A GPU-Accelerated 3-D Coupled Subsample Estimation Algorithm for Volumetric Breast Strain Elastography

Convolutional Neural Network-Based Speckle Tracking for Ultrasound Strain Elastography: An Unsupervised Learning Approach

Accurate and computationally efficient motion estimation is a critical component of real-time ultrasound strain elastography (USE). With the advent of deep-learning neural network models, a growing body of work has explored supervised convolutional neural network (CNN)-based optical flow in the framework of USE. However, the above-said supervised learning was often done using simulated ultrasound data. The research community has questioned whether simulated ultrasound data containing simple motion can train deep-learning CNN models that can reliably track complex in vivo speckle motion. In parallel with other research groups' efforts, this study developed an unsupervised motion estimation neural network (UMEN-Net) for USE by adapting a well-established CNN model named PWC-Net. Our network's input is a pair of predeformation and postdeformation radio frequency (RF) echo signals. The proposed network outputs both axial and lateral displacement fields. The loss function consists of a correlation between the predeformation signal and the motion-compensated postcompression signal, smoothness of the displacement fields, and tissue incompressibility. Notably, an innovative correlation method known as the globally optimized correspondence (GOCor) volumes module developed by Truong et al. was used to replace the original Corr module to enhance our evaluation of signal correlation. The proposed CNN model was tested using simulated, phantom, and in vivo ultrasound data containing biologically confirmed breast lesions. Its performance was compared against other state-of-the-art methods, including two deep-learning-based tracking methods (MPWC-Net++ and ReUSENet) and two conventional tracking methods (GLUE and BRGMT-LPF). In summary, compared with the four known methods mentioned above, our unsupervised CNN model not only obtained higher signal-to-noise ratios (SNRs) and contrast-to-noise ratios (CNRs) for axial strain estimates but also improved the quality of the lateral strain estimates.

Hadamard-Encoded Synthetic Transmit Aperture Imaging for Improved Lateral Motion Estimation in Ultrasound Elastography

Lateral motion estimation has been a challenge in ultrasound elastography mainly due to the low resolution, low sampling frequency, and lack of phase information in the lateral direction. Synthetic transmit aperture (STA) can achieve high resolution due to two-way focusing and can beamform high-density image lines for improved lateral motion estimation with the disadvantages of low signal-to-noise ratio (SNR) and limited penetration depth. In this study, Hadamard-encoded STA (Hadamard-STA) is proposed for the improvement of lateral motion estimation in elastography, and it is compared with STA and conventional focused wave (CFW) imaging. Simulations, phantom, and in vivo experiments were conducted to make the comparison. The normalized root mean square error (NRMSE) and the contrast-to-noise ratio (CNR) were calculated as the evaluation criteria in the simulations. The results show that, at a noise level of -10 dB and an applied strain of -1% (compression), Hadamard-STA decreases the NRMSEs of lateral displacements by 46.92% and 35.35%, decreases the NRMSEs of lateral strains by 52.34% and 39.75%, and increases the CNRs by 9.70 and 9.75 dB compared with STA and CFW, respectively. In the phantom experiments performed on a heterogeneous tissue-mimicking phantom, the sum of squared differences (SSD) between the reference and the motion-compensated RF data, and the CNR were calculated as the evaluation criteria. At an applied strain of -1.80%, Hadamard-STA is found to decrease the SSDs by 20.91% and 30.99% and increase the CNRs by 14.15 and 24.66 dB compared with STA and CFW, respectively. In the experiments performed on a breast phantom, Hadamard-STA achieves better visualization of the breast inclusion with a clearer boundary between the inclusion and the background than STA and CFW. The in vivo experiments were performed on a patient with a breast tumor, and the tumor could also be better visualized with a more homogeneous background in the strain image obtained by Hadamard-STA than by STA and CFW. These results demonstrate that Hadamard-STA achieves a substantial improvement in lateral motion estimation and maybe a competitive method for quasi-static elastography.

Digital image technology based on PCA and SVM for detection and recognition of foreign bodies in lyophilized powder

BACKGROUND:

Digital image technology has made great progress in the field of foreign body detection and classification, which is of great help to drug purity extraction and impurity analysis and classification.

OBJECTIVE:

The detection and classification of foreign bodies in lyophilized powder are important. The method which can obtain a higher accuracy of recognition needs to be proposed.

METHODS:

We used digital image technology to detect and classify foreign bodies in lyophilized powder, and studied the process of image preprocessing, median filtering, Wiener filtering and average filtering balance to better detect and classify foreign bodies in lyophilized powder.

RESULTS:

Through industrial small sample data simulation, test results show that in the process of image preprocessing, 3 × 3 median filtering is best. In the aspect of foreign body recognition, the recognition based on principal component analysis (PCA) and support vector machine (SVM) algorithm and the recognition based on PCA and Third-Nearest Neighbor classification algorithm are compared and results show that the PCA+SVM algorithm is better.

CONCLUSION:

We demonstrated that integrating PCA and SVM to classify foreign bodies in lyophilized powder. Our proposed method is effective for the prediction of essential proteins.

Neural-network-based Motion Tracking for Breast Ultrasound Strain Elastography: An Initial Assessment of Performance and Feasibility

Accurate tracking of tissue motion is critically important for several ultrasound elastography methods. In this study, we investigate the feasibility of using three published convolution neural network (CNN) models built for optical flow (hereafter referred to as CNN-based tracking) by the computer vision community for breast ultrasound strain elastography. Elastographic datasets produced by finite element and ultrasound simulations were used to retrain three published CNN models: FlowNet-CSS, PWC-Net, and LiteFlowNet. After retraining, the three improved CNN models were evaluated using computer-simulated and tissue-mimicking phantoms, and in vivo breast ultrasound data. CNN-based tracking results were compared with two published two-dimensional (2D) speckle tracking methods: coupled tracking and GLobal Ultrasound Elastography (GLUE) methods. Our preliminary data showed that, based on the Wilcoxon rank-sum tests, the improvements due to retraining were statistically significant (p < 0.05) for all three CNN models. We also found that the PWC-Net model was the best neural network model for data investigated, and its overall performance was on par with the coupled tracking method. CNR values estimated from in vivo axial and lateral strain elastograms showed that the GLUE algorithm outperformed both the retrained PWC-Net model and the coupled tracking method, though the GLUE algorithm exhibited some biases. The PWC-Net model was also able to achieve approximately 45 frames/second for 2D speckle tracking data investigated.

Modeling Uncertainty of Strain Ratio Measurements in Ultrasound Breast Strain Elastography: A Factorial Experiment

Strain elastography (SE) is a technique in which images of localized tissue strains are used to detect the relative stiffness of tissues. The application of SE in differentiating malignant breast lesions from benign ones has been broadly investigated. The strain ratio (SR) between the background and the breast tumor has been used and its results have been mixed. Due to the complex nature of tissue elasticity and how it relates to the strain fields measured in SE, the exact reason is not known. In this study, we apply a novel design-of-experiments-based metamodeling approach to mechanical simulation of SE in the human breast. To our knowledge, such a study has not been reported in the ultrasound SE literature. More specifically, we first conduct a screening study to identify the biomechanical factors/simulation inputs that most strongly determine SR. We then apply a response surface experimental design to these factors to produce a metamodel of SR as a function of said factors. Results from the screening study suggest that the SR measurements are primarily influenced by three factors: the initial shear modulus of the lesion, the elastic nonlinearity of the lesion, and the precompression applied during acquisition. In order to investigate the implications of these results, stochastic inputs for these three factors associated with the malignant and benign cases were applied to the resulting response surface. The resulting optimal cutoffs, sensitivity, and specificity were generally in line with a majority (>60%) of 19 clinical trials in the literature.

Accelerating 3-D GPU-based Motion Tracking for Ultrasound Strain Elastography Using Sum-Tables: Analysis and Initial Results

Now, with the availability of 3-D ultrasound data, a lot of research efforts are being devoted to developing 3-D ultrasound strain elastography (USE) systems. Because 3-D motion tracking, a core component in any 3-D USE system, is computationally intensive, a lot of efforts are under way to accelerate 3-D motion tracking. In the literature, the concept of Sum-Table has been used in a serial computing environment to reduce the burden of computing signal correlation, which is the single most computationally intensive component in 3-D motion tracking. In this study, parallel programming using graphics processing units (GPU) is used in conjunction with the concept of Sum-Table to improve the computational efficiency of 3-D motion tracking. To our knowledge, sum-tables have not been used in a GPU environment for 3-D motion tracking. Our main objective here is to investigate the feasibility of using sum-table-based normalized correlation coefficient (ST-NCC) method for the above-mentioned GPU-accelerated 3-D USE. More specifically, two different implementations of ST-NCC methods proposed by Lewis et al. and Luo-Konofagou are compared against each other. During the performance comparison, the conventional method for calculating the normalized correlation coefficient (NCC) was used as the baseline. All three methods were implemented using compute unified device architecture (CUDA; Version 9.0, Nvidia Inc., CA, USA) and tested on a professional GeForce GTX TITAN X card (Nvidia Inc., CA, USA). Using 3-D ultrasound data acquired during a tissue-mimicking phantom experiment, both displacement tracking accuracy and computational efficiency were evaluated for the above-mentioned three different methods. Based on data investigated, we found that under the GPU platform, Lou-Konofaguo method can still improve the computational efficiency (17–46%), as compared to the classic NCC method implemented into the same GPU platform. However, the Lewis method does not improve the computational efficiency in some configuration or improves the computational efficiency at a lower rate (7–23%) under the GPU parallel computing environment. Comparable displacement tracking accuracy was obtained by both methods.

A comparison of hyperelastic constitutive models applicable to shear wave elastography (SWE) data in tissue-mimicking materials

Shear wave elastography (SWE) techniques have received substantial attention in recent years. Strong experimental data in SWE suggest that shear wave speed changes significantly due to the known acoustoelastic effect (AE). This presents both challenges and opportunities toward in vivo characterization of biological soft tissues. In this work, under the framework of continuum mechanics, we model a tissue-mimicking material as a homogeneous, isotropic, incompressible, hyperelastic material. Our primary objective is to quantitatively and qualitatively compare experimentally measured acoustoelastic data with model-predicted outcomes using multiple strain energy functions. Our analysis indicated that the classic Neo-Hookean and Mooney-Rivlin models are inadequate for modeling the AE in tissue-mimicking materials. However, a subclass of strain energy functions containing both high-order/exponential term(s) and second-order invariant dependence showed good agreement with experimental data. Based on data investigated, we also found that discrepancies may exist between parameters inversely estimated from uniaxial compression and SWE data. Overall, our findings may improve our understanding of clinical SWE results.

GPU Accelerated Multilevel Lagrangian Carotid Strain Imaging

A multilevel Lagrangian carotid strain imaging algorithm is analyzed to identify computational bottlenecks for implementation on a graphics processing unit (GPU). Displacement tracking including regularization was found to be the most computationally expensive aspect of this strain imaging algorithm taking about 2.2 h for an entire cardiac cycle. This intensive displacement tracking was essential to obtain Lagrangian strain tensors. However, most of the computational techniques used for displacement tracking are parallelizable, and hence GPU implementation is expected to be beneficial. A new scheme for subsample displacement estimation referred to as a multilevel global peak finder was also developed since the Nelder-Mead simplex optimization technique used in the CPU implementation was not suitable for GPU implementation. GPU optimizations to minimize thread divergence and utilization of shared and texture memories were also implemented. This enables efficient use of the GPU computational hardware and memory bandwidth. Overall, an application speedup of was obtained enabling the algorithm to finish in about 50 s for a cardiac cycle. Last, comparison of GPU and CPU implementations demonstrated no significant difference in the quality of displacement vector and strain tensor estimation with the two implementations up to a 5% interframe deformation. Hence, a GPU implementation is feasible for clinical adoption and opens opportunity for other computationally intensive techniques.

A 3-D Region-Growing Motion-Tracking Method for Ultrasound Elasticity Imaging

A 3-D region-growing motion-tracking (RGMT) method for ultrasound elasticity imaging is described. This 3-D RGMT method first estimates the displacements at a sparse subset of points, called seeds; uses an objective measure to determine, among those seeds, which displacement estimates to trust; and then performs RGMT in three dimensions to estimate displacements for the remaining points in the field. During the growing process in three dimensions, the displacement estimate at one grid point is employed to guide the displacement estimation of its neighboring points using a 3-D small search region. To test this algorithm, volumetric ultrasound radiofrequency echo data were acquired from one phantom and five in vivo human breasts. Displacement estimates obtained with the 3-D RGMT method were compared with a published 2-D RGMT method via motion-compensated cross-correlation (MCCC) of pre- and post-deformation radiofrequency echo signals. For data from experiments with the phantom, the MCCC values in the entire tracking region of interest averaged approximately 0.95, and the contrast-to-noise ratios averaged 4.6 for both tracking methods. For all five patients, the average MCCC values within the region of interest obtained with the 3-D RGMT were consistently higher than those obtained with the 2-D RGMT method. These results indicate that the 3-D RGMT algorithm is able to track displacements with increased accuracy and generate higher-quality 3-D elasticity images than the 2-D RGMT method.

A Normalized Shear Deformation Indicator for Ultrasound Strain Elastography in Breast Tissues: An In Vivo Feasibility Study

The shear deformation under loads contains useful information for distinguishing benign breast lesions from malignant ones. In this study, we proposed a normalized shear deformation indicator (NSDI) that was derived from the concept of principal strains. Since the NSDI requires both high-quality axial and lateral (parallel and perpendicular to the beam, resp.) displacement estimates, a strategy combining high-quality speckle tracking with signal “denoising” was employed. Both techniques were previously published by our group. Finite element (FE) models were used to identify possible causes for elevated NSDI values in and around breast lesions, followed by an analysis of ultrasound data acquired from 26 biopsy-confirmed in vivo breast lesions. We found that, theoretically, the elevated NSDI values could be attributed to two factors: significantly hardened tissue stiffness and increasing heterogeneity. The analysis of in vivo data showed that the proposed NSDI values were higher (p < 0.05) among malignant cancers as compared to those measured from benign ones. In conclusion, our preliminary results demonstrated that the calculation of NSDI value is feasible and NSDI could add value to breast lesion differentiation with current clinical equipment as a postprocessing tool.

Efficient Two-Pass 3-D Speckle Tracking for Ultrasound Imaging

Speckle tracking based on block matching is the most common method for multi-dimensional motion estimation in ultrasound elasticity imaging. Extension of two-dimensional (2-D) methods to three dimensions (3-D) has been problematic because of the large computational load of 3-D tracking, as well as performance issues related to the low frame (volume) rates of 3-D images. To address both of these problems, we have developed an efficient two-pass tracking method suited to cardiac elasticity imaging. PatchMatch, originally developed for image editing, has been adapted for ultrasound to provide first-pass displacement estimates. Second-pass estimation uses conventional block matching within a much smaller search region. 3-D displacements are then obtained using correlation filtering previously shown to be effective against speckle decorrelation. Both simulated and in vivo canine cardiac results demonstrate that the proposed two-pass method reduces computational cost compared to conventional 3-D exhaustive search by a factor of 10. Moreover, it outperforms one-pass tracking by a factor of about 3 in terms of root-mean-square error relative to available ground-truth displacements.