology images were then compared with local H-scan US image features. Overall, local H-scan US image intensity exhibited a significant correlation with both cancer cell nuclear size (R2 > 0.27, p < 0.001) and the inverse relationship with nuclear spacing (R2 > 0.17, p < 0.001).

Ultrasound Image Segmentation and Classification of Benign and Malignant Thyroid Nodules on the Basis of Deep Learning

This study aimed to investigate the effect of an image denoising algorithm based on weighted low-rank matrix restoration on thyroid nodule ultrasound images. A total of 1000 original ultrasound image data sets of thyroid nodules were selected as the study samples. The nodule segmentation data set of thyroid ultrasound region of interest (ROI) images was drawn and acquired. By introducing multiscale features and an attention mechanism to optimize the U-Net model, an ultrasound image segmentation model (F-U-Net) was constructed. The performance of the traditional U network model and full convolutional neural network model (FCN) was analyzed and compared by simulation experiments. The results showed that the dice coefficient, accuracy, and recall of the improved loss function in this study were significantly higher than those of the traditional cross entropy loss function and dice coefficient loss function, and the differences were statistically significant (P &lt; 0.05). The Dice coefficient, accuracy, and recall of the F-U-net model were significantly higher than those of the traditional FCN model and U-net model (P &lt; 0.05). The diagnostic sensitivity, specificity, accuracy, and positive predictive value of the F-U-net model for benign and malignant thyroid nodules were significantly higher than those of the FCN model and U-net model (P &lt; 0.05). In summary, the proposed F-U network can effectively process the ultrasound images of thyroid nodules, improve the image quality, and help to improve the diagnostic effect of benign and malignant thyroid nodules. It provides a data reference for segmentation and reconstruction of benign and malignant ultrasound images of thyroid nodules.

Acoustic Velocity Measurement for Enhancing Laser UltraSound Imaging Based on Time Domain Synthetic Aperture Focusing Technique

A method to enhance laser ultrasound (LUS) image reconstruction with the time-domain synthetic aperture focusing technique (T-SAFT) is presented, in which the acoustic velocity is extracted in situ with curve fitting. The operational principle is provided with the help of a numerical simulation, and the confirmation is provided experimentally. In these experiments, an all-optic LUS system was developed by using lasers for both excitation and detection of ultrasound. The acoustic velocity of a specimen was extracted in situ by fitting a hyperbolic curve to its B-scan image. The needle-like objects embedded within a polydimethylsiloxane (PDMS) block and a chicken breast have been successfully reconstructed using the extracted in situ acoustic velocity. Experimental results suggest that knowing the acoustic velocity in the T-SAFT process is important not only in finding the depth location of the target object but also for producing a high resolution image. This study is expected to pave the wave to the development and usage of all-optic LUS for bio-medical imaging.

Simulation-to-real generalization for deep-learning-based refraction-corrected ultrasound tomography image reconstruction

Objective. The image reconstruction of ultrasound computed tomography is computationally expensive with conventional iterative methods. The fully learned direct deep learning reconstruction is promising to speed up image reconstruction significantly. However, for direct reconstruction from measurement data, due to the lack of real labeled data, the neural network is usually trained on a simulation dataset and shows poor performance on real data because of the simulation-to-real gap. Approach. To improve the simulation-to-real generalization of neural networks, a series of strategies are developed including a Fourier-transform-integrated neural network, measurement-domain data augmentation methods, and a self-supervised-learning-based patch-wise preprocessing neural network. Our strategies are evaluated on both the simulation dataset and real measurement datasets from two different prototype machines. Main results. The experimental results show that our deep learning methods help to improve the neural networks’ robustness against noise and the generalizability to real measurement data. Significance. Our methods prove that it is possible for neural networks to achieve superior performance to traditional iterative reconstruction algorithms in imaging quality and allow for real-time 2D-image reconstruction. This study helps pave the path for the application of deep learning methods to practical ultrasound tomography image reconstruction based on simulation datasets.

A Unifying Approach to Inverse Problems of Ultrasound Beamforming and Deconvolution

Beamforming is an essential step in the ultrasound image formation pipeline and has recently attracted growing interest. An important goal of beamforming is to increase the image spatial resolution, or in other words to narrow down the system Point Spread Function (PSF). In parallel to beamforming approaches, deconvolution methods have also been explored in ultrasound imaging to mitigate the adverse effects of PSF. Unfortunately, these two steps have only been considered separately in a sequential approach. Herein, a novel framework for unifying beamforming and deconvolution in ultrasound image reconstruction is introduced. More specifically, the proposed formulation is a regularized inverse problem including two linear models for beamforming and deconvolution plus additional sparsity constraint. We take advantage of the alternating direction method of multipliers algorithm to find the solution of the joint optimization problem. The performance evaluation is presented on a set of publicly available simulations, real phantoms, and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">in vivo</i> data. As compared to Delay-And-Sum (DAS) beamforming, simulation results indicate improvements of 45% and 44% in terms of axial and lateral resolution, respectively. Moreover, the proposed method improves the contrast of simulation data by 6.7% in comparison to DAS. The superiority of the proposed approach in comparison with the sequential approach as well as the state-of-the-art beamforming and deconvolution approaches is also shown.

Bandwidth Improvement in Ultrasound Image Reconstruction Using Deep Learning Techniques.

Ultrasound (US) imaging is a medical imaging modality that uses the reflection of sound in the range of 2-18 MHz to image internal body structures. In US, the frequency bandwidth (BW) is directly associated with image resolution. BW is a property of the transducer and more bandwidth comes at a higher cost. Thus, methods that can transform strongly bandlimited ultrasound data into broadband data are essential. In this work, we propose a deep learning (DL) technique to improve the image quality for a given bandwidth by learning features provided by broadband data of the same field of view. Therefore, the performance of several DL architectures and conventional state-of-the-art techniques for image quality improvement and artifact removal have been compared on in vitro US datasets. Two training losses have been utilized on three different architectures: a super resolution convolutional neural network (SRCNN), U-Net, and a residual encoder decoder network (REDNet) architecture. The models have been trained to transform low-bandwidth image reconstructions to high-bandwidth image reconstructions, to reduce the artifacts, and make the reconstructions visually more attractive. Experiments were performed for 20%, 40%, and 60% fractional bandwidth on the original images and showed that the improvements obtained are as high as 45.5% in RMSE, and 3.85 dB in PSNR, in datasets with a 20% bandwidth limitation.

Applications of ultrasound image reconstruction using deep learning

Deep learning has gained tremendous popularity as a tool for ultrasound beamforming and image reconstruction. In previous work, we trained deep neural networks (DNNs) to estimate the echogenicity of a medium, to improve acoustical and electronic signal-to-noise ratio (SNR) in channel data, and to detect targeted microbubbles nondestructively for real-time ultrasound molecular imaging. Here, we present several advancements to each application. First, we compare the speckle- and noise-reducing performance of DNNs trained with simple linear Field II simulations of photographic images versus that of DNNs trained with full wave finite-difference time domain numerical simulations containing realistic abdominal walls and the resulting image degradation artifacts. We further extend our nondestructive molecular imaging DNN to incorporate spatiotemporal information using an extended simulation study to increase specificity for stationary bound microbubbles and further improve nondestructive molecular imaging.