Plaque hemorrhage in carotid artery disease: Pathogenesis, clinical and biomechanical considerations

Abstract

Stroke remains the most prevalent disabling illness today, with internal carotid artery luminal stenosis due to atheroma formation responsible for the majority of ischemic cerebrovascular events. Severity of luminal stenosis continues to dictate both patient risk stratification and the likelihood of surgical intervention. But there is growing evidence to suggest that plaque morphology may help improve pre-existing risk stratification criteria. Plaque components such a fibrous tissue, lipid rich necrotic core and calcium have been well investigated but plaque hemorrhage (PH) has been somewhat overlooked. In this review we discuss the pathogenesis of PH, its role in dictating plaque vulnerability, PH imaging techniques, marterial properties of atherosclerotic tissues, in particular, those obtained based on in vivo measurements and effect of PH in modulating local biomechanics.

1. Introduction

Carotid artery disease is responsible for about 30% of ischemic strokes (Levy et al., 2008), which is the third leading cause of mortality and the primary cause of disability in developed countries (Roger et al., 2011). Currently, carotid luminal stenosis is the only validated diagnostic criterion for clinical risk stratification. Large multicentre clinical trials have shown that carotid endarterectomy (CEA) provides maximum benefit to patients with a significant carotid stenosis (≥70%) (Barnett et al., 1998). The overall risk-to-benefit ratio for CEA however becomes marginal in patients with moderate stenosis (50–69%). Since patients with moderate carotid stenosis constitute the majority of individuals suffering from clinical events (Barnett et al., 1998, Rothwell et al., 2003), there is therefore a need to identify better risk stratification tools besides luminal stenosis.

The primary mechanism for an ischemic stroke due to carotid atherosclerotic disease is believed to be an embolic event from a ruptured carotid plaque (Rothwell et al., 2000). As a multi-component structure, plaque stability is determined by both its structure and local haemodynamics, including arterial pressure and flow. A typical carotid atherosclerotic plaque is composed of lipid-rich necrotic core (LRNC), plaque hemorrhage (PH) and calcium, all covered by an overlying fibrous cap (FC). Large LRNC with or without PH and thin FC characterize a high-risk plaque. Although plaque components such as LRNC, FC and calcium have been extensively studied, PH has been much less investigated. There is growing evidence that PH is critical in dictating plaque vulnerability and is associated with subsequent ischemic events (Kolodgie et al., 2003, Michel et al., 2011, Sadat et al., 2010b). PH has been observed to be more prevalent in acutely symptomatic patients than in asymptomatic individuals (Imparato et al., 1983, Sadat et al., 2009), while prospective studies have confirmed that PH confers additional risk in both symptomatic and asymptomatic patients (Altaf et al., 2008, Altaf et al., 2007, Eliasziw et al., 1994, Sadat et al., 2010b, Singh et al., 2009, Takaya et al., 2006).

Compositional features provide complementary information to luminal stenosis in carotid plaque vulnerability assessment. Indeed, about 60% symptomatic patients exhibit PH or FC rupture at baseline (Gao et al., 2007, Milei et al., 2003), yet only 10–20% will go on to experience a recurrent event at 1 year (Sadat et al., 2010b, U-King-Im et al., 2009). It is therefore clear that composition alone cannot serve as a marker for prospective cerebrovascular risk and there is a need to identify novel biomarkers. Under the physiological conditions, carotid plaques are subjected to mechanical loading driven by pulsatile blood pressure. FC rupture may occur when this loading exceeds its material strength. Therefore, a strategy integrating both high risk anatomical plaque features and mechanical conditions may improve patient risk stratification and ultimately guide clinical treatment. With this consideration, this review will focus on (1) the pathological features of PH and its clinical significance; (2) the non-invasive in vivo imaging techniques used to depict PH; (3) the material properties of atherosclerotic components, including PH; and (4) the critical mechanical conditions within a hemorrhagic plaque.

2. Pathology of plaque hemorrhage and clinical significance

2.1. Plaque hemorrhage and associated neovascularization

The involvement of hemoglobin-rich plaque hemorrhage in the transformation of a stable to unstable atherosclerotic lesion was proposed as early as 1936 (Paterson, 1936). At the same time it was suggested that rupture of neovessels could be the initiating event for PH (Paterson, 1938, Wartman, 1938). After this initial period, the clinical and biological significance of PH became rather overlooked. The majority of biological studies focussed on either lipid metabolism or the inflammatory response within the plaque. Imparato et al. (1979) reported the relationship between PH observed in CEA samples and neurological symptoms. Since then various studies have assessed this relationship which are summarized by Gao et al. (2007). Others have investigated the relationship between PH and plaque neovessels (Fryer et al., 1987, McCarthy et al., 1999). Observations that erythrocyte extravasation and PH are related to high density of plaque neovessels, in the absence of plaque fissuring, support the common viewpoint that PH is related to neovessels leakage (Virmani et al., 2005), due to leaky endothelial junctions (Jeziorska and Woolley, 1999) (Fig. 1). An alternate viewpoint is that repeated plaque fissuring and associated formation of non-occlusive luminal thrombus gets incorporated into the plaque (Davies and Thomas, 1985). Either way, the extracorpuscular hemoglobin released following the phagocytosis of red blood cells in PH acts as a pro-inflammatory agent, promoting local inflammation (Fig. 2) and plaque progression (Moreno et al., 2012). PH also carries proteolytic enzymes, causing thinning of the FC and making the plaque liable to rupture (Michel et al., 2012). Readers are directed to other excellent articles (Levy and Moreno, 2006, Michel et al., 2011) for further relevant pathological details, as they are beyond the scope of this review.

Fig. 1.

Atherosclerotic plaque sample collected from carotid endarterectomy of a 72-year old symptomatic patient (male) showing both old thrombus and plaque hemorrhage around neovessels: (A and D) Van Gieson's (EVG) and Hematoxylin and eosin (HE) stains showing elastin (black in EVG), collagen (pink in HE) and erythrocyte and its debris (brown in EVG and red in HE); (B and E) local views of 1 in (A) and 3 in (B) showing neovessels in a old thrombus resulted from previous rupture (marked by arrow in A and D); (C and F) local views of 2 in (A) and 4 in (B) showing erythrocytes (fresh plaque hemorrhage) around neovessels; red asterisk: arterial lumen; green asterisk: the lumen of neovessels. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 2.

Aggregation of macrophages around neovessels (A) hematoxylin and eosin (HE) stain showing neovessels within a fibrotic region; (B) CD68 stain showing the surrounding macrophages; green asterisks: lumens of neovessels. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

2.2. Prevalence of plaque hemorrhage (PH) in patients with carotid artery disease

The prevalence of PH in symptomatic and asymptomatic patients has been mostly obtained from ex vivo histological examination of CEA or cadaveric tissue. In the 29 histological studies (Table 1) including both symptomatic and asymptomatic patients, PH was reported to range from 18.6% to 97.1% (median [inter quartile range (IQR)]: 72.8% [53.1, 85.9]) in symptomatic patients and from 7.4% to 91.3% (52.6% [37.3, 72.6]) in asymptomatic patients. Eight studies showed that PH had a higher prevalence in symptomatic patients; however 21 studies found no difference in PH prevalence between patient groups. Importantly none of studies conducted after 1995 found a difference in the rates of PH. This could be a result of preventive treatment (including statins), which came into widespread clinical use around this time. Nevertheless, if data were pooled, PH was more prevalent in symptomatic compared to the asymptomatic patients [70.3% (1937 out of 2754) vs. 52.0% (669 out of 1287); p<0.0001].

Table 1.

The prevalence of histologically identified plaque haemorrhage (PH) in symptomatic and asymptomatic patients.

Reference (year)	Symptomatic			Asymptomatic			p-Value
	PH (%)	PH (n)	Patient (n)	PH (%)	PH (n)	Patient (n)
Lusby et al. (1982a)	92.5	49	53	26.9	7	26	<0.0001
Reilly et al. (1983)	73.0	27	37	53.8	7	13	0.354
Persson et al. (1983)	97.1	33	34	52.4	11	21	0.0002
Ammar et al. (1984)	90.9	40	44	78.1	25	32	0.218
Persson (1986)	95.0	93	98	61.3	38	62	<0.0001
Ricotta et al. (1986)	76.5	26	34	23.5	8	34	<0.0001
Ammar et al. (1986)	95.5	42	44	78.1	25	32	0.051
Lennihan et al. (1987)	45.9	56	122	52.6	40	76	0.439
Fryer et al. (1987)	81.7	58	71	45.0	9	20	0.003
Bassiouny et al. (1989)	67.7	21	31	85.7	12	14	0.368
AbuRahma et al. (1990)	60.4	61	101	9.4	5	53	<0.0001
Bornstein et al. (1990)	85.9	61	71	83.3	5	6	0.663
Sterpetti et al. (1991)	68.2	30	44	37.3	25	67	0.003
von Maravic et al. (1991)	93.3	14	15	91.3	21	23	0.699
Feeley et al. (1991)	56.8	25	44	37.5	3	8	0.532
Arapoglou et al. (1994)	77.2	44	57	37.5	12	32	0.0005
Fisher et al. (1994)	53.1	17	32	72.6	45	62	0.098
Sitzer et al. (1995)	22.2	6	27	7.4	2	27	0.251
Seeger et al. (1995)	47.7	10	21	54.5	12	22	0.888
Carr et al. (1996)	84.2	16	19	56.0	14	25	0.096
Bassiouny et al. (1997)	18.6	11	59	17.5	7	40	0.920
Milei et al. (1998)	50.6	43	85	51.3	41	80	0.920
Montauban van Swijndregt et al. (1999)	93.9	31	33	71.4	10	14	0.101
McCarthy et al. (1999)	61.5	8	13	20.0	3	15	0.063
Tegos et al. (2000)	67.4	31	46	76.0	19	25	0.624
Milei et al. (2003)	23.3	31	133	29.1	43	148	0.340
Turu et al. (2006)	49.1	27	55	37.9	11	29	0.454
Hellings et al. (2010)	72.8	484	665	69.9	107	153	0.543
Derksen et al. (2011)	81.4	542	666	79.7	102	128	0.740

χ² test in R 2.10.1 (The R Foundation for Statistical Computing) was used for the statistical analysis and significant difference was assumed if p<0.05.

High resolution, multi-contrast MRI has been shown to be the most promising technique to depict PH with the sensitivity and specificity being 89.7±6.8% and 85.6±9.1% respectively (Bitar et al., 2008, Chu et al., 2004, Kampschulte et al., 2004, Moody et al., 2003, Ota et al., 2010, Qiao et al., 2011, Yim et al., 2008). MRI-based clinical studies (Table 2) have confirmed that PH has a higher prevalence in symptomatic lesions (symptomatic 33.3%; 192/576 vs. asymptomatic 19.4%; 131/675; p<0.0001). This suggests a potential role of PH in the etiology of stroke. It is only through well designed future studies, with uniformity in the definition and evaluation of PH, that the full mechanism will be elucidated (Gao et al., 2007).

Table 2.

MRI-depicted plaque haemorrhage (PH) in symptomatic and asymptomatic patients/sides.

Reference (year)	Symptomatic			Asymptomatic			p-value
	PH (%)	PH (n)	Patient (n)	PH (%)	PH (n)	Patient (n)
Saam et al. (2006)a	100.0	23	23	87.0	20	23	0.232
U-King-Im et al. (2008)	44.4	8	18	15.8	3	19	0.122
Sadat et al. (2009)	45.0	18	40	5.0	1	20	0.004
DeMarco et al. (2010)b	67.0	4	6	25.0	11	44	0.106
DeMarco et al. (2010)c	43.0	3	7	36.0	12	33	0.920
Cheung et al. (2011)a	13.0	31	217d	7.0	14	217e	0.045
Turc et al. (2012)	33.0	38	114	31.0	37	120	0.791
Lindsay et al. (2012)	22.0	9	41	15.0	6	40	0.603
Qiao et al. (2012)	62.5	15	24	30.4	7	23	0.056
Millon et al. (2013)	39.0	20	52	16.0	16	102	0.003
Grimm et al. (2013)	67.6	23	34	11.8	4	34	<0.0001

χ² test in R 2.10.1 (The R Foundation for Statistical Computing) was used for the statistical analysis and significant difference was assumed if p<0.05.

^aBilateral studies.

^bMild to moderate luminal stenosis.

^cSevere luminal stenosis.

^d233 Plaques.

^e201 Plaques.

2.3. Association between plaque hemorrhage (PH) and plaque progression

As stated earlier, PH is known to promote plaque progression. In an 18-month follow-up study of asymptomatic patients with 50–70% luminal stenosis, percentage change in wall volume and lipid-rich necrotic core volume was significantly higher in the group of patients with PH (Takaya et al., 2005). Those with PH at baseline were also more likely to have new PH at 18 months compared to controls (43% vs. 0%; p=0.006). This finding was consistent with that of Spagnoli et al. (2004) who had observed fresh thrombus in carotid plaques from CEA samples of patients, up to 24 months after an ischemic stroke. In a long-term (54 months) follow-up study, the development of PH was found to have a marked effect on plaque progression (Sun et al., 2012). A multicentre study with 6-month follow-up corroborated these results (Sun et al., 2013). Although a few studies have failed to show association between PH and plaque progression (Kwee et al., 2012, Kwee et al., 2010), they did surprisingly reported an association between PH and recurrent cerebrovascular events (Kwee et al., 2013).

2.4. Association between plaque hemorrhage (PH) and subsequent ischemic events

Atherosclerotic disease progresses over decades (slow progression) but can, under certain circumstances, lead to acute symptomatic events (rapid progression) (Fuster, 1994, Kiechl and Willeit, 1999). The presence of PH can be a contributing factor (Lusby et al., 1982a, Lusby et al., 1982b). In a longitudinal MRI study involving 154 asymptomatic patients with 50–79% luminal stenosis, PH was observed to increase the risk of subsequent ischemic cerebrovascular events by five times during a mean follow-up period of 38.2 months (Takaya et al., 2006). The authors highlighted that the ability to distinguish LRNC with PH plaque from the one without PH was critical in identifying higher risk patients. In a complementary study, six cerebrovascular events occurred in 36 asymptomatic patients (50%–70% luminal stenosis) with PH, whilst no clinical events occurred in the 39 patients without PH during a mean follow-up of 24.9 months (Singh et al., 2009). Other investigators have found a similar association between PH and subsequent ischemic events (Altaf et al., 2007, Kurosaki et al., 2011), but not always (Sun et al., 2012). Recently, Hosseini et al. (2013) reported 63.7% of hemorrhage prevalence in 179 symptomatic patients with ≥50% luminal stenosis and 92% (57 out 62) patients who showed PH suffered recurrent ipsilateral events. The risk of recurrent clinical events in symptomatic patients with PH and mild to moderate (30–69%) luminal stenosis has been tabulated in Table 3.

Table 3.

The risk nature of plaque haemorrhage in symptomatic patients with mild to moderate luminal stenosis (30–69%).

Reference (year)	Patients (n)	Hazard ratio (HR)	95% Confident interval (CI)	p-Value	Follow-up period	Description
Altaf et al. (2008)	64	9.8	1.3-75	0.03	28 [26–30] months	13 recurrent events occurred in 39 arteries with IPH.
Sadat et al. (2010c)	61	5.85	1.27-26.77	0.02	514 [255–718] days	–
Teng et al. (2011)	21	–	–	–	262 [10–610] days	11 patient with juxtaluminal haemorrhage/thrombus experienced recurrent events.
Kurosaki et al. (2011)	34	–	–	–	Mean 9.1 months	5 patients experienced recurrent ischemic events.
Yoshida et al. (2012)	25	–	–	–	31.3±16.4 months	11 patients developed a total of 30 recurrent ischemic events.
Kwee et al. (2013)	126	3.542	1.058-11856	0.04	One year	-

3. Imaging modalities for assessing focal carotid plaque hemorrhage (PH)

3.1. Ultrasonography

Ultrasound imaging allows qualitative and quantitative analysis of carotid plaque. It has the ability to identify the location, internal characteristics, and surface detail of plaque adequately. Ultrasonography identifies two plaque categories based on the heterogeneous and homogeneous echo patterns of the lesion studied (Reilly et al., 1983). However, at present it is unable to give detailed morphometric information regarding plaque components such as FC thickness and PH age, known important features of plaque vulnerability. Furthermore, ultrasound is unable to provide sufficient information about the underlying pathological processes involved in atherosclerosis, including inflammatory neovascularization. Ultrasonic contrast agents have been introduced to improve image resolution and specificity (Demos et al., 1999, Lindner et al., 2000). They have shown promising results in various feasibility studies in identifying the inflammatory neovessels in carotid atheroma (Coli et al., 2008, Vicenzini et al., 2007), but this methodology is still in its infancy and will require significant further development before translation into clinical practice.

3.2. Computerized tomography (CT)

With the recent advent of multi-slice computerized tomography (CT) technology, high resolution imaging in the arterial phase of arteries from the aortic arch to the skull base can be rapidly acquired. From the axial source images, image post-processing can be performed to produce an angiographic display, identifying plaque calcium, luminal surface ulceration (Saba et al., 2007, U-King-Im et al., 2010) and intraluminal thrombus (Eesa et al., 2010). However, limitations of CT and CT angiography (CTA) include exposure to ionizing radiation in addition to artefacts induced by calcification and implanted metallic devices. Moreover, CT angiography cannot depict plaque soft components (Walker et al., 2002), including PH, due to the wide distribution of CT density (U-King-Im et al., 2010). With improved technology such as multi-slice detector-rows or dual energy CT (DECT) machines, the accuracy of CTA in depicting soft atherosclerotic content may be further increased. DECT has been reported to improve the ex vivo differentiation between soft tissues (Zachrisson et al., 2010). Results from 31 patients indicated that multidetector CT showed a high level of sensitivity and a moderate level of specificity in detecting atherosclerotic carotid plaques complicated with hemorrhage (Ajduk et al., 2009). It was found that the density difference (∆HU) between early and delayed phases was associated with tissue type in carotid plaques. Histology confirmed that ∆HU was positively correlated with fibrous tissue and negatively correlated with a lipid-rich necrotic core with hemorrhage (Horie et al., 2012). However, further studies are needed for confirmation of these promising results.

3.3. Magnetic resonance imaging (MRI)

High resolution, multi-contrast MRI has shown a great capacity in depicting various atherosclerotic components within carotid plaques (Gortler et al., 1995, Toussaint et al., 1996, Underhill et al., 2010). Detection of plaque hemorrhage by carotid MRI relies on the degradation of hemorrhage into magnetic susceptible hemosiderin and ferritin, two iron-storing proteins present in macrophages, which shorten the longitudinal relaxation time (T₁) of surrounding protons. The consequent imaging effect in carotid MR imaging is a bright signal on T₁-weighted sequences such as time-of-flight (TOF) and standard T₁-weighted fast spin echo (FSE) sequences (Fig. 3A), with low signal intensity in T₂-weighted sequences. These sequences have been extensively validated with histology (Chu et al., 2004, Kampschulte et al., 2004, Toussaint et al., 1996, Yuan et al., 2001). TOF maximum intensity projection (MIP) images have high overall specificity [100.0%, 95% confident interval (CI): 93.0–100.0%], but relatively low sensitivity (32%, 95% CI: 20.8–47.9%) for the identification of PH (Yamada et al., 2012). Contrast enhanced-MR angiography has been suggested as a method to improve the sensitivity in PH detection, when compared to TOF sequence (Qiao et al., 2011).

Fig. 3.

A large plaque hemorrhage depicted by MR imaging from a 72-year old asymptomatic patient (A) and (B) 3D-CUBE T₁-weighted MR image in coronal and transverse planes; (C) and (D) direct thrombus image (DTI) at the same location in (A) in coronal and transverse planes; red asterisk: arterial lumen. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

PH identified on MRI can also be described according to location; intraplaque when the FC is intact and juxtaluminal when there is luminal extension to the thrombus. TOF pulse sequence is particularly helpful in this description (Kampschulte et al., 2004). Juxtaluminal hemorrhage is characterized by discontinuity of dark band in TOF (indicating FC rupture) and high signal intensity in T₁-weighted image, while intraplaque hemorrhage has a smooth luminal contour (Kampschulte et al., 2004, Teng et al., 2011). PH can also be classified according to time of formation, either as fresh/acute (<1 week), recent (1–6 weeks) or remote/old (>6 weeks) (Lusby et al., 1982a). Fresh hemorrhage or thrombus has a shorter T₂ (Toussaint et al., 1996) appearing hypointense in T₂-weighted images. Following PH organization, the intensity of the susceptibility effect is diminished, appearing hyperintense on T₂-weighted images. Old (or remote) hemorrhage appears hypointense in all TOF, T₁- and T₂-weighted images (Chu et al., 2004).

As the pathological and clinical importance of PH has become established, efforts have been made on developing more efficient MR sequences for PH identification. Direct thrombus imaging (DTI), a T₁-weighted MR sequence, has been widely used to depict PH by suppressing the fat signal normally seen on T₁-weighted sequences (Fig. 3B). It does not however provide information about the location or age of PH (Chu et al., 2004). A 3-dimentional (3D) T₁-weighted magnetization prepared rapid acquisition gradient echo (MPRAGE) sequence (Mugler and Brookeman, 1990) has been introduced to detect PH with a bigger contrast-to-noise ratio, improving both sensitivity and specificity (Ota et al., 2010). An optimized 3D spoiled gradient recalled echo pulse Sequence for Hemorrhage assessment using INversion recovery and multiple Echoes (3D SHINE) has been developed to differentiate between fresh and recent PH (Zhu et al., 2010). To further improve intraplaque hemorrhage to wall contrast with suppression of blood signal, a slab-selective phase-sensitive inversion-recovery (SPI) technique is proposed (Wang et al., 2010). A simultaneous non-contrast angiography and PH (SNAP) MR imaging technique was also proposed recently to image both lumen size and carotid PH in a single scan (Wang et al., 2013). Nonetheless, MR offers the most detailed assessment of PH in humans, in terms of its location within a plaque and its age, whilst also offering ability to have a detailed concomitant assessment of other high-risk plaque features.

4. Local biomechanical stresses and carotid plaque hemorrhage

Since carotid plaques are located at the location of the carotid artery bifurcation, the effect of structural biomechanical stresses cannot be overlooked. The biomechanical forces acting on a plaque, which is weakened structurally by the underlying inflammation driven by agents like PH (Altaf et al., 2013), can trigger plaque rupture due to excess loading on the thin FC. Although this remains a hypothesis, circumstantial evidence supports this mechanism. Due to the close relationship between plaques and local haemodynamics, the integration of plaque morphology and mechanical analysis has the potential to improve plaque vulnerability assessment.

For a successful mechanical analysis, knowledge of several important variables is required including governing equations, material properties, geometrical determinants, loading and boundary conditions. In particular, the material behavior of various plaque components, such as FC, lipid and PH are necessary for the accuracy. Moreover, other pre-processing procedures such as geometric reconstruction, the shrinkage procedure and mesh generation are involved. As these components of the simulation are essential they shall be discussed in detail.

4.1. General aspects in biomechanical analysis with carotid plaques

Plaque is a multi-component structure with irregular geometries. The material properties of each component are highly non-linear. The entire plaque undergoes a large deformation due to blood pressure and flow, while the governing equations describing the motion are non-linear. These properties make the mechanical analysis complex, particularly when both plaque structure and blood flow are considered in a computational model [i.e., fluid–structure interaction model (FSI)].

The mechanical properties of atherosclerotic tissues are usually from ex vivo direct material measurements. However this approach is short of clinical potential. Elastography (de Korte et al., 2002, Hu et al., 2011, Shi et al., 2008, Zhang et al., 2010a, Zhang et al., 2010b) has been developed to quantify the material properties of plaque using in vivo data. Some numerical techniques have been also developed for similar purposes (Liu et al., 2012, Schulze-Bauer and Holzapfel, 2003). An accurate measurement of geometric determinants, including FC thickness, LRNC size and lumen curvature, is essential for an accurate numerical prediction. Accuracy is further determined by the imaging resolution, which clearly depends on the imaging technique chosen. For MR carotid imaging, the resolution is determined by the magnetic strength, number of coil element arrays and imaging parameters. The resolution of traditional 2D turbo sequences usually is 0.39×0.39×2(or 3)mm³ (Fuster et al., 2005, Trivedi et al., 2004) with an anatomical coverage of 15–30 mm. It can be improved to about 0.5×0.5×1 mm³ with much larger coverage up to 100–150 mm using newly developed 3D isotropic sequences (Balu et al., 2011, Balu et al., 2009, Li et al., 2010). Such improvement is beneficial for a more accurate geometric reconstruction. Loading conditions include arterial blood pressure and velocity. Non-invasive central blood pressure measurements (carotid artery and thoracic aorta) have been developed and widely validated for this purpose (Papaioannou et al., 2009, Siebenhofer et al., 1999). Blood velocity can be quantified using phase contrast MR imaging (Gatehouse et al., 2005, Long et al., 2000). Residual stress exists in the arterial wall and atherosclerotic structure, but this cannot be accounted for using imaging alone. The influence of residual stress on the carotid bifurcation (Delfino et al., 1997) and atherosclerotic structure (Ohayon et al., 2007) has been previously quantified.

The central issue is in the modeling of the plaque. Different strategies have been adopted to predict stress within the plaque structure. The most comprehensive could be the full coupled 3D FSI simulation (Bluestein et al., 2008, Leach et al., 2010, Tang et al., 2009b). Considering the difficulty of convergence, one-way 3D FSI (Gao et al., 2009) and 2D FSI (Kock et al., 2008, Thrysoe et al., 2010) have been used as alternatives. 3D (Kiousis et al., 2009, Teng et al., 2011) and 2D (Sadat et al., 2011a, Tang et al., 2009a) structure-only modeling have also been used for population based analyses. Tang et al. (2009a) showed a similar stress distribution pattern between 3D full coupled FSI and 2D structure-only analysis using data from one patient. How much the results of each model differ from another awaits comprehensive comparison.

4.2. Material properties of atherosclerotic tissue

4.2.1. Derived from ex vivo direct measurements

The direct material measurement with human atherosclerotic tissues is very limited; readers are referred to summaries on the direct measurements with atherosclerotic tissue in the following references (Chai et al., 2013, Sadat et al., 2010a). In recent years, a few ex vivo direct measurements within carotid tissues have been reported. Fourier transform infrared (FTIR) spectroscopy was used to quantify the amount of mineral and lipid in each tissue region then tested (n=10) (Ebenstein et al., 2009). The results demonstrated that the stiffness of plaque tissue increases with increasing mineral content. Maher et al. (2009) tested fresh CEA samples (n=14). Their results indicated that calcified plaques had the stiffest response, while echolucent plaques were the least stiff. Holzapfel's group has performed direction and layer-specific studies in both normal and atherosclerotic carotid tissues on the residual strain and material properties (Sommer et al., 2010, Tong et al., 2011). By testing the whole plaque sample (n=14) in its entirety, it was found that experimental Green strains at rupture varied from 0.299 to 0.588 and the Cauchy stress observed in the experiments was between 0.131 and 0.779 MPa (Lawlor et al., 2011). Biaxial test has also been employed to characterize the anisotropic behavior of diseased carotid arteries (n=5) using planar biaxial testing (Kural et al., 2012).

Some studies treat plaque tissue as a linear elastic material (Cheng et al., 1993, Loree et al., 1994), in particular those based on in vivo measurements (see next section). Most material models assume each type of tissue, including FC and healthy media, as non-linear, hyper-elastic, homogeneous and incompressible. Various strain energy density functions (SEDF) have been used to describe the material behavior of carotid atherosclerotic tissues, treating the tissue to be either isotropic or anisotropic. The following SEDFs can be found in literature; general polynomial hyperelastic strain–energy function (Maher et al., 2009), single-term Ogden model (Barrett et al., 2009), neo-Hookean material (Chai et al., 2013, Holzapfel, 2000), Yeoh SEDF (Lawlor et al., 2011), Fung, Choi-Vito and modified Mooney–Rivlin SEDFs (Kural et al., 2012). As a fiber-oriented structure, considering the anisotropic behavior of atherosclerotic plaque it is essential for an accurate stress and strain prediction (Yang et al., 2009). However, current in vivo imaging techniques are unable to depict the fiber orientation within the plaque. With this in mind, isotropic material models may be more practical for clinically oriented studies.

The question therefore arises as to whether the stress prediction depends primarily on the constitutive law chosen or material constants involved. Fortunately, it seems that stress analysis can be used with confidence, as material properties variability generates relatively small errors in the prediction of wall stresses. Stresses within the arterial wall, fibrous plaque, calcified plaque, and lipid pool have low sensitivities for variation in the elastic modulus and material models. Even a ±50% variation in elastic modulus leads to <10% change in stress at the site of rupture (Williamson et al., 2003). Sensitivity to variations in elastic modulus is comparable between isotropic nonlinear, isotropic nonlinear with residual strains and transversely isotropic linear models (Williamson et al., 2003). This was confirmed when the change in stress and strain values predicted, using parameters from two very different specimens, were relatively minor (within 20%) despite the substantial difference in stiffness and direction of anisotropy (Kural et al., 2012). However, such insensitivity may only stand under a certain circumstance and need further investigations. It was observed that for a thin FC (0.05 mm), the peak stress in a soft intima (Young’s modulus: 33 kPa) was about 50% as intermediate (Young's modulus: 500 kPa) or stiff (Young’s modulus: 1000 kPa) intima (Akyildiz et al., 2011).

4.2.2. Derived from in vivo measurements

Results from ex vivo direct measurements indicate that the constants describing the material properties of atherosclerotic tissues vary largely from study to study and from location to location, indicating the high non-homogeneity of tissues. Despite the reported insensitivity of material properties in stress prediction, a window may exist as stress and strain predictions can vary significantly if the variation of material properties/models exceeds a certain level (Yang et al., 2009). There is therefore still a need to obtain patient and location-specific material properties. Although this is a challenging goal, advancing imaging techniques may make it achievable. In this section material properties obtained with in vivo measurement are reviewed, particularly those obtained using elastography.

Elastography deals with calculation from elastograms and hence, begins with image registration, i.e., identify how each image point corresponding to a single material point of tissue that is moving within different frames. The gradient of the displacement field, computed in the registration step, generates the strain field (Ophir et al., 1991). For vascular application, the pulsatile intraluminal pressure is used as the deformation force. Ophir et al. (1991) first described the concept of strain imaging using ultrasound data, with the periodicity of ultrasound allowing accurate estimation of displacement. Due to the heterogeneous nature of material stiffness within the atherosclerotic plaque, the displacement varies. Ultrasound-based elasticity imaging can portray such differences in the elastic properties of soft tissues. The accuracy of elastography in capturing deformation/strain has been widely assessed using phantom and finite element analysis (Allen et al., 2011, Dumont et al., 2011, Park et al., 2010, Ribbers et al., 2007). Its capacity in quantifying the material properties has been testified by comparing the results obtained in and ex vivo (Kanai et al., 2003, Viola et al., 2004, Xie et al., 2005). The technique has been employed widely to characterize plaque type (hard or soft) based on the displacement amplitude, allowing differentiation of vulnerable and stable lesions (Allen et al., 2011, Dahl et al., 2009, Schmitt et al., 2007).

With the assumption of linear elasticity, the material properties of arterial wall and atherosclerotic tissues in various artery beds have been investigated. In an ex vitro elastography study, average elasticity was 81±40 kPa in the lipid region of 9 human iliac arteries and 1.0±0.63 MPa where there was a mixture of smooth muscle and collagen fiber (Kanai et al., 2003). The difference between the obtained value in vivo and ex vivo was approximately 8%. In the same report, obtained from the in vivo data, the elasticity of healthy human common carotid artery was approximately 1 MPa (n=2) and that of lipid was about 100 kPa (n=2). Young's moduli for the common carotid artery were compared in groups with (n=25) and without (37) plaque (Paini et al., 2007). The value did not appear to significantly differ between groups (751±560 kPa vs. 677±427 kPa; p>0.05). Based on in vivo cine and multi-contrast MR images, Liu et al. (2012) quantified material properties from 12 patients using modified Mooney–Rivlin SEDF. The effective Young's modulus varied from 137 kPa to 1435 kPa within the stretch range of [1.0, 1.3]. The non-linear material behavior of atherosclerotic tissues can also be captured with more sophisticated methods. Karimi et al. (2008) used a combination of non-linear finite element methods with a genetic algorithm to determine the material constants in the exponential form of the Mooney–Rivlin SEDF using optical coherence tomography images.

4.2.3. Material properties of plaque hemorrhage (PH)

Ultrasound-based elasticity imaging has also been used to capture the variation of Young's modulus of venous thrombosis in rats (n=4) during maturation (Aglyamov et al., 2007). The Young's modulus of venous thrombosis in rat, measured by in vivo elastography, increased from 8.53±2.66 kPa at day 3 to 34.73±8.26 kPa at day 10 (n=12) (Xie et al., 2005). The material properties of PH and intraluminal thrombus could be very similar considering their similar biological content. We have recently performed direct uni-axial material tests with PH and intraluminal thrombus, which demonstrate their non-linear material behavior. Modified Mooney–Rivlin SEDF was used to describe its stress–stretch relationship. In total, 12 PH tissue strips were obtained from CEA samples from 5 symptomatic patients. The fitted material constants were: c₁: 16.57×10⁻⁸ kPa [5.27×10^–8, 0.10]; D₁: 3.76 kPa [0.41, 2.67] and D₂: 8.90 [4.50, 12.44] (details of methodology are in the Appendix).

4.3. Shrinking procedure

In vivo plaque imaging, either from ultrasound, CT or MR, is acquired under pressurized conditions. The computational simulation should at least start from a zero pressure configuration due to the large deformation and non-linear material properties if the residual stress was ignored. Several techniques have been employed for this purpose. Inverse design analysis has been used in studies (Gee et al., 2009, Govindjee and Mihalic, 1996, Lu et al., 2007) to determine a truly stress-free configuration. A backward incremental method, which is a modified updated Lagrangian formulation, has also been applied to achieve a computational start shape in studies of abdominal aortic aneurysms (de Putter et al., 2007, Gee et al., 2009, Merkx et al., 2009, Speelman et al., 2009) and atherosclerotic plaques (Speelman et al., 2011). A pull-back algorithm was recently developed to determine the unloaded geometry in aneurysms (Riveros et al., 2013). These approaches lead to an acceptable recovery of vessel geometry with reasonable computational load. The uniform shrinkage method proposed by Huang et al. (2009) is an alternate approach (Tang et al., 2009a). It involves uniform shrinkage of the lumen contour about its geometric center before external loading is applied. The shrinkage rate is verified by comparing the pressurized geometry with the in vivo contour. Being a phenomenon-based technique, this is relatively easy to apply, but can fail to recover the in vivo configuration when the luminal shape is irregular. A non-uniform shrinking procedure has been proposed to handle this problem (Huang et al., 2011). It imposes further stretching in the major axis and compressing in the minor axis after the uniform shrinking. If the shrinking procedure is omitted, the stress prediction will be over-estimated in the plaque with a circular lumen shape (Tang et al., 2009a) and under-estimated in that with an irregular lumen shape (Huang et al., 2011).

4.4. Biomechanical stress analyses of carotid plaques with and without plaque hemorrhage (PH)

The association between mechanical loading and formation of plaque hemorrhage was proposed about 50 years ago. Constantinides, 1967, Constantinides, 1984 suggested that FC microfissures could result from mechanical factors, such as bursts of hypertension or the constant bending and torsion of the arterial wall. The resulting effect could be the formation of an intraluminal thrombus; indeed FC erosion is thought to be responsible for approximately 40% of lethal coronary thromboses (Virmani et al., 2000). About 20 years later, Lusby et al. (1982b) proposed that mechanical/tensile stresses probably played an important role in the sudden transition from the asymptomatic to the symptomatic stage, most likely due to rupture of neovessels producing PH. It was also suggested that higher mechanical stresses could produce PH by stretching the intima and pushing the lesion further into the lumen. Due to limited numerical analysis techniques and computing power in the early 1980s, the tensile stress was approximated using Laplace's law. Since then only a few groups have studied this relationship. Compared to the worldwide interest in histological and imaging-based clinical studies on carotid plaque vulnerability, biomechanical analyses comprise only a fraction of the published research.

The association between neovessel rupture and mechanical loading was recently investigated by quantifying the mechanical conditions around each neovessel (Teng et al., 2012). It was found that neovessels with red blood cells nearby were subjected to a higher stretch level compared with those without red blood cells. High local stretch level may therefore lead to neovessel rupture, resulting in the formation and expansion of PH.

The association between PH and elevated stress at a patient level has also been explored. In a MRI-based study involving 45 symptomatic patients, the stress level over FC in 28 individuals with PH was significantly higher than those of asymptomatic patients (315 kPa [247–434] vs. 200 kPa [171–282]; p=0.003; median [inter quartile range]) (Sadat et al., 2011b). Similarly, Huang et al. (2010) performed 3D FSI on 5 patients showing that stress over FC near PH was higher than those locations without PH (75.6 kPa vs. 68.1 kPa; p=0.0003). The same group further reported that the size of PH had a significant impact on stress level over FC (Huang et al., 2012). The impact of hemorrhage evolution on the stress level over FC has also been explored. When the hemorrhage passes the acute period, stress decreased 30% from 159 kPa [114–253] to 118 kPa [79–189] (p=0.001; n=23) (Sadat et al., 2011c). High stretch concentration around the rupture site in plaques with juxtaluminal hemorrhage (Teng et al., 2011) may prevent the plaque healing by inducing apoptosis of both smooth muscle and endothelial cells (Kainulainen et al., 2002, O’Brien et al., 1998). The peak stretch ratio in lesions with juxtaluminal hemorrhage was 2.10±0.53 (n=21), which was significantly larger than lesions without such features (1.21±0.08; p<0.0001) (Teng et al., 2011). Large stretch may contribute to increasing plaque vulnerability, as 11 (11 out of 21 patients) recurrent events occurred in the first year in the patients with PH.

5. Expert commentary

PH is an important plaque component which significantly increases plaque vulnerability. Combined with other vulnerable plaque features, such as thin FC and large LRNC, it can form an important component towards more accurate patient risk stratification. Clinical studies have demonstrated the strong association between PH and future cerebrovascular events. Although biomechanical studies are few, they appear to support results of clinical studies. Advancements with in vivo imaging techniques, particularly MRI, have helped the investigation into the mechanisms and significance of PH. Determination of in vivo material properties for both PH and other plaque components would significantly assist patient-specific biomechanical simulations, hopefully leading to a refined risk assessment tool. However the complementary value provided by biomechanical analysis to the luminal stenosis and plaque morphological and compositional features in assessing plaque vulnerability needs further investigations. A collaborative effort by a multidisciplinary team of pathologists, radiologist, biomechanical engineers and clinicians is required to achieve this future goal.

Conflict of interest statement

Authors do not have any conflict of interest to be declared.

Appendix A. Supplementary materials

Supplementary data