|Home | About | Journals | Submit | Contact Us | Français|
Elastography is a new imaging modality where elastic tissue parameters related to the structural organization of normal and pathological tissues are imaged. Basic principles underlying the quasi-static elastography concept and principles are addressed. The rationale for elastographic imaging is reinforced using data on elastic properties of normal and abnormal soft tissues. The several orders of magnitude difference between the elastic modulus of normal and abnormal tissues which is the primary contrast mechanism in elastographic imaging underlines the probability of success with this imaging modality. Recent advances enabling the clinical practice of elastographic imaging in real-time on clinical ultrasound systems is also discussed.
In quasi-static elastography, radiofrequency echo signals acquired before and after a small (about 1%) of applied deformation are correlated to estimate tissue displacements. Local tissue displacement vector estimates between small segments of the pre- and post-deformation signals are estimated and the corresponding strain distribution imaged. Elastographic imaging techniques are based on the hypothesis that soft tissues deform more than stiffer tissue, and these differences can be quantified in images of the tissue strain tensor or the Young’s modulus.
Clinical applications of quasi-static elastography have mushroomed over the last decade, with the most commonly imaged areas being the breast, prostate, thyroid, cardiac, treatment monitoring of ablation procedures and vascular imaging applications.
Imaging of the viscoelastic properties of tissue for diagnosis and treatment has gained popularity over the last decade because of the ability to provide noninvasive and new diagnostic information [1–29]. Elastography has been likened to manual palpation of tissue, utilized by clinicians for centuries to aid in clinical diagnosis. The clinical popularity of manual palpation is due to the fact that pathologic and stiffness changes in the body are generally well-correlated , and irregular and stiffer masses commonly present as warning signs of diseases in organs such as the breast, liver, and prostate. For example, breast scirrhous carcinomas on palpation are felt to be extremely hard nodules [31, 32], while liver tissue with cirrhosis is also known to be significantly stiffer than normal healthy liver tissue . However, manual palpation is generally limited to superficial structures and depends largely on the ability of the physician performing the examination. Stiffer masses located deep in the body almost certainly cannot be detected by surface manual palpation. The biggest advancement in the field of tissue elasticity imaging has come with the advent and advances in ‘elastography’ over the last two decades. The motivation for the use of elastographic techniques stems from the fact that large differences in stiffness or modulus contrast exist between surrounding normal and pathological tissues that may otherwise possess similar image contrasts with conventional clinical imaging modalities.
The term ‘Elastography’ was coined by Ophir et al. , to refer to an ultrasound based imaging technique where local axial strains were estimated by computing the gradient of axial shifts in echo arrival times along the ultrasound beam direction following a quasi-static tissue deformation. Elastography, however has now been used as a more general term to identify methods that image tissue stiffness, using different imaging modalities for example ultrasound, magnetic resonance imaging, optical coherence tomography, X-ray computed tomography [6, 26–29], different perturbation techniques to deform tissue [5, 6, 15, 24, 34–37] and based on the elasticity parameter being measured or imaged [6, 38–40]. In this review, we will focus on quasi-static ultrasound based elastography, since this group of methods form the more commonly used approaches for clinical elasticity imaging.
The practice of quasi-static ultrasound elastography was initially based on the estimation of the axial tissue displacement and strain (corresponding to displacements estimated along both the direction of insonification and tissue deformation) by analyzing ultrasonic radiofrequency echo signals obtained from standard clinical ultrasound diagnostic equipment. Frames of radiofrequency echo-signals acquired before and after a small amount (about 1%) of quasi-static deformation were correlated to estimate differential displacements along small data segments or regions of interest using classical time-delay estimation techniques [41, 42]. Finally, the axial strain distribution was computed from the gradient of the time-delays or tissue displacements .
Algorithms for displacement and strain estimation have progressed from one-dimensional (1D) tracking of the displacement along the insonification and deformation direction [6, 9, 43–47] to two-dimensional (2D) tracking of the displacement vector within the scan plane [48–52] to three-dimensional (3D) [53–55] methods for tracking the complete displacement vector. These methods can also be classified based on whether they utilize phase information present in the radiofrequency data, namely time-domain cross-correlation [6, 56], phase-tracking techniques , and phase root tracking methods . Estimators that do not utilize phase information that include optical flow-based speckle tracking , analysis of envelope or B-mode signals , power spectral methods[45, 58, 59], and methods like sum amplitude or sum-squared difference that mimic cross-correlation [60, 61]. Algorithms to reduce signal decorrelation such as temporal stretching [4–7] of the post-deformation signal to align the radiofrequency peaks with the pre- deformation signal, and multi-compression averaging [6–9], which reduces signal decorrelation by using small deformations, have been utilized. Displacement estimates from the multiple small deformations have been accumulated [8, 9], averaged or compounded [6–9] to improve strain and modulus images. Displacement estimates obtained from multiple angular insonifications have also been averaged to improve the noise properties of the strain images [62, 63]. 2D signal processing methods to estimate the displacement vector for data acquired using curvilinear arrays have also been developed , since most of the previous algorithms catered to data acquired using linear arrays [48–52]. Displacement estimation in areas where the tissue continuity assumption does not hold namely for vascular tissue have also been developed [65, 66], by processing data using a coarse to a fine approach in multiple estimation stages [48, 50, 51]. Real-time implementations of strain imaging on both custom  and clinical ultrasound scanners , for direct visualization of the strain distribution while scanning the patient has been reported.
Only the axial-strain distribution i.e. strains along the insonification and deformation were imaged until recently , while lateral (perpendicular to the beam propagation direction and within the same scan plane) and elevational (perpendicular to the beam propagation direction and scan plane) displacements were usually not estimated. Algorithms have been developed for the estimation of the complete displacement vector and consequently the strain tensor components [67–69]. Since the components of the strain tensor are coupled, accurate estimations of all components are necessary for a complete visualization of the 3D strain distribution incurred in tissue. Lubinski et al.  computed local lateral displacements from axial displacements assuming tissue incompressibility (Poisson’s ratio of 0.495). Konofagou and Ophir  describe a method that utilized weighted interpolation between neighboring radiofrequency A-lines in the lateral direction, along with iterative corrections of lateral and axial displacements to estimate displacement vectors. Another method using radiofrequency data acquired along multiple angular insonification directions (using data from a phased array transducer) was described by Techavipoo et al.  to estimate components of the displacement vector, which was later adapted to linear array transducers with electronic beam steering .
The normal strain tensor components are also essential in the estimation of other important parameters such as the shear strains [68, 73, 74] and lateral to axial strain ratios (equivalent to the Poisson’s ratio under specific conditions) . Imaging of the shear strain distribution has been utilized to evaluate lesion mobility, in order to differentiate between benign masses such as fibroadenomas that are surrounded by a capsule and loosely attached to background normal tissue and malignant masses that are firmly attached to the normal background tissue through infiltration and the accompanying desmoplastic reaction [75, 76]. The lateral to axial strain ratio has been utilized as a marker to characterize poroelastic tissue enabling differentiation, for example between normal and edematous tissues .
However, the strain tensor distribution is not indicative of the absolute elastic properties of tissue, since it is significantly dependent on the applied deformation or stress distribution [78, 79]. The additional information with the complete strain tensor distribution can be utilized to obtain unique solutions for the inverse problem [38, 80]. Many investigators have focused on the estimation of the Young’s Modulus in tissue, which if estimated accurately would provides an absolute or quantitative distribution of the underlying tissue elastic properties. However, there are many challenges that have to be addressed for accurate estimation of the Young’s modulus, including accurate estimation of the stress distribution and the associated boundary conditions.
Techniques to obtain the local stress distribution patterns using force sensor arrays during tissue deformation have been reported . Temporal and spatial maps of the stress distribution are obtained to evaluate breast masses. The major limitation is the absence of depth dependent information which is being addressed using analytic and finite element methods to model the stress distribution.
The improvements in the quality of the strain and modulus images have dramatically improved over the last two decades. This has led to the commercialization of quasi-static elastography, with four manufactures currently offering strain imaging modes on commercial clinical systems. The following sections lead the reader through a classification of quasi-static elastographic techniques; provide a brief description of the fundamental principles underpinning quasi-static elastography. The review then concludes with a description of the current clinical applications of quasi-static strain and modulus imaging.
Elastographic techniques can be broadly classified based on the mechanical stimuli applied into quasi-static and dynamic techniques. Under quasi-static techniques, which are the focus of this review paper, we can further classify the approaches into three categories, namely: 1) Steady-state quasi-static excitation [6–14], 2) Steady-state quasi-static low frequency excitation on the order of 5–10 Hz, and 3) Imaging of steady-state quasi-static deformations due to physiological excitation [35–37], as illustrated in Fig. 1. Dynamic methods can also be further classified into approaches that utilize 1) Harmonic excitations on the order of 10–1000 Hz [3–5, 23, 82], and 2) Transient dynamic excitation and subsequent estimation of the transient response [20–22, 83].
Techniques that utilize a steady state quasi-static excitation of tissue using the ultrasound transducer represent some of the most commonly utilized perturbation approaches to elastographic imaging. Radiofrequency data are collected before and after a known applied deformation, applied as a single deformation step or as multiple deformation steps [6, 7, 9, 84]. The pre- and post-deformation data are then compared utilizing different algorithms to estimate differential local displacements between the two tissue states [41, 85–87]. Local strains are then estimated from the gradient of the displacement using forward differences or least-square based methods.
Steady-state quasi-static deformations have been applied using either under stepper motor control (with the transducer held in a fixture) or freehand compression using the transducer. Approaches where the deformations are applied using intracavitary probes and ablation electrodes have also been described . Reports on the significant improvement in the strain contrast with electrode displacement when compared to external deformation methods for quasi-static elastography have been presented . In addition, 3D strain imaging  and modulus reconstruction (solution of the inverse problem)  has also been demonstrated using this approach.
The methods described in this category utilize low-frequency deformations on the order of 1–10 Hz, to perturb tissue. Since the deformation frequencies are quite low, these are considered to be quasi-static in nature, and do not generate appreciable shear waves in tissue. Both well-controlled [24, 34] and freehand-based  approaches have been described. Hall et al. , described an approach that coupled real-strain strain imaging (7 frames per second) and visualization with freehand deformations termed ‘palpation imaging’. This approach was implemented initially on the Siemens Elegra, and later on the Antares and the S2000 (Siemens Ultrasound, Seattle, WA, USA), and constitutes one of the earliest commercially available clinical elastography systems on the market.
Physiological stimuli due to respiratory , cardiac muscle deformations [36, 37] and cardiovascular sources [9, 18, 45, 92–96] have been used for elastographic imaging. Investigators have utilized physiological stimuli to obtain both forward [35–37] and inverse problem  solutions for elastography. Physiological deformation sources, however, introduce challenges due to the non-uniform deformations introduced and the need for gating of the data to obtain reproducible results.
The success of elastographic imaging is predicated primarily on the several orders of the Young’s modulus contrast that exist between normal and abnormal tissue types [1, 2, 6]. This concept is shown as the initial modulus distribution in Fig. 2, whose depiction on an image is sought by all the elasticity imaging methods. A singular advantage of techniques based on exploiting these modulus variations, is that pathological changes in tissue generally alter the underlying elasticity of tissue. However, these same pathological changes may not alter the contrast mechanisms utilized by conventional imaging modalities.
The practice of elastography can be framed either as the solution of the forward problem, that relies on the estimation of the displacement vector, strain tensor or other related parameters, or the inverse problem that attempts to reconstruct the initial modulus distribution as shown in Fig. 2. In general, both the forward and inverse problem solution requires accurate and precise estimation of the local displacements with high spatial resolution and signal-to noise . This is in contrast to dynamic methods for tissue elasticity measurement that utilize variations in the velocity of the shear waves generated in tissue to obtain modulus distributions [22, 98, 99]. Dynamic methods for elasticity imaging and reconstruction are not discussed in this paper, which focuses primarily on quasi-static elastography.
Several groups have reported results on the Young’s modulus of soft tissues such as the lung, tendon, breast, prostate, liver, uterus and brain tissue. However, the measurement methods used vary as widely as the values reported [100–105]. Relatively fewer published results exist on human tissues such as breast, prostate, liver, uterus which are of interest for quasi-static elastography [106–111]. Measurements on human tissue stiffness by Sarvazyan et al. , Parker et al. , Walz et al. , Wellman , Krouskop , and Kiss , demonstrate the existence of stiffness contrast among normal tissues, and between normal and pathological tissues in the breast, prostate, thyroid, liver and the uterus.
Krouskop et al.  presented initial reproducible results on the modulus variations on breast and prostate tissue and also reported on their nonlinear stress strain behavior. These results were corroborated by Wellman , indicating that fibrous breast tissue was stiffer than glandular tissues, which were stiffer than adipose or fatty infiltrated tissues. Krouskop , also reported on the Young’s modulus values of tumors in the breast with infiltrating ductal carcinomas being significantly stiffer than in situ ductal tumors, which were in turn softer than the glandular, fibrous and fatty tissue in most instances. He also reported on prostate tissue with normal prostate tissue being stiffer than benign prostate hyperplasia and cancers of the prostate being significantly stiffer than the normal prostate.
Young’s Modulus measurements on liver tissue by Yeh et al. , show that liver tissue stiffness increase with cirrhosis. Hepatic tumors such as cholangiocarcinomas, focal nodular hyperplasia and hemangiomas also exhibit increased stiffness. However, hepatocellular carcinomas were found to be softer than healthy liver . Elastic modulus variations of liver tissue with different fibrosis grades were reported by Wen-Chun et al. .
This above section and associated references provides a glimpse into the several order of magnitude in the Young’s modulus of normal and pathological tissue types that has been utilized to provide new information on either the modulus of strain distribution in tissue.
The solution of the forward problem in elastography is based on the utilization of displacement vector and strain tensor images for clinical diagnosis. However, in order to obtain the strain distribution, the underlying tissue has to be deformed or stressed for the modulus distribution to be converted to a strain distribution. The contrast transfer efficiency concept describes the manner by which the modulus distribution is converted to an ideal strain distribution based on purely mechanical considerations as illustrated in Fig. 2. The contrast transfer efficiency concept was utilized by Ponnekanti et al.  to illustrate that the axial strain distribution was more efficient in depicting the modulus distribution of stiffer inclusions embedded in a softer background when compared to softer inclusions in a stiffer background. This study was corroborated by Kallel et al.  who also developed an analytic model to verify the finite element results presented by Ponnekanti et al. .
Tradeoffs in the displacement vector and strain tensor estimation process from the noisy pre- and post-deformed ultrasound radiofrequency signals has been described statistically using the ‘Strain Filter’ concept [11, 47, 118]. The strain filter has been defined as the statistical upper bound of the variation in the elastographic signal-to-noise ratio (SNRe) (ratio of the mean strain estimate to the theoretical lower bound of the standard deviation [41, 85–87]) as a function of the applied deformation. The strain filter provides a transfer characteristic that describes the filtering process in the strain domain that enables visualization of only a limited range of the local strains generated in tissue as a function of the applied deformation. The strain filter concept allows prediction of this limited range of strains based on the ultrasound system parameters and the signal processing algorithms and parameters used for strain estimation .
The forward problem in quasi-static elastography as illustrated in Fig. 2, is based on the three concepts or principles discussed in the preceding paragraphs . Conversion of the underlying modulus distribution to the strain distribution is clearly elucidated using the contrast transfer efficiency concept to obtain the ideal strain distribution. The statistical analysis of strain estimation described using the strain filter concept, demonstrates the tradeoffs between the system parameters and signal processing approaches in depicting the strain visualized in the estimated strain image. Finally, the combination of the contrast transfer efficiency and the strain filter principles lead to the prediction of the upper bound on the contrast-to-noise ratios obtainable in strain images, which quantifies the forward problem [119, 120].
The cross-correlation window length and the overlap between adjacent processing windows are two of the primary signal-processing parameters that impact the axial resolution obtained with quasi-static strain imaging [10, 121–123]. The length of the cross-correlation window at a fixed overlap was initially used as a measure of the axial resolution . Alam et al.  demonstrated that axial resolution can be expressed as a bilinear function of the window length and overlap, with the overlap being the more important factor. Longer duration windows also provide improved sensitivity to strain estimation, and the ability to estimate a larger range of local strains and improved SNRe and CNRe (as long as the increased signal decorrelation within the gated window was negated by the increased information content) at the expense of spatial resolution . In general, we can tradeoff reductions in the spatial resolution to significant improvements in the SNRe and CNRe and vice-versa [11, 124]. In a recent paper, Righetti et al.  illustrated that the ultimate limit on the axial resolution was directly proportional to the wavelength or inversely proportional to the fractional bandwidth of the ultrasound system. On the other hand, the ultrasound lateral beam width limits the achievable lateral resolution for quasi-static strain imaging . In a similar manner, the elevational beam-width would impact the elevational resolution. The axial, lateral and elevational resolution become important with the use of 2D transducers for 3D strain imaging .
Solution of the inverse problem in quasi-static elastography involve the utilization of the estimated displacement vector and strain tensor information to reconstruct the underlying modulus distribution as illustrated in Fig. 2. Accurate and reproducible determination of the modulus distribution would be the preferred image for clinical diagnosis since it would provide a direct image of the underlying modulus distribution. In addition, it would provide quantitative information as opposed to the qualitative information provided with displacement vector and strain tensor images, which depend to a large extent on the applied deformation.
However, the solution of the inverse problem is not straightforward and requires additional information on both the boundary conditions in effect during the applied deformation and the local stress distribution in tissue. Unfortunately the local stress distribution is generally unknown other than at the tissue surface where the deformation is applied. The difficulty in the computation of the local stress distribution within the tissue is a major stumbling block for modulus reconstruction. This has led to two different approaches for modulus reconstruction described in the literature, namely 1) Iterative modulus reconstruction [38, 91, 126] based on assumption on the uniformity of the stress distribution ., and 2) model (finite element or analytic) based reconstruction [97, 127] approaches.
Iterative methods [38, 91, 126] attempt to converge to a final modulus distribution based on the estimated displacement field using assumptions on the boundary conditions and uniformity of the stress distribution. Kallel et al.  utilized a Newton-Raphson algorithm to find the modulus distribution that provides the best reproduction of the estimated axial displacement field using Tikhonov regularization to achieve convergence in 8–10 steps. Model based methods [97, 127] on the other hand utilize prior information on the geometry of the imaged object to obtain analytic or finite element solutions that are utilized to aid in the convergence.
The ill-posed nature of the inverse solutions also introduce large noise artifacts and may result in non-unique solutions of the underlying modulus distribution . Barbone and Bamber  demonstrate that knowledge of only the displacement and boundary conditions remain insufficient to obtain unique solutions of the modulus distribution. Additional information such as the modulus itself or its derivatives on the boundary and stress distributions or additional strain tensor information are essential for unique modulus reconstruction [38, 80].
Although quasi-static elastography has been under development over the last two decades, it is only recently that commercial clinical ultrasound system manufacturers have introduced clinical products based on elastography. The availability of elastography modes on clinical system will rapidly increase clinical applications and organ system where elastography will be utilized in the future.
In general, two criteria have to be satisfied for successful clinical elastographic imaging, namely the ability to apply a quasi-static deformation and the ability to ultrasonically image the tissue being deformed. Superficial organs such as skin, breast, neck, thyroid, lymph nodes, deep vein thrombi are some of the best candidates for strain or modulus imaging [13, 15, 75, 76, 128–131]. Approaches to deform organs located deeper in the body have been reported, that use intracavitary ultrasound transducers, intracavitary balloons for imaging the prostate gland [132, 133], saline infusion for imaging the uterus  etc.. As described previously, physiological stimuli have been utilized to image the heart, vasculature, liver, and the coronary and carotid arteries [36, 37, 51, 135–140]. We discuss some of the clinical applications of quasi-static strain and modulus imaging in the following paragraphs.
Strain [13, 15, 75, 76, 128–130] and modulus  imaging of masses in the breast have been widely reported in the literature. Strain imaging in the breast was one of the first reported clinical application of quasi-static elastography , since many breast lesions are superficial and are known to be significantly stiffer than surrounding normal breast tissue . Strain imaging of the breast has been reported using well-controlled stepper motor based deformations (both single and multiple deformation steps) , real-time freehand palpation , and also the utilization of physiological deformation of the left breast due to cardiac activity.
Some of the early parameters utilized to differentiate benign from malignant masses was also reported on breast masses to differentiate malignant breast tumors from benign fibroadenomas. Since breast tumors are significantly stiffer (depicted in a darker gray scale as opposed to the lighter gray scale used for softer tissue) than surrounding normal tissue, the stiffness or strain contrast was initially used to differentiate breast masses, as shown in Fig. 3 for a patient with an invasive ductal carcinoma. Note that although the margins of the mass are not clearly visualized in the B-mode image in Fig. 3(a), it is clearly seen in the axial strain image in Fig. 3(b) as a darker (stiffer) region. In addition, it was found that invasive cancers in the breast were depicted as larger areas of stiffness in the strain images, when compared to the corresponding B-mode images , observed from comparing Figs, 3 (a) and (b). The increased stiffer regions observed in the strain images were hypothesized to be due to the desmoplastic reaction from the infiltration of the tumor cells into surrounding breast tissue. A second parameter referred to as the ‘size ratio’ derived from the ratio of the lesion dimensions derived from the strain and B-mode images was then proposed to differentiate between benign fibroadenomas from malignant tumors[13, 15]. Another parameter utilized in the differentiation is related to lesion mobility; since cancers are firmly attached to surrounding tissue, when compared to fibroadenomas that are more mobile and hence slip during the applied deformation [75, 76]. Parameters related to the normalized shear strain around the breast masses haven been utilized to quantify the attachment of the mass or lesion to the surrounding tissue as shown in Fig. 3(c). Observe the large shear strain areas (blue and red regions) that are clearly seen towards the top of the mass margins are indicative of firmly attached masses [75, 76]. Elastography also provides clear classification of cystic masses which are depicted with a central decorrelation spot appearing as a bright spot or region on the strain image. Finally, methods to assess the viscoelastic response by evaluating tissue creep have also been described .
The thyroid gland is another superficial gland that is a viable target for quasi-static elastography. Due to the accessibility of the thyroid gland, external deformation of the thyroid using the ultrasound transducer has been utilized by several groups [131, 142, 143]. Deformations introduced from pulsations due to blood flow through the carotid artery has also been utilized as a deformation source .
Assessments of the stiffness of lymph nodes represent another clinical application area for quasi-static elastography due to the superficial nature of these nodes enabling application of external deformations for strain imaging .
Deep vein thrombi or blood clots are known to progressively increase in stiffness with age, and a means to stage thrombi age is essential in their treatment. Quasi-static modulus imaging has been utilized to determine the thrombi age, since newer clots can be treated more effectively when compared to older clots .
Another area where in-vivo strain imaging has made inroads is in the imaging of the prostate gland. The prostate gland presents a more challenging environment for the application of strain imaging, due to its location, primarily for the application of the deformation required for strain imaging. Deformation states of the prostate has been varied utilizing different levels of saline within the balloon , and deformations applied using the trans-rectal ultrasound transducer (TRUS) . The TRUS transducer also provides data in a curvilinear format when compared to linear array transducers utilized for the breast, small parts and imaging of the carotid.
Evaluation of focal uterine masses has also been explored using trans-vaginal ultrasound transducer (TVUS) and utilizing saline infusion to provide the deformation . Ex-vivo quasi-static strain imaging of different uterine pathology has been reported in the literature . Another application area for strain and modulus imaging is the evaluation of cervical stiffness [149, 150] which may have implications for the evaluation of patients at risk for pre-term labor.
Treatment monitoring of ablative therapies is probably one of the more natural applications for strain and modulus elastography, since heating tissue induce denaturation of proteins which in turn elevates the Young’s modulus of ablated tissue. Strain imaging has been utilized to monitor ablative therapies such as high-intensity focused ultrasound , radiofrequency and microwave ablation procedures . An interesting offshoot is the utilization of the radiofrequency or microwave electrode itself to introduce the quasi-static deformation for both strain  and modulus  imaging. This method has been utilized for in vivo imaging of thermal lesions in the kidney and liver [132, 133]. 3D strain imaging has also been reported with electrode displacement elastography . An example of in-vivo electrode displacement based strain imaging is illustrated in Fig, 4, for a thermal lesion created in the liver of a porcine animal model using a Cool-tip™ radiofrequency ablation system (Valleylab, Boulder, CO).
Quasi-static elastography has also been utilized for imaging the poroelastic properties of tissue that accumulate fluid, for example for evaluating lymphedema, which involves an abnormal interstitial accumulation of lymphatic fluid causing tissue swelling [39, 77, 151]. Variations in the ratio of the lateral to axial strain values with deformation provide estimates of the fluid content and the ability to differentiate between normal and edematous tissues .
Intravascular ultrasound (IVUS) based strain imaging of coronary arteries using pulsations introduced due to cardiac activity has been widely described in the literature and is probably the most common clinical application involving physiological stimuli[9, 18, 45, 92, 93]. The most common implementation utilizes radiofrequency data acquired at two different intraluminal pressure levels near diastole. IVUS elastography involves the acquisition of high-frequency (20 MHz or higher) radiofrequency data from single element or array transducers located at the tip of a catheter that is inserted into the coronary artery to be evaluated. In addition to the strain distribution modulus maps have also been generated . One of the challenges in intravascular elastography and carotid strain imaging described in the next sub-section is the identification and differentiation of ‘vulnerable plaque’ or plaque prone to rupture .
Most of the reported strain and modulus imaging studies for plaque characterization utilize intravascular ultrasound on coronary arteries [139, 140] as described above. However, the carotid artery is another superficial location that provides easy access to clinical ultrasound equipment with linear array transducers [51, 135–138], and is also amenable to strain imaging. Figure 5, presents a B-mode and axial strain image for a patient with a strip of softer plaque observed at a depth of 2 cm and extending from 1.5 to 4 cm. The strain image also classifies the remainder of the plaque in the vessel as relatively stiffer plaque visualized as the mid-gray gray scale in Fig. 5(b) [51, 135, 138]. Investigators have reported on estimation of the Young’s modulus by solving the inverse problem using the estimated strains and mechanical models of the carotid artery . However, both the carotid and coronary artery models have to account for hemodynamic parameters that vary significantly with stenosis for inverse problem solutions.
Cardiac-elastography or myocardial strain imaging where the strain distribution is imaged over the entire contraction and relaxation of the heart (cardiac cycle) [36, 37] is another clinical application that utilizes physiological stimuli. B-mode speckle tracking for strain imaging is currently utilized, primarily due to limitations associated with tissue Doppler-derived velocity and strain estimates [153–155]. Both General Electric Medical Systems (GE Healthcare, Milwaukee, WI, USA) and Siemens (Siemens Ultrasound, Mountain View, CA, USA) have rolled out clinical cardiac ultrasound systems equipped with 2D speckle tracking methods for strain imaging. Strain imaging that utilizes radiofrequency signals would provide significantly improved strain sensitivity, when compared to the B-mode based approaches [36, 156]. A short-axes B-mode and strain image obtained using radiofrequency data, estimated using a hybrid 2D algorithm developed for curvilinear transducers  is shown in Fig. 6. Observe that the strains depicted in the image for the cardiac muscle is mostly compressive (blue color), since this image was obtained during systole. However, the frame rates at which radiofrequency data are acquired is an important factor to obtain unbiased and robust estimation of tissue displacements and strain . Frame rates for B-mode data are significantly higher than that for radiofrequency data. Approaches to obtain high precision radial, circumferential and longitudinal strains are also necessary for the widespread clinical application of this modality. Estimation of principal strain components, which are angle-independent, is one such approach that may provide reproducible diagnostic measures for evaluating cardiac disease .
Deformations of the liver and other abdominal organs introduced due to respiration, for example diaphragmatic deformations has been utilized for imaging thermal lesions created n the liver . Cardiovascular motion, has also been utilized for in-vivo elastographic imaging of the liver. However, all of these approaches require gating of the data acquisition to the respiratory or cardiac waveform to ensure similar deformation increments and for reproducible imaging.
The last two decades have seen tremendous advances in the development and clinical practice of quasi-static ultrasound elastography in the depiction of quality displacement vector, strain tensor and modulus distributions. Currently four different commercial ultrasound system manufacturer’s offer quasi-static based strain imaging modes on their clinical systems. Palpation based strain imaging is currently available on both the Siemens Antares and the S2000 systems (Siemens Ultrasound, Seattle, WA, USA). Hitachi Medical systems (Hitachi Ltd., Tokyo, Japan) provide elastography modes on their scanners. Ultrasonix Medical Corporation (Vancouver, BC, Canada) recently introduced the SonixTOUCH system with an elastography mode. General Electric Medical Systems (GE Healthcare, Milwaukee, WI, USA) and Siemens have also introduced 2D speckle tracking on their cardiac ultrasound systems. The availability of quasi-static based strain imaging modes on clinical system will further spur the development of this new imaging modality, since it provides additional information not present with current clinical imaging modalities.
Although, quantitative i.e. modulus imaging with quasi-static elastography would provide the best visualization of the underlying modulus distribution in tissue, this mode requires additional development to reduce noise artifacts and to obtain unique solutions. This would require the development of techniques for accurate and precise estimate of both the local strain and stress distributions at high spatial resolutions, and research is ongoing in this area. We anticipate continued improvements in the spatial resolution and signal to noise ratios in the estimated strain tensor distribution with improved algorithms and processing techniques. Development of methods for accurate estimation of the local stress distribution has to be developed, with techniques that use force sensors on ultrasound transducers under evaluation.
This paper does not review dynamic elastography based methods, which have also made significant strides over the last decade. Commercialization of these approaches include the FibroScan® system developed by Echosens (Echosens SA, Paris, France), a device that tracks shear wave speed for monitoring and staging hepatic fibrosis , and the S2000 (Siemens Ultrasound, Seattle, WA, USA) equipped with the Virtual Touch(TM) software a acoustic radiation force based imaging mode. Some of these recent developments to obtain modulus estimates of small regions in tissue may enable more accurate staging of diffuse diseases. Several excellent review papers exist in the literature that would provide the interested reader with additional insights into this novel imaging modality [160–164].
I would like to thank Dr. Min Rao, Ph.D, Dr. Hao Chen Ph.D, Mr. Matthew McCormick, Mr. Nick Rubert and Ms. Haiyan Xu for providing the results used in this paper. This work was funded in part by NIH grants R01 CA112192-02, R21 CA140939-01 and Komen Foundation grant BCTR0601153.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.