Shear Induced Non-Linear Elasticity Imaging: Elastography for Compound Deformations

Mechanics of Small-Scale Spherical Inclusions Using Nonlocal Poroelasticity Integrated with Light Gradient Boosting Machine

Detecting inclusions in materials at small scales is of high importance to ensure the quality, structural integrity and performance efficiency of microelectromechanical machines and products. Ultrasound waves are commonly used as a non-destructive method to find inclusions or structural flaws in a material. Mathematical continuum models can be used to enable ultrasound techniques to provide quantitative information about the change in the mechanical properties due to the presence of inclusions. In this paper, a nonlocal size-dependent poroelasticity model integrated with machine learning is developed for the description of the mechanical behaviour of spherical inclusions under uniform radial compression. The scale effects on fluid pressure and radial displacement are captured using Eringen's theory of nonlocality. The conservation of mass law is utilised for both the solid matrix and fluid content of the poroelastic material to derive the storage equation. The governing differential equations are derived by decoupling the equilibrium equation and effective stress-strain relations in the spherical coordinate system. An accurate numerical solution is obtained using the Galerkin discretisation technique and a precise integration method. A Dormand-Prince solution is also developed for comparison purposes. A light gradient boosting machine learning model in conjunction with the nonlocal model is used to extract the pattern of changes in the mechanical response of the poroelastic inclusion. The optimised hyperparameters are calculated by a grid search cross validation. The modelling estimation power is enhanced by considering nonlocal effects and applying machine learning processes, facilitating the detection of ultrasmall inclusions within a poroelastic medium at micro/nanoscales.

Possible depth-resolved reconstruction of shear moduli in the cornea following collagen crosslinking (CXL) with optical coherence tomography and elastography

Corneal collagen crosslinking (CXL) is commonly used to prevent or treat keratoconus. Although changes in corneal stiffness induced by CXL surgery can be monitored with non-contact dynamic optical coherence elastography (OCE) by tracking mechanical wave propagation, depth dependent changes are still unclear if the cornea is not crosslinked through the whole depth. Here, phase-decorrelation measurements on optical coherence tomography (OCT) structural images are combined with acoustic micro-tapping (AµT) OCE to explore possible reconstruction of depth-dependent stiffness within crosslinked corneas in an ex vivo human cornea sample. Experimental OCT images are analyzed to define the penetration depth of CXL into the cornea. In a representative ex vivo human cornea sample, crosslinking depth varied from ∼100 µm in the periphery to ∼150 µm in the cornea center and exhibited a sharp in-depth transition between crosslinked and untreated areas. This information was used in an analytical two-layer guided wave propagation model to quantify the stiffness of the treated layer. We also discuss how the elastic moduli of partially CXL-treated cornea layers reflect the effective engineering stiffness of the entire cornea to properly quantify corneal deformation.

Possible depth-resolved reconstruction of shear moduli in the cornea following collagen crosslinking (CXL) with optical coherence tomography and elastography

Corneal collagen crosslinking (CXL) is commonly used to prevent or treat keratoconus. Although changes in corneal stiffness induced by CXL surgery can be monitored with non-contact dynamic optical coherence elastography (OCE) by tracking mechanical wave propagation, depth dependent changes are still unclear if the cornea is not crosslinked through the whole depth. Here, phase-decorrelation measurements on optical coherence tomography (OCT) structural images are combined with acoustic micro-tapping (A$\mu$T) OCE to explore possible reconstruction of depth-dependent stiffness within crosslinked corneas in an ex vivo human cornea sample. Experimental OCT images are analyzed to define the penetration depth of CXL into the cornea. In a representative ex vivo human cornea sample, crosslinking depth varied from $\sim 100\mu m$ in the periphery to $\sim 150\mu m$ in the cornea center and exhibited a sharp in-depth transition between crosslinked and untreated areas. This information was used in an analytical two-layer guided wave propagation model to quantify the stiffness of the treated layer. We also discuss how the elastic moduli of partially CXL-treated cornea layers reflect the effective engineering stiffness of the entire cornea to properly quantify corneal deformation.

Cross-Image Dependency Modeling for Breast Ultrasound Segmentation

We present a novel deep network (namely BUSSeg) equipped with both within- and cross-image long-range dependency modeling for automated lesions segmentation from breast ultrasound images, which is a quite daunting task due to (1) the large variation of breast lesions, (2) the ambiguous lesion boundaries, and (3) the existence of speckle noise and artifacts in ultrasound images. Our work is motivated by the fact that most existing methods only focus on modeling the within-image dependencies while neglecting the cross-image dependencies, which are essential for this task under limited training data and noise. We first propose a novel cross-image dependency module (CDM) with a cross-image contextual modeling scheme and a cross-image dependency loss (CDL) to capture more consistent feature expression and alleviate noise interference. Compared with existing cross-image methods, the proposed CDM has two merits. First, we utilize more complete spatial features instead of commonly used discrete pixel vectors to capture the semantic dependencies between images, mitigating the negative effects of speckle noise and making the acquired features more representative. Second, the proposed CDM includes both intra- and inter-class contextual modeling rather than just extracting homogeneous contextual dependencies. Furthermore, we develop a parallel bi-encoder architecture (PBA) to tame a Transformer and a convolutional neural network to enhance BUSSeg's capability in capturing within-image long-range dependencies and hence offer richer features for CDM. We conducted extensive experiments on two representative public breast ultrasound datasets, and the results demonstrate that the proposed BUSSeg consistently outperforms state-of-the-art approaches in most metrics.

A Non-invasive Method to Estimate the Stress-Strain Curve of Soft Tissue Using Ultrasound Elastography

Ultrasound elastography performed under small strain conditions has been intensively studied. However, small deformations may be not sufficiently large to differentiate some abnormal tissues. By combining quasi-static and shear wave elastography, we developed a non-invasive method to estimate the localized stress- strain curve of materials. This method exerts progressive multistep uniaxial compression on the materials, and shear wave measurements were performed at every compression step. This method estimates the 2-D displacements between steps via a 2-D region growing motion tracking method and accumulates these displacements to obtain the large material displacements with respect to the initial configuration. At each step, the shear modulus and stress were calculated according to linear elastic theory. The proposed method was tested on custom-made tissue-mimicking phantoms. Mechanical compression testing was conducted on the samples made of the same material as the phantoms and taken as the reference. The stress-strain curves for the same material from the proposed method and from mechanical testing are in good agreement. The root mean square error (RMSE) and area percentage error (APE) of the stress-strain curve between ultrasound measurement and mechanical testing for soft materials ranged from 0.18 to 0.26 kPa and from 5.6% to 7.8%, respectively. The RMSE and APE for stiff materials ranged from 0.56 to 1.17 kPa and 8.0% to 17.9%. Therefore, our method was able to provide good estimates of the stress-strain curve for tissue-mimicking materials.

Imaging the Local Nonlinear Viscoelastic Properties of Soft Tissues: Initial Validation and Expected Benefits

Imaging tissue mechanical properties has shown promise in noninvasive assessment of numerous pathologies. Researchers have successfully measured many linear tissue mechanical properties in laboratory and clinical settings. Currently, multiple complex mechanical effects such as frequency-dependence, anisotropy, and nonlinearity are being investigated separately. However, a concurrent assessment of these complex effects may enable more complete characterization of tissue biomechanics and offer improved diagnostic sensitivity. In this work, we report for the first time a method to map the frequency-dependent nonlinear parameters of soft tissues on a local scale. We recently developed a nonlinear elastography model that combines strain measurements from arbitrary tissue compression with radiation-force-based broadband shear wave speed (WS) measurements. Here, we extended this model to incorporate local measurements of frequency-dependent shear modulus. This combined approach provides a local frequency-dependent nonlinear parameter that can be obtained with arbitrary, clinically realizable tissue compression. Initial assessments using simulations and phantoms validate the accuracy of this approach. We also observed improved contrast in nonlinearity parameter at higher frequencies. Results from ex-vivo liver experiments show 32, 25, 34, and 38 dB higher contrast in elastograms than traditional linear elasticity, elastic nonlinearity, viscosity, and strain imaging methods, respectively. A lesion, artificially created by injection of glutaraldehyde into a liver specimen, showed a 59% increase in the frequency-dependent nonlinear parameter and a 17% increase in contrast ratio.

Shear Wave Elasticity Imaging Using Nondiffractive Bessel Apodized Acoustic Radiation Force

The acoustic radiation force impulse (ARFI) has been widely used in transient shear wave elasticity imaging (SWEI). For SWEI based on focused ARFI, the highest image quality exists inside the focal zone due to the limitation of the depth of focus and diffraction. Consequently, the areas outside the focal zone and in the near field present poor image quality. To address the limitations of the focused beam, we introduce Bessel apodized ARFI that enhances image quality and improves the depth of focus. The objective of this study is to evaluate the feasibility of SWEI based on Bessel ARF in simulation and experiment. We report measurements of elastogram image quality and depth of field in tissue-mimicking phantoms and ex vivo liver tissue. Our results demonstrate improved depth of field, image quality, and shear wave speed (SWS) estimation accuracy using Bessel push beams. As a result, Bessel ARF enlarges the field of view of elastograms. The signal-to-noise ratio (SNR) of Bessel SWEI is improved 26% compared with focused SWEI in homogeneous phantom. The estimated SWS by Bessel SWEI is closer to the measured SWS from a clinical scanner with an error of 0.3% compared to 2.4% with a focused beam. In heterogeneous phantoms, the contrast-to-noise ratios (CNRs) of shallow and deep inclusions are improved by 8.79 and 3.33 dB, respectively, under Bessel ARF. We also compare the results between Bessel SWEI and supersonic shear imaging (SSI), and the SNR of Bessel SWEI is improved by 8.1%. Compared with SSI, Bessel SWEI shows more accurate SWS estimates in high stiffness inclusions. Finally, Bessel SWEI can generate higher quality elastograms with less energy than conventional SSI.

Second-Generation Dual Scan Mammoscope With Photoacoustic, Ultrasound, and Elastographic Imaging Capabilities

We recently developed the photoacoustic dual-scan mammoscope (DSM), a system that images the patient in standing pose analog to X-ray mammography. The system simultaneously acquires three-dimensional photoacoustic and ultrasound (US) images of the mildly compressed breast. Here, we describe a second-generation DSM (DSM-2) system that offers a larger field of view, better system stability, higher ultrasound imaging quality, and the ability to quantify tissue mechanical properties. In the new system, we doubled the field of view through laterally shifted round-trip scanning. This new design allows coverage of the entire breast tissue. We also adapted precisely machined holders for the transducer-fiber bundle sets. The new holder increased the mechanical stability and facilitated image registration from the top and bottom scanners. The quality of the US image is improved by increasing the firing voltage and the number of firing angles. Finally, we incorporated quasi-static ultrasound elastography to allow comprehensive characterization of breast tissue. The performance of the new system was demonstrated through in vivo human imaging experiments. The experimental results confirmed the capability of the DSM-2 system as a powerful tool for breast imaging.

Domain Knowledge Powered Deep Learning for Breast Cancer Diagnosis Based on Contrast-Enhanced Ultrasound Videos

In recent years, deep learning has been widely used in breast cancer diagnosis, and many high-performance models have emerged. However, most of the existing deep learning models are mainly based on static breast ultrasound (US) images. In actual diagnostic process, contrast-enhanced ultrasound (CEUS) is a commonly used technique by radiologists. Compared with static breast US images, CEUS videos can provide more detailed blood supply information of tumors, and therefore can help radiologists make a more accurate diagnosis. In this paper, we propose a novel diagnosis model based on CEUS videos. The backbone of the model is a 3D convolutional neural network. More specifically, we notice that radiologists generally follow two specific patterns when browsing CEUS videos. One pattern is that they focus on specific time slots, and the other is that they pay attention to the differences between the CEUS frames and the corresponding US images. To incorporate these two patterns into our deep learning model, we design a domain-knowledge-guided temporal attention module and a channel attention module. We validate our model on our Breast-CEUS dataset composed of 221 cases. The result shows that our model can achieve a sensitivity of 97.2% and an accuracy of 86.3%. In particular, the incorporation of domain knowledge leads to a 3.5% improvement in sensitivity and a 6.0% improvement in specificity. Finally, we also prove the validity of two domain knowledge modules in the 3D convolutional neural network (C3D) and the 3D ResNet (R3D).