Shear Wave Elastography for the Assessment of Carotid Plaque Vulnerability: A Systematic Review

,


Introduction
Treatment of carotid artery disease remains controversial, most especially in relation to both asymptomatic carotid stenosis as well as possibly symptomatic nonstenosing carotid plaques, without concrete outlined clinical recommendations [1,2]. Patients are mostly selected for carotid intervention depending on the extent of carotid luminal stenosis and the presence of clinical symptoms [2,3]. The occurrence of ipsilateral stroke, however, is not solely dependent on the extent of carotid stenosis, but also on the extent of plaque stability [4]. Carotid plaques which cause ipsilateral stroke are believed to be unstable due to their structure and composition and could therefore be a source of atherogenic emboli [5]. This characteristic of plaques is described as plaque vulnerability. Carotid imaging is therefore desirable, irrespective of the degree of stenosis, for risk stratification which could help in identifying patients with vulnerable plaques and high risk of cerebral thromboembolism, i.e., patients who might benefit from aggressive treatment including carotid surgery, stenting, or prolonged dual antiplatelet therapy. On the other hand, asymptomatic patients with stable carotid plaques could be spared the risks of unnecessary and potentially harmful procedures and treatments including peri-interventional ischemic or hemorrhagic stroke, and even death [3].
Plaque vulnerability can best be assessed histologically through the invasive acquisition of plaque tissue [6]. However, the associated risks render this procedure not ideal for the sole purpose of diagnostics. Contrast-enhanced magnetic resonance plaque imaging (MRPI), which has high diagnostic accuracy in assessing individual characteristics of plaques and carotid plaque vulnerability [4], however, is expensive and not easily available. Likewise, conventional B-mode ultrasound enables evaluating individual characteristics of plaque vulnerability including lipid-rich necrotic core volume, fibrous cap thickness, and intraplaque hemorrhage but is challenged by high interoperator variability [7,8]. In assessing plaque vulnerability, contrast-enhanced ultrasound (CEUS) may be used to detect intraplaque neovascularization and ulceration of plaques. Neovascularization can also be assessed by superb microvascular imaging (SMI) through visualization of intraplaque microvascular low-velocity blood flow [9]. Computed tomography angiography (CTA) can be used to detect plaque calcification, ulceration, or intraplaque hemorrhage [10,11]. MRPI, CEUS, and CTA expose patients to contrast agents, which might be inconvenient in patients with contraindications, with the latter additionally exposing patients to radiation (Supplemental Table S1) [12].
Shear wave elastography (SWE) is a novel ultrasound technique which offers, in addition to the functions of conventional B-mode ultrasound, a multidimensional colorimetric map representation of tissues for the quantification of tissue stiffness [13]. It employs acoustic radiation force impulse (ARFI), which induces the propagation of mechanical shear waves in tissues ( Figure 1). The propagated shear wave, which travels orthogonally with respect to the ARFI [14], is then tracked by means of ultrafast ultrasound wave imaging allowing quantitative mapping of mechanical tissue properties in the form of a real-time colorimetric map [15]. The tissue stiffness, expressed in Young's modulus (YM), based on the formula YM = 3ρc 2 (ρ, tissue density; c, shear wave velocity) [16], is proportional to the squared velocity of the propagated shear waves. Quantification of tissue stiffness could therefore be expressed in YM (kPa) or speed (meter/second). Since plaque vulnerability is invariably dependent on the extent of stiffness, quantification of plaque stiffness could therefore predict its vulnerability [17].
This systematic review, compared to previous studies [18,19], includes more and recent prospective studies and narrowed its scope on evaluating the feasibility of SWE in carotid plaque risk stratification in relation to ischemic stroke (PROSPERO registration number CRD42022309709).

Materials and Methods
This study was conducted in accordance with the Preferred Reporting Item for Systematic Reviews and Meta-Analysis (PRISMA). We systematically searched the databases PubMed, Cochrane, Web of Science, and Embase for all published studies from their inception until November 1, 2022, which evaluated the feasibility of SWE in assessing carotid plaque vulnerability (Supplemental Material). Quality assessment was conducted in accordance with the quality assessment of diagnostic accuracy studies-(QUADAS-) 2 tool [20]. Search and quality assessment was performed by two independent reviewers, and disagreements were resolved through a third reviewer. Further details are described in the Supplementary Material. If not specified, plaque stiffness values are presented as mean.

Results
The search yielded 593 studies out of which 18 studies that investigated our primary aim of evaluating the feasibility of SWE in carotid plaque risk stratification were included in the systematic review ( Figure 2). The sample size of human studies ranged from 20 to 199 patients. Eleven out of 13 human studies did this by comparing vulnerable/symptomatic with stable/asymptomatic plaques ( Table 1). Ten of these eleven studies as well as two additional human studies compared SWE to other plaque imaging modalities ( Table 2). In all, five human studies used histology as a standard to assess plaque vulnerability [10,[21][22][23][24]. Five studies evaluated the feasibility of SWE in carotid plaque risk stratification by assessing tissues of varying mechanical stiffness in carotid phantom models (   Five out of seven studies which compared plaques of symptomatic and asymptomatic patients reported YM of symptomatic plaques to be lower (27-52 kPa) compared to asymptomatic plaques (38-115 kPa) (p < 0:05) [9,[25][26][27][28]. In line with these findings, one study found mean shear wave velocity (SWV) to be negatively associated with symptomatic ischemic stroke (binary logistic regression coeffi-cient -0.624, constant -0.355, p = 0:004, odds ratio 0.54, 95% confidence interval 0.35-0.82) [29]. In the remaining study, SWE could not distinguish between plaques of symptomatic and asymptomatic patients (Table 1) [24].

SWE Compared to Other Plaque Imaging Modalities.
In the only study obtained from our search that used MRPI as a comparator, SWE could well discriminate the particularly rupture-prone American Heart Association (AHA) type VI plaques (7.3 vs. 3.1-3.6 m/s, p < 0:004) [22]. All other evaluated AHA plaque types (Supplemental Table S2), however, indicated similar shear wave velocities (Table 2) [22].
Eight studies compared SWE to grayscale quantification using B-mode ultrasound ( Table 2) [23][24][25][26][28][29][30][31]. Hyperechoic plaques [32] corresponded to higher YM (51-57 kPa) and higher shear wave velocity (3.9-4.7 m/s), while echolucent plaques on the other hand corresponded to lower YM (11-17 kPa) and lower shear wave velocity (1.5-2.6 m/s) (Supplemental Table S2) [29,30]. Data from one study showed a strong correlation between YM values and grayscale median (GSM) (r = 0:728, p < 0:05) [25], and three studies found SWE to be superior to GSM in identifying vulnerable or stable plaques using histology as reference [23] and symptomatic or asymptomatic plaques based on clinical assessment [26,28]. Despite showing a wide range of variation of absolute, mean, and median YM values across each plaque type, one pilot study observed a high correlation between YM and plaque echogenicity [31]. In another study, none of the image texture parameters  Figure 2: Mechanism of shear wave elastography (SWE). Acoustic radiation force impulse (ARFI) emitted by the SWE probe causes tissue displacement, inducing the propagation of mechanical shear waves, which travel orthogonally relative to ARFI. Tissue displacement is monitored using real-time B-mode imaging. Based on tissue displacement, the velocity of shear wave propagation (c) within the tissue is calculated, which is then incorporated in Young's modulus (YM) formula, i.e., YM = 3ρc2, to estimate the tissue stiffness. Tissue stiffness is represented as a colorimetric map superimposed on the acquired B-mode image in real time.      (Table 2), although secondary findings from the same study as previously reported hinted on the feasibility of SWE in carotid plaque risk stratification (Table 1) [24].
No comparative studies with SWE were found for ultrasound-based strain elastography, PET, and SPECT.

SWE Assessment of Tissues with Varying Mechanical
Stiffness in Carotid Phantom Models. Data from all five carotid phantom models showed lower and higher YM values to be representative of soft and harder tissues, respectively [33][34][35][36][37]. Results from four carotid plaque phantom studies showed the feasibility of SWE in discriminating vessel wall from soft [33][34][35][36] and hard [36] plaques using mechanical testing as standard (Table 3). One additional study indicated the reproducibility of SWE in discriminating vessel wall from both soft and hard plaques (Table 3) [37].

Reproducibility of SWE in Carotid Plaque Risk
Stratification. Data from four studies showed a similar interobserver coefficient of variation in SWE application, depicting a good reproducibility of SWE. These were 0.16 [25], 0.18 [29], 0.22 for the vessel wall and 0.19 for carotid plaques [26], and 0.08-0.20 in a phantom study [37]. Intraobserver coefficient of variation ranging from 0.07-0.22 was reported by one study [37] and 0.14 by another [29]. Further results showed 97% of agreement between two independent observers (p < 0:05) [25].

Discussion
In this study, we evaluated the feasibility of SWE in carotid plaque risk stratification in relation to ischemic stroke, which could potentially provide a cheaper and more convenient alternative to guide treatment strategies in patients with carotid artery disease.
Four out of five studies showed a good correlation of SWE with histology, which is the most accurate means of plaque vulnerability assessment, and also the feasibility of SWE to distinguish vulnerable from stable plaques [10,[21][22][23]. Although the remaining one study could not significantly distinguish vulnerable and stable plaques by mean stiffness values, secondary findings suggest feasibility in carotid plaque risk stratification [24].
Plaques which stand the risk of rupture and distal embolization, and therefore, increase the risk of ischemic strokes 8 Acta Neurologica Scandinavica are termed "vulnerable" [38]. It should be noted, that the term "symptomatic plaques," as used in some literature, might be ambiguous since these plaques have been described as symptomatic based on clinical assessment and not on their histological vulnerability. In some cases, the term "symptomatic plaques" may have been related to hemodynamically relevant carotid stenosis or may even have been masked by the stroke of other etiologies, e.g., lacunar infarction. Nonetheless, out of seven human studies which assessed symptomatic and asymptomatic plaques using SWE, five [9, 25-28] showed a significant difference in YM values between these two groups and one [29] reported a significant correlation of lower stiffness values with symptomatic patients (Table 1). A combination of YM with a degree of stenosis could thus improve diagnostic performance and provide more reliable risk stratification [25]. Studies involving human participants, although seemingly similar in study design, reported different YM values for vulnerable/symptomatic as well as for stable/asymptomatic plaques. Nonetheless, YM values differed significantly between vulnerable/symptomatic and stable/asymptomatic plaques within each study, and results were highly reproducible [25,29,37]. However, an overlap in mean YM values between studies seems to exist (Table 1). Therefore, the use of the absolute YM values obtained from these studies to distinguish between vulnerable/symptomatic and stable/ asymptomatic carotid plaques might not be applicable in clinical practice. Overlap of YM values between studies might potentially be related to differences in SWE devices as well as other technical variables such as acquisition and imaging settings [14,15], and the type of velocity analysis [14][15][16]36]. Aside from technical differences, procedural variations including the type of imaging plane [19] and influences of pulsatile movements of the artery by blood flow [19,39] or breathing movements [25] might compromise interpretation through the generation of artefacts [10]. We summarize possible means to address these problems in Supplemental Table S3.
Imaging techniques and modalities used for assessing plaque vulnerability are based on their potential to detect one or more factors of plaque vulnerability. These factors, which are most accurately assessed by means of histology, include the thickness of the fibrous cap, the size of the lipid-rich core, and the presence of an inflammatory infiltrate, intraplaque hemorrhage, and/or ulcerations [6]. Due to the higher proportion of lipid core, susceptibility to hemorrhage, and ability to form thrombus, vulnerable plaques are described as "soft plaques." These are echolucent on conventional B-mode ultrasound and have lower YM values ( Figure 3) [23]. On the other hand, stable plaques are more calcified with a higher fibrous content on histology, echogenic on conventional ultrasound, and have higher YM 9 Acta Neurologica Scandinavica values [40]. Data from six out of eight studies confirm this association between SWE and B-mode ultrasound, by means of grayscale quantification (Table 2) [23,25,26,[29][30][31]. Data from three of these eight studies even hint at the superiority of SWE to GSM in discriminating histologically confirmed vulnerability from stable plaques [23] and clinically symptomatic from asymptomatic plaques [26,28]. SWE shows a similar diagnostic capability to SMI, which is well known for detecting intraplaque neovascularization, in discriminating symptomatic and asymptomatic plaques in the study population [9]. CEUS and CTA are known to facilitate the assessment of plaque vulnerability through detecting plaque neovascularization as well as ulceration [10,11,21]. In detecting plaque neovascularization and ulceration, data show that SWE is equally sensitive as compared to CEUS and CTA [10]. Compared to CEUS and SWE, CTA has a higher specificity [10], probably due to its capability of additionally characterizing plaque vulnerability through the detection of intraplaque hemorrhage, and to some extent, the lipid-rich necrotic core [11].
MRPI is currently the gold standard for carotid plaque imaging [4,5]. With MRPI, carotid plaques are stratified according to the modified AHA classification, which is based on histological hallmark (Supplemental Table S2) [41,42]. Of most, relevance to carotid plaque vulnerability is AHA type VI plaques, which are complex plaques with possible surface defect and are prone to hemorrhage and thrombus formation [41]. Data shows the feasibility of SWE in distinguishing AHA type VI from other plaque types as assessed on MRPI [22]. Although SWE cannot discriminate between other AHA plaque types [22], for the purpose of carotid plaque risk stratification, this plays no role since the ultimate aim is to differentiate vulnerable from stable plaques.
Supporting the data gathered from real patients with carotid artery disease, studies using SWE in carotid artery phantom models show the capability of SWE in distinguishing plaques of different stiffness as well as the vessel wall itself (Table 3). YM values for vessel walls and soft/hard plaques obtained by either mechanical testing or SWE seemed to vary between studies which is not surprising due to differences in phantom construction. Out of four studies which used mechanical testing as reference, three showed comparable tissue stiffness values to that obtained by SWE, which reflects the potential of SWE for clinical application [33,34,36].
As a limitation of SWE, the estimation of YM is based on the assumption that the shear wave is propagated in an incompressible, homogeneous, and isotropic tissue [16]. Most plaques on the other hand are small and anisotropic with irregular morphologies and heterogeneous composition, compromising the conversion of SWV to YM. Thus, reporting SWV rather than YM as an estimation of plaque stiffness may avoid values that are based on incorrect assumptions [24]. Also, access to carotid plaques is hindered by factors such as plaques near the origin of the common carotid artery or plaques in the posterior and medial walls of the internal carotid artery [25], as well as in obese patients, due to excessive signal attenuation by underlying tissues [26]. Available data does not report any safety concerns of SWE, and the force of the wave pulse is known to be a thousand times lower compared to the force of blood pressure on the vessel wall [43]. However, external plaque compression by the transducer during examination just as in any ultrasound-based carotid plaque imaging [44] does not only pose the threat of increased plaque instability and distal embolization especially in vulnerable plaques, but also compromises tissue compression stiffness measurement by increasing the inherent tissue stiffness (Supplemental Table S3) [45][46][47][48]. It should also be taken into consideration that the different devices and device programming used in these studies could also be a potential source of variability [15].

Limitations of Study
The use of histology as a comparator for plaque vulnerability was reported by five studies. There was heterogeneity in the remaining human studies which assessed the feasibility of SWE by comparing SWE to other modalities of carotid plaque imaging. These comparators have their respective inherent limitation in assessing plaque vulnerability and are, as such, not the most ideal for assessing plaque vulnerability. Three of the studies compared SWE to other carotid plaque diagnostic modalities without using a gold standard as a reference [29][30][31]. Other sources of heterogeneity in the included studies were with respect to models (i.e., phantom and human) and methodology, for which reason we resorted to conducting a systematic review without a meta-analysis.

Conclusions and Future Considerations
Although SWE cannot be used to evaluate individual characteristics of plaque vulnerability, i.e., thin fibrous cap, ulceration, large lipid-rich necrotic core, and intraplaque neovascularization, SWE seems to assess plaque vulnerability through quantification of YM independent of the underlying cause of the vulnerability and for that matter, with good reproducibility. Several studies have hinted on the feasibility of SWE in carotid plaque risk stratification. In spite of the limitations, SWE could potentially facilitate the identification of patients with vulnerable carotid plaques for timely intervention and therapy most especially in institutions where alternatives to MRPI, which is expensive and not readily available, might be needed. On the other hand, patients with stable plaques could be spared possible complications associated with thromboendarterectomy, stenting, and/or dual antiplatelet therapy. However, in order to derive more accurate vascular acquisition settings for the reproducibility of SWE in assessing carotid plaque vulnerability, defining YM thresholds for vulnerable plaques and for predicting restroke, larger prospective multicentric clinical studies would be necessary. These studies would benefit from using histopathology where applicable (i.e., when thromboendarterectomy is originally indicated) or MRPI as comparators to SWE.

Abbreviations
AHA: American Heart Association ARFI: Acoustic radiation force impulse