No sonographer, no radiologist: Assessing accuracy of artificial intelligence on breast ultrasound volume sweep imaging scans.

Breast ultrasound provides a first-line evaluation for breast masses, but the majority of the world lacks access to any form of diagnostic imaging. In this pilot study, we assessed the combination of artificial intelligence (Samsung S-Detect for Breast) with volume sweep imaging (VSI) ultrasound scans to evaluate the possibility of inexpensive, fully automated breast ultrasound acquisition and preliminary interpretation without an experienced sonographer or radiologist. This study was conducted using examinations from a curated data set from a previously published clinical study of breast VSI. Examinations in this data set were obtained by medical students without prior ultrasound experience who performed VSI using a portable Butterfly iQ ultrasound probe. Standard of care ultrasound exams were performed concurrently by an experienced sonographer using a high-end ultrasound machine. Expert-selected VSI images and standard of care images were input into S-Detect which output mass features and classification as "possibly benign" and "possibly malignant." Subsequent comparison of the S-Detect VSI report was made between 1) the standard of care ultrasound report by an expert radiologist, 2) the standard of care ultrasound S-Detect report, 3) the VSI report by an expert radiologist, and 4) the pathological diagnosis. There were 115 masses analyzed by S-Detect from the curated data set. There was substantial agreement of the S-Detect interpretation of VSI among cancers, cysts, fibroadenomas, and lipomas to the expert standard of care ultrasound report (Cohen's κ = 0.73 (0.57-0.9 95% CI), p<0.0001), the standard of care ultrasound S-Detect interpretation (Cohen's κ = 0.79 (0.65-0.94 95% CI), p<0.0001), the expert VSI ultrasound report (Cohen's κ = 0.73 (0.57-0.9 95% CI), p<0.0001), and the pathological diagnosis (Cohen's κ = 0.80 (0.64-0.95 95% CI), p<0.0001). All pathologically proven cancers (n = 20) were designated as "possibly malignant" by S-Detect with a sensitivity of 100% and specificity of 86%. Integration of artificial intelligence and VSI could allow both acquisition and interpretation of ultrasound images without a sonographer and radiologist. This approach holds potential for increasing access to ultrasound imaging and therefore improving outcomes related to breast cancer in low- and middle- income countries.

Radiological Society of North America/Quantitative Imaging Biomarker Alliance Shear Wave Speed Bias Quantification in Elastic and Viscoelastic Phantoms.

OBJECTIVES: To quantify the bias of shear wave speed (SWS) measurements between different commercial ultrasonic shear elasticity systems and a magnetic resonance elastography (MRE) system in elastic and viscoelastic phantoms. METHODS: Two elastic phantoms, representing healthy through fibrotic liver, were measured with 5 different ultrasound platforms, and 3 viscoelastic phantoms, representing healthy through fibrotic liver tissue, were measured with 12 different ultrasound platforms. Measurements were performed with different systems at different sites, at 3 focal depths, and with different appraisers. The SWS bias across the systems was quantified as a function of the system, site, focal depth, and appraiser. A single MRE research system was also used to characterize these phantoms using discrete frequencies from 60 to 500 Hz. RESULTS: The SWS from different systems had mean difference 95% confidence intervals of ±0.145 m/s (±9.6%) across both elastic phantoms and ± 0.340 m/s (±15.3%) across the viscoelastic phantoms. The focal depth and appraiser were less significant sources of SWS variability than the system and site. Magnetic resonance elastography best matched the ultrasonic SWS in the viscoelastic phantoms using a 140 Hz source but had a - 0.27 ± 0.027-m/s (-12.2% ± 1.2%) bias when using the clinically implemented 60-Hz vibration source. CONCLUSIONS: Shear wave speed reconstruction across different manufacturer systems is more consistent in elastic than viscoelastic phantoms, with a mean difference bias of < ±10% in all cases. Magnetic resonance elastographic measurements in the elastic and viscoelastic phantoms best match the ultrasound systems with a 140-Hz excitation but have a significant negative bias operating at 60 Hz. This study establishes a foundation for meaningful comparison of SWS measurements made with different platforms.

Technology trends and applications of deep learning in ultrasonography: image quality enhancement, diagnostic support, and improving workflow efficiency.

In this review of the most recent applications of deep learning to ultrasound imaging, the architectures of deep learning networks are briefly explained for the medical imaging applications of classification, detection, segmentation, and generation. Ultrasonography applications for image processing and diagnosis are then reviewed and summarized, along with some representative imaging studies of the breast, thyroid, heart, kidney, liver, and fetal head. Efforts towards workflow enhancement are also reviewed, with an emphasis on view recognition, scanning guide, image quality assessment, and quantification and measurement. Finally some future prospects are presented regarding image quality enhancement, diagnostic support, and improvements in workflow efficiency, along with remarks on hurdles, benefits, and necessary collaborations.

Attenuation of Shear Waves in Normal and Steatotic Livers.

Shear wave propagation in the liver has been a robust subject of research, with shear wave speed receiving the most attention. The correlation between increased shear wave speed and increased fibrosis in the liver has been established as a useful diagnostic tool. In comparison, the precise mechanisms of shear wave attenuation, and its relation to diseased states of the liver, are less well-established. This study focused on the hypothesis that steatosis adds a viscous (lossy) component to the liver, which increases shear wave attenuation. Twenty patients' livers were scanned with ultrasound and with induced shear wave propagation, and the resulting displacement profiles were analyzed using recently developed estimators to derive both the speed and attenuation of the shear waves within 6-cm2 regions of interest. The results were compared with pathology scores obtained from liver biopsies taken under ultrasound guidance. Across these cases, increases in shear wave attenuation were linked to increased steatosis score. This preliminary study supports the hypothesis and indicates the possible utility of the measurements for non-invasive and quantitative assessment of steatosis.

Group versus Phase Velocity of Shear Waves in Soft Tissues.

Across the varieties of waves that have been studied in physics, it is well established that group velocities can be significantly greater than or less than phase velocities measured within comparable frequency bands, depending on the particular mechanisms involved. The distinction between group and phase velocities is important in elastography, because diagnoses are made based on shear wave speed estimations from a variety of techniques. We review the general definitions of group and phase velocity and examine their specific relations within an important general class of rheological models. For the class of tissues and materials exhibiting power law dispersion, group velocity is significantly greater than phase velocity, and simple expressions are shown to interrelate the commonly measured parameters. Examples are given from phantoms and tissues.

Analysis of Transient Shear Wave in Lossy Media.

The propagation of shear waves from impulsive forces is an important topic in elastography. Observations of shear wave propagation can be obtained with numerous clinical imaging systems. Parameter estimations of the shear wave speed in tissues, and more generally the viscoelastic parameters of tissues, are based on some underlying models of shear wave propagation. The models typically include specific choices of the spatial and temporal shape of the impulsive force and the elastic or viscoelastic properties of the medium. In this work, we extend the analytical treatment of 2-D shear wave propagation in a biomaterial. The approach applies integral theorems relevant to the solution of the generalized Helmholtz equation, and does not depend on a specific rheological model of the tissue's viscoelastic properties. Estimators of attenuation and shear wave speed are derived from the analytical solutions, and these are applied to an elastic phantom, a viscoelastic phantom and in vivo liver using a clinical ultrasound scanner. In these samples, estimated shear wave group velocities ranged from 1.7 m/s in the liver to 2.5 m/s in the viscoelastic phantom, and these are lower-bounded by independent measurements of phase velocity.

Shear wave dispersion in lean versus steatotic rat livers.

OBJECTIVES: The precise measurement of fat accumulation in the liver, or steatosis, is an important clinical goal. Our previous studies in phantoms and mouse livers support the hypothesis that, starting with a normal liver, increasing accumulations of microsteatosis and macrosteatosis will increase the lossy viscoelastic properties of shear waves in a medium. This increase results in an increased dispersion (or slope) of the shear wave speed in the steatotic livers. METHODS: In this study, we moved to a larger animal model, lean versus obese rat livers ex vivo, and a higher-frequency imaging system to estimate the shear wave speed from crawling waves. RESULTS: The results showed elevated dispersion in the obese rats and a separation of the lean versus obese liver parameters in a 2-dimensional parameter space of the dispersion (slope) and shear wave speed at a reference frequency of 150 Hz. CONCLUSIONS: We have confirmed in 3 separate studies the validity of our dispersion hypothesis in animal models.

Shear wave speed and dispersion measurements using crawling wave chirps.

This article demonstrates the measurement of shear wave speed and shear speed dispersion of biomaterials using a chirp signal that launches waves over a range of frequencies. A biomaterial is vibrated by two vibration sources that generate shear waves inside the medium, which is scanned by an ultrasound imaging system. Doppler processing of the acquired signal produces an image of the square of vibration amplitude that shows repetitive constructive and destructive interference patterns called "crawling waves." With a chirp vibration signal, successive Doppler frames are generated from different source frequencies. Collected frames generate a distinctive pattern which is used to calculate the shear speed and shear speed dispersion. A special reciprocal chirp is designed such that the equi-phase lines of a motion slice image are straight lines. Detailed analysis is provided to generate a closed-form solution for calculating the shear wave speed and the dispersion. Also several phantoms and an ex vivo human liver sample are scanned and the estimation results are presented.

Mouse liver dispersion for the diagnosis of early-stage Fatty liver disease: a 70-sample study.

The accumulation of fat droplets within the liver is an important marker of liver disease. This study assesses gradations of steatosis in mouse livers using crawling waves, which are interfering patterns of shear waves introduced into the liver by external sources. The crawling waves are detected by Doppler ultrasound imaging techniques, and these are analyzed to estimate the shear wave speed as a function of frequency between 200 and 360 Hz. In a study of 70 mice with progressive increases in steatosis from 0% to >60%, increases in steatosis are found to increase the dispersion, or frequency dependence, of shear wave speed. This finding confirms an earlier, smaller study and points to the potential of a scoring system for steatosis based on shear wave dispersion.

Elasticity estimates from images of crawling waves generated by miniature surface sources.

We describe a surface-based approach to the generation of shear wave interference patterns, called crawling waves (CrW), within a medium and derive local estimates of biomechanical properties of tissue. In previous experiments, elongated bars operating as vibration sources were used to generate CrW propagation in samples. In the present study, however, a pair of miniature circular vibration sources was applied to the overlying skin to generate the CrW within the medium. The shape and position of the miniature sources make this configuration more applicable for in vivo implementation. A modified ultrasound imaging system is used to display the CrW propagation. A shear speed mapping algorithm is developed using a detailed analysis of the CrW. The proposed setup is applied to several biomaterials including a homogeneous phantom, an inhomogeneous phantom and an ex vivo human liver. The data are analyzed using the mapping algorithm to reveal the biomechanical properties of the biomaterials.

Quantitative crawling wave sonoelastography of benign and malignant thyroid nodules.

OBJECTIVE: The purpose of this study is to determine if crawling wave elastography, a novel sonoelastography technique, can be used to provide quantitative measurements of thyroid tissue shear velocity (a measure of tissue stiffness) and distinguish between benign and malignant thyroid nodules. STUDY DESIGN: Diagnostic test assessment. SETTING: Academic university. SUBJECTS AND METHODS: Fresh thyroid specimens (n = 20) with 44 regions of interest were imaged ex vivo with crawling wave sonoelastography over a 9-month period in 2010 at a single institution. Using the sonoelastography technique, shear velocity estimations and contrast-to-noise ratios were calculated. The higher the shear velocity (SV) and contrast-to-noise ratio (CNR), the greater the tissue stiffness. Histological diagnosis was correlated with shear velocity and contrast-to-noise ratio values. RESULTS: Both the shear velocity and contrast-to-noise values of papillary thyroid carcinoma (n = 10, CNR = 5.29, SV = 2.45 m/s) were significantly higher than benign nodules (n = 22, CNR = -0.41, SV = 1.90 m/s). There is a maximum sensitivity and specificity of 100% and 90.9%, respectively, for differentiating papillary thyroid carcinoma from benign nodules using contrast-to-noise ratio values. There is a maximum sensitivity and specificity of 83.3% and 72.7%, respectively, for differentiating papillary thyroid carcinoma from benign nodules using shear velocity values. Insufficient samples were obtained for comparison with other histological types. CONCLUSION: Crawling wave sonoelastography can provide quantitative estimations of shear velocity, thereby depicting the elastic properties of thyroid nodules. The shear velocity and contrast-to-noise ratio can differentiate between benign thyroid nodules and papillary thyroid carcinoma with high specificity and sensitivity.

Experimental validation of acoustic radiation force induced shear wave interference patterns.

A novel elasticity imaging system founded on the use of acoustic radiation forces from a dual beam arrangement to generate shear wave interference patterns is described. Acquired pulse-echo data and correlation-based techniques were used to estimate the resultant deformation and to visualize tissue viscoelastic response. The use of normal versus axicon focal configurations was investigated for effects on shear wave generation. Theoretical models were introduced and shown in simulation to accurately predict shear wave propagation and interference pattern properties. In a tissue-mimicking phantom, experimental results are in congruence with theoretical predictions. Using dynamic acoustic radiation force excitation, results confirm that shear wave interference patterns can be produced remotely in a particular tissue region of interest (ROI). Overall, preliminary results are encouraging and the system described may prove feasible for interrogating the viscoelastic properties of normal and diseased tissue types.

Modeling shear waves through a viscoelastic medium induced by acoustic radiation force.

In this study, a finite element model of a tissue-mimicking, viscoelastic phantom with a stiffer cylindrical inclusion subjected to an acoustic radiation force (ARF) is presented, and the resulting shear waves through the heterogeneous media are simulated, analyzed, and compared with experimental data. Six different models for the ARF were considered and compared. Each study used the same finite element model, but applied the following: (1) full radiation push; (2) focal region push; (3) single element focal point source; or (4) various thresholds of the full radiation push. For each case, displacements at discrete locations were determined and compared. The finite element simulation results for the full radiation push matched well with the experimental data with respect to replicating the shear wave speed and attenuation in the peak displacements through the background medium and inclusion, but did not illustrate comparable recovery after the peak displacements. As a result of this study, it has been shown that a focal region or point source push is not adequate to accurately model the effects of the full radiation push, but thresholding the full push can produce comparable results and reduce computation time.

Shear wave dispersion measures liver steatosis.

Crawling waves, which are interfering shear wave patterns, can be generated in liver tissue over a range of frequencies. Some important biomechanical properties of the liver can be determined by imaging the crawling waves using Doppler techniques and analyzing the patterns. We report that the dispersion of shear wave velocity and attenuation, that is, the frequency dependence of these parameters, are strongly correlated with the degree of steatosis in a mouse liver model, ex vivo. The results demonstrate the possibility of assessing liver steatosis using noninvasive imaging methods that are compatible with color Doppler scanners and, furthermore, suggest that liver steatosis can be separated from fibrosis by assessing the dispersion or frequency dependence of shear wave propagations.

Integration of crawling waves in an ultrasound imaging system. Part 1: system and design considerations.

An ultrasound system (GE Logiq 9) was modified to produce a synthetic crawling wave using shear wave displacements generated by the radiation force of focused beams formed at the left and the right edge of the region of interest (ROI). Two types of focusing, normal and axicon, were implemented. Baseband (IQ) data was collected to determine the left and right displacements, which were then used to calculate an interference pattern. By imposing a variable delay between the two pushes, the interference pattern moves across the ROI to produce crawling waves. Also temperature and pressure measurements were made to assess the safety issues. The temperature profiles measured in a veal liver along the focal line showed the maximum temperature rise less than 0.8°C, and the pressure measurements obtained in degassed water and derated by 0.3 dB/cm/MHz demonstrate that the system can operate within FDA safety guidelines.

Integration of crawling waves in an ultrasound imaging system. Part 2: signal processing and applications.

This paper introduces methods to generate crawling wave interference patterns from the displacement fields generated from radiation force pushes on a GE Logiq 9 scanner. The same transducer and system provides both the pushing pulses to generate the shear waves and the tracking pulses to measure the displacements. Acoustic power and system limitations result in largely impulsive displacement fields. Measured displacements from pushes on either side of a region-of-interest (ROI) are used to calculate continuously varying interference patterns. This technique is explained along with a brief discussion of the conventional mechanical source-driven crawling waves for comparison. We demonstrate the method on three example cases: a gelatin-based phantom with a cylindrical inclusion, an oil-gelatin phantom and mouse livers. The oil-gelatin phantom and the mouse livers demonstrate not only shear speed estimation, but the frequency dependence of the shear wave speeds.

Crawling wave detection of prostate cancer: preliminary in vitro results.

PURPOSE: The focus of this article is to develop signal and imaging processing methods to derive an accurate estimation of local tissue elasticity using the crawling wave (CrW) sonoelastography method. The task is to reduce noise and to improve the contrast of the elasticity map. METHODS: The protocol of the CrW approach was first tested on heterogeneous elastic phantoms as a model of prostate cancers. Then, the contrast-to-noise ratio of the estimation was calculated iteratively with various sequences of algorithms to determine the optimal signal processing settings. Finally, the optimized signal processing was applied to ex vivo prostate cancer detection. The comparison of the segmented elasticity map and the histology tumor outline was made by quadrants to evaluate the diagnostic performance of the protocol. Furthermore, the CrW approach was combined with amplitude-sonoelastography to achieve a higher specificity. RESULTS: This study demonstrated the feasibility of the proposed approach for clinical applications. In the application to ex vivo prostate cancer detection, the established approach was tested on 43 excised prostate glands. The combination of the CrW approach and amplitude-sonoelastography achieved an accuracy of over 80% for finding tumors larger than 4 mm in diameter. The elasticity values and contrast found by the CrW approach were in agreement with the previous results derived from mechanical testing. CONCLUSIONS: Crawling waves can be applied to detect prostate cancer with accuracy approaching 80% and can quantify the stiffness or shear modulus of both cancerous and noncancerous tissues. The technique therefore shows promise for guiding biopsies to suspect regions that are otherwise difficult to identify.

Crawling waves from radiation force excitation.

Crawling waves are generated by an interference of two oscillating waves traveling in opposite directions, with a progressive movement resulting from a frequency difference or a phase difference between the sources. While the idea has been applied to numerous applications, all the previous reports used mechanical sources to vibrate the medium. It is shown, through experiments and simulation, that crawling waves can be generated from focused beams that produce radiation force excitation within the tissue. Some examples are also shown.

Analysis and evaluation of a novel quality assurance device for ultrasonic medical imaging systems.

A novel test device that has the potential to facilitate day-to-day maintenance and quality assurance of ultrasonic medical imaging systems was developed and tested. The device consists of small echogenic particles deposited on a thin film substrate in patterns similar to those found on conventional eye charts, which can be scanned to produce recognizable images that can be quantitatively evaluated in a quick and repeatable fashion. An evaluation system consisting of this device and a related assessment methodology was tested at several locations by both physicians and sonographers. The results of the testing and responses of the users were both encouraging and insightful.