Computer-assisted Carotid Plaque Analysis: Characteristics of Plaques Associated with Cerebrovascular Symptoms and Cerebral Infarction

Transcription

1 Eur J Vasc Endovasc Surg 19, (2000) doi: /ejvs , available online at on Computer-assisted Carotid Plaque Analysis: Characteristics of Plaques Associated with Cerebrovascular Symptoms and Cerebral Infarction L. M. Pedro, M. M. Pedro, I. Gonçalves, T. F. Carneiro, C. Balsinha, R. Fernandes e Fernandes and J. Fernandes e Fernandes Objective: to correlate the echostructural characteristics of carotid plaques with neurological symptoms and cerebral infarcts. Materials: one hundred and five plaques were studied in 74 patients by colour-flow duplex ultrasound: 39 were symptomatic and 37 were associated with brain infarction on CT scanning. Methods: the images were digitalised for computer-assisted image standardisation and divided in homogenous (n=67) and heterogenous (n=38) groups. Parameters of the plaque image were analysed. Results: global echogenicity was significantly lower in symptomatic and in CT + plaques (p<0.05). For homogenous plaques, an echogenic cap (EC) was visualised in 8.3% of symptomatic vs. 33.9% of asymptomatic (p<0.05) and in 7.7% for plaques that were CT + vs. 37.7% for CT (p<0.05). Surface disruption was visualised in 50% of symptomatic vs. 8.5% of asymptomatic (p=0.002) and in 46% of CT + vs. 9.4% of CT plaques (p=0.002). For heterogenous plaques, the echolucent region was juxtaluminal in 67% of symptomatic and CT + plaques vs. 33% in asymptomatic and CT (p<0.01). Conclusion: echolucent plaques are associated with a higher neurological risk. In homogenous plaques the absence of an echogenic cap and disruption of the plaque surface also correlates with symptoms. In heterogenous plaques, juxtaluminal location of the echolucent region is an additional marker of increased risk. Key Words: Carotid plaque; Computer-assisted plaque analysis; High-risk carotid plaque; High definition ultrasonography. Introduction The structure and composition of atheromatous carotid plaques may influence the neurological risk of emboli. Several aspects of plaque structure have been studied. Based on the observation of surgically removed specimens, a positive correlation between intra-plaque haemorrhage or ulceration and neurological events has been reported. 1 8 Previous multicentric studies have confirmed the importance of the degree of stenosis in the prediction of outcome in symptomatic patients with severe carotid artery disease However, for asymptomatic or less severe forms of carotid stenosis, there is some controversy concerning the risk of neurological events and the possible role of surgery in the prevention of stroke. 11 Histological studies performed in specimens removed during carotid endarterectomy, analysing the morphology, structure and composition of the lesion, have been able to correlate some of these characteristics with appropriate neurological symptoms Please address all correspondence to: L. Mendes Pedro, Rua D. Estefânia Esq., 1000 Lisboa, Portugal. B-mode ultrasonography, particularly with highdefinition imaging capabilities, provides a suitable technique for evaluation of structural aspects of atheromatous plaques as shown in previous publications Limitations of this method are its subjective nature, dependency upon the experience and ability of the observer and also the quality and set-up of the equip- ment. The introduction of high-definition ultra- sonography and the association of computer-assisted methods for carotid-plaque analysis have been shown to provide the possibility of standardisation of dif- ferent image characteristics, thus providing an ob- jective assessment of morphological plaque parameters. 23,24 The aim of the present study was to assess the possibility of recognition of objective parameters of plaque echo-structure by computer-assisted analysis, and to establish which factors may have a significant correlation with appropriate neurological symptoms and evidence of brain infarcts detected by CT scanning. Materials and Methods The study included 105 carotid bifurcation plaques from 74 patients (58 men and 16 women). The mean age /00/ $35.00/ Harcourt Publishers Ltd.

2 Computer-assisted Carotid Plaque Analysis 119 was 67 years (38 80). All patients had a neurological Table 1. Analysis of the characteristics of the plaques. evaluation and the carotid lesion was considered Symptomatic Plaques 39 37% symptomatic when associated with amaurosis fugax or Asymptomatic Plaques 66 63% appropriate neurological deficits (transitory, reversible CT Scan % or established) related to the ipsilateral carotid ter- CT Scan 68 65% ritory. A CT scan was obtained in all patients and Stenosis >70% 54 51% Stenosis 51 69% 21 20% was considered positive when cerebral infarcts were Stenosis 30 50% 30 29% present in the area of the ipsilateral anterior and Homogenous Plaques 67 64% middle cerebral arteries. Lacunar infarcts were not considered relevant for this analysis and, when present as isolated finding, the CT scan was classified as negative. Heterogenous Plaques 38 36% The examination used colour-flow duplex-scan plaques the following characteristics were analysed: equipment (ATL Advanced Technology Laboratories global echogenicity (GSM, P40), the presence of echo- model HDI 3000) which provided high-definition genic cap (EC), evidence of plaque-surface disruption, images of the carotid bifurcation. Equipment set-up relation between EC thickness and plaque thickness. included 7 10 MHz probe, 60 db dynamic range and In the subgroup of heterogenous plaques the following post-processing linear maps. All the ultrasonographic parameters were analysed: global echogenicity (GSM, examinations were performed by the same observer P40), localisation of echolucent region of the plaque (LMP) without knowledge of the clinical and CT-scan (basal or juxtaluminal), and the percentage of echoevaluation. The images were recorded in a S-VHS lucent area in relation to the overall plaque area. videotape and digitalised directly by the duplex-scan For each lesion the degree of stenosis was quantified machine into an optical disk for later computer ana- using end-systolic and end-diastolic velocity criteria 26 lysis. Images were obtained in longitudinal and trans- combined with cross-section area reduction measured verse section in order to obtain full visualisation of in transverse images of the lesion. Statistical analysis the carotid bifurcation and extent of the lesion. How- used the software STATA 4.0 with categorical variables ever, for analysis, the best longitudinal section with analysed by Chi-squared and Fisher s exact test and full display of the plaque and adventitia was used and continuous variables analysed by variance analysis this image was recorded in black and white. The computer-assisted analysis used the software and Student s test. ADOBE PHOTOSHOP 3.0 for standardisation and echogenicity analysis and the software INSIGHT 5.0 for linear and area measurements. This software provides the possibility of statistical analysis of the histogram of grey-scale pixel distribution of the image. Results Standardisation was achieved as previously reported. 25 A normalised value of 190 for the adventitia and 0 for Thirty-nine (37%) plaques were considered sympthe blood was set, thus producing a modification of tomatic and 66 (63%) were asymptomatic. Thirty-seven the characteristics of the entire image, which was (35%) plaques were associated with ipsilateral cerebral adjusted according to these values, equal for all ex- infarction and in 68 (65%) the CT scan was negative. aminations. In this standardised image the plaque was The mean degree of stenosis was 68%, and the severity delineated by drawing around its entire contour, and of the stenosis was assessed for each lesion and its a histogram of the grey-scale distribution of pixels of distribution was as follows: more than 70% stenosis the entire area of the plaque was obtained. The fol- 54 lesions (51%); stenosis between 51 69% 21 lesions lowing parameters were determined from the histo- (20%); stenosis between 30 50% 30 lesions (29%). gram and used in the study: (i) the grey-scale median Sixty-seven plaques (64%) were classified as homo- (GSM); (ii) the percentile 40 (P40), which represents genous and 38 lesions (36%) were considered heterothe percentage of black pixels ( echolucent pixels). genous (Table 1). These two parameters were considered as a measure The analysis of echogenicity of the overall group of of the whole echogenicity of the plaque and analysed plaques showed that the GSM of symptomatic plaques in all the plaques. was significantly lower than of asymptomatic plaques Carotid plaques were classified in homogenous and heterogenous according to its echo-structure, thus defining two subgroups. In the subgroup of homogenous (respectively 33 vs. 46, p=0.005). A similar relationship was observed in plaques associated with normal or abnormal CT scan (respectively 32 vs. 44, p=0.005)

3 120 L. M. Pedro et al. Fig. 1. Distribution of the grey-scale median (GSM) in symptomatic and asymptomatic plaques and in plaques associated with positive and negative CT scan (for ipsilateral brain infarction). Fig. 3. Distribution of the grey-scale median (GSM) and percentile 40 (P40) according to the degree of stenosis. significant difference was observed in symptomatic lesions when compared with asymptomatic plaques. In those that were considered symptomatic, echographic evidence of echogenic cap was present only in 8.3%, and in asymptomatic plaques it was clearly visible in 33.9% (p<0.05). The presence of an echogenic cap was observed in 7.7% of the lesions in the group with positive CT scans and 37.7% in those with negative CT scans (p<0.05). Evidence of plaque surface dis- ruption was diagnosed in 50% of the symptomatic plaques and in 46% of those associated with cerebral infarction, but only in 8.5% of asymptomatic (p=0.002) and 9.4% of those with a negative CT scan (p=0.002). The relationship between echogenic cap thickness and overall plaque thickness was also studied in the group of homogenous plaques, and the results showed similar values in symptomatic and asymptomatic plaques (respectively 13.9 and 11.9, p=ns) and in plaques with and without cerebral infarction (respectively 15.7 and 12.1, p=ns). In heterogenous plaques, the GSM was also significantly different concerning the surrogate endpoints: presence of symptoms and brain infarcts on CT scan evaluation. The P40 parameter was significantly different according to the presence of brain infarct, but was of no significant value in the discrimination between symptomatic vs. asympto- matic plaques; the observed trend in the distribution of its values did not reach statistical significance. In symptomatic plaques and in those associated with brain infarction there was evidence of a juxtaluminal location of the echolucent region in 67% against 33% in asymptomatic plaques and plaques not associated with cerebral infarction (p<0.01). The percentage of echolucent region relatively to the overall plaque area did not show significant differences between symp- tomatic and asymptomatic plaques (respectively 63% and 60%, p=ns) and between plaques associated with positive and negative CT scan (respectively 69% and 57%, p=ns). Fig. 2. Distribution of the percentile 40 (P40) in symptomatic and asymptomatic plaques and in plaques associated with positive and negative CT scan (for ipsilateral brain infarction). (Fig. 1). The parameter P40 was also significant in the discrimination of symptomatic vs. asymptomatic plaques (respectively 60 vs. 47, p=0.001), and for plaques associated with normal or abnormal CT scan (respectively 62 vs. 49, p=0.002) (Fig. 2). The search for a cut-off point revealed that the best discrimination level was a value of 32 for the GSM and of 43 for the P40. Sixty per cent of plaques with GSM below 32 were symptomatic (against 26% in the group with GSM>32) and 66% were associated with brain in- farction (against 25% in the group with GSM>32). Fifty- one per cent of plaques with P40>43 were symptomatic (against 14% in the group with P40<43) and 47% were associated with brain infarction (against 24% in the group with P40<43). The echogenicity of the plaques was also studied according the degree of stenosis, as assessed by colour-flow duplex scan. Lesions were classified in three groups: <50%, 51 69% and >70%, and the GSM was, respectively, 57, 45 and 33 (p<0.001) and P40 was, respectively, 37, 48 and 61 (p<0.001) (Fig. 3). Subsequently, analysis was performed separately in the two subgroups of homogenous and heterogenous plaques (Table 2), in order to establish a correlation between the several parameters previously analysed and the following surrogate endpoints: presence of symptoms and evidence of brain infarcts on CT scan. In the subgroup of homogenous plaques, a statistical

4 Computer-assisted Carotid Plaque Analysis 121 Table 2. Results of the separate analysis of the subgroups of homogenous and heterogenous plaques. Symptomatic Asymptomatic p CT + CT p Homogenous GGM p= p=0.005 P p= p=0.002 Presence of echogenic cap 8.3% 33.9% p< % 37.7% p<0.05 Surface disruption 50% 8.5% p= % 9.4% p=0.002 EC thickness/plaque thickness NS NS Heterogenous GSM p< p<0.01 P NS p=0.01 Echolucent region juxtaluminal 67% 33% p< % 33% p<0.01 % echolucent region 63% 60% NS 69% 57% NS Discussion dependency upon the operator performing the test, and the quality of the equipment. The introduction The relationship between carotid artery disease and of high-definition equipment with improvements in stroke was by the preliminary work of Egas Moniz 27 image quality, and the association of computerised (1937), with the use of arteriography, and Miller methods to standardise the assessment of carotid Fisher 28 (1951) who established a definitive correlation plaques, were remarkable developments that allowed between carotid stenosis, transient ischaemic attacks us to overcome limitations due to individual and and stroke. 29 Structural alterations within the atheromatous technical variability, thus facilitating comparisons beand carotid plaque may predispose to instability tween different examinations and the monitorisation facilitate the occurrence of embolisation to the of the progression of carotid stenosis. Previous pub- brain. Histological and biochemical studies from speci- lications confirmed the feasibility of objective quantification mens obtained from carotid endarterectomy helped to of heterogeneity and echolucency, and also clarify the relationship between some features of the that these parameters had a correlation with symptoms lesion and the predisposition to cause symptoms. The and ipsilateral brain infarcts. 23,24,34,35 increase of lipid 30,31 and blood 2,6 content in the plaque, We have expanded the non-invasive investigation to the proximity of a central necrotic core from the overlying accommodate other components of the atheromatous fibrous cap 16 and the presence of ulceration 3,7,13 plaque that could be measured with this combined were shown to be more common in symptomatic technology and that were previously shown, in histo- lesions, when compared with specimens from logical studies, to be associated with unstable, symptomatic asymptomatic stenosis. plaques. The aim of this study was to add B-mode ultrasonography provides a reliable non- refinements to the discrimination between different invasive method for evaluation of the morphological lesions in order to improve the identification of those aspects of the atheromatous plaque. Studies using this that would have a higher risk of neurological complications, technology helped to classify plaques in different types thus obtaining a comprehensive pattern of according to its ultrasonographic appearance. One of unstable plaques which could be associated with the these classifications divides plaques in homogenous surrogate events: appropriate neurological symptoms and heterogenous. 17 Gray-Wheale, 32 Stephens 33 and and/or ipsilateral brain hemispheric infarcts. Our Geroulakos 22 developed a more complete classification, results showed that the overall echogenicity assessed based on ultrasonographic plaque structure and echo- from the plaque grey-scale pixel distribution is an genicity, which includes both homogenous and important discriminating factor in both homogenous heterogenous plaques. Echolucency and heterogeneity and heterogenous lesions, with symptomatic plaques seem to be associated with plaque instability and being more echolucent. The definition of a cut-off level with predisposition to respond differently to a similar of echogenicity that provides the higher discrimination haemodynamic alteration caused by the jet effect of capacity is important and should be mentioned in increased velocity and turbulence caused by a severe future studies. El-Barghouty 34 reported a cut-off point stenosis. Atheroembolism could be more frequent in of 32 for the GSM in a study of computer-assisted this type of plaque than in those with a fibrotic, dense plaque analysis without standardisation of the images. and homogenous structure. Interestingly, we observed the same cut-off level of 32 The problem related with the systematic use of the for the GSM and the discrimination capacity of GSM Gray Wheale classification is the lack of uniformity, and P40 was not significantly different, thus suggesting

5 122 L. M. Pedro et al. the need to pursue research of better discrimination 5Fisher CM, Ojemann RG. A clinico-pathologic study of carotid endarterectomy plaques. Rev Neurol 1986; 142: levels and parameters in future studies. 6 AbuRhama AF, Boland JP, Robinson P et al. Antiplatelet therapy In homogenous plaques the absence of an iden- and carotid plaque hemorrhage and its clinical implications. J tifiable echogenic cap, an ultrasonographic equivalent Cardiovasc Surg 1990; 31: Carr S, Farb A, Pearce WH et al. Atherosclerotic plaque rupture of the fibrous cap, and evidence of plaque surface in symptomatic carotid artery stenosis. J Vasc Surg 1996; 23: disruption are significantly more common in plaques associated with the surrogate endpoints: symptoms 8Park AE, McCarthy WJ, Pearce WH et al. Carotid plaque morphology correlates with presenting symptomatology. J Vasc and brain infarction. The ultrasonographic detection Surg 1998; 27: of disruption of plaque surface with erosion or ul- 9European Carotid Surgery Trialists Collaborative Group. ceration is a matter of controversy. Some authors report MRC European Carotid Surgery Trial. Interim results for symp- high levels of accuracy, but others fail to reproduce tomatic patients with severe (70 99%) and with mild (0 29%) stenosis. Lancet 1991; 337: these results. 36,37 Newer HDU equipments providing 10 North American Symptomatic Carotid Endarterectomy better image quality improve the accuracy in the de- Trial Collaborators. Beneficial effect of carotid endarterectomy in symptomatic patients with high grade carotid tection of plaque disruption through a more clear stenosis. NEnglJMed1991; 325: visualisation of plaque surface and fibrous cap. 11 Executive Committee for the Asymptomatic Carotid Bassiouny et al. 16 published a histological study that Atherosclerosis Study. Endarterectomy for asymptomatic carotid artery stenosis. JAMA 1995; 273: looked at the structure of the plaque and the spacial relationship between the various components of the 12 Imparato AM, Riles TS, Gorstein F. The carotid bifurcation plaque: pathologic findings associated with cerebral ischemia. lesion. They concluded that, in symptomatic plaques, Stroke 1979; 10: proximity of the central core was twice that in asymptoplaques: correlation of clinical and anatomical findings. Ann Vasc 13 Avril G, Batt M, Guidoin R et al. Carotid endarterectomy matic plaques, but found no significant relevance of Surg 1991; 5: the percentage of central core on symptoms associated 14 Feeley TM, Leen EJ, Colgan MP et al. Histologic characteristics with the plaque. Our study used non-invasive tech- of carotid artery plaque. J Vasc Surg 1991; 13: nology to visualise these structural components of 15 Carr SC, Cheanvechai MD, Virmani RV et al. Histology and clinical significance of the carotid atherosclerotic plaque: imthe lesions and confirmed their findings, with the plications for endovascular treatment. J Endovasc Surg 1997; 4: observation that in heterogenous plaques it is the juxtaluminal location of the echolucent region, rather 16 Bassiouny HS, Sakaguchi Y, Mikucki SA et al. Juxtaluminal than its percentage, that was significantly correlated location of plaque necrosis and neoformation in symptomatic carotid stenosis. J Vasc Surg 1997; 26: with the clinical endpoints of the study. 17 Reilly LM, Lusby RJ, Hughes L. Carotid plaque histology using In conclusion, overall echolucency in both homo- real-time ultrasonography. Am J Surg 1983; 146: genous and heterogenous plaques with GSM<32 and 18 Ratliff DA, Gallagher PJ, Hames TK et al. Characterization of carotid artery disease: comparison of duplex scan with histology. P40>43, disruption of plaque surface and absence of Ultrasound Med Biol 1985; 11: echogenic cap for homogenous plaques, and juxta- 19 O Donnell TF, Erdoes L, Mackey WC et al. Correlation of B- luminal location of the echolucent area for heterofindings at carotid endarterectomy. Arch Surg 1985; 120: mode ultrasound imaging and arteriography with pathologic genous plaques provide a set of objective parameters 20 Reilly LM. Carotid intraplaque hemorrhage: Non-invasive dethat help to define a profile of plaque instability. These tection and clinical significance. In: Bernstein EF, ed. Non-Invasive parameters correlate with the presence of appropriate Diagnostic Techniques in Vascular Surgery. The C.V. Mosby Com- neurological symptoms and brain infarction in the pany, 1990: Merritt CR, Bluth EI. Ultrasonic identification of plaque comipsilateral carotid territory, and provide more objective position. In: Labs KH, ed. Diagnostic Vascular Ultrasound. Arnold, identification of high-risk carotid stenosis. London, 1992: Geroulakos G, Ramaswami G, Nicolaides AN et al. Characterization of symptomatic and asymptomatic carotid plaques using high-resolution real-time ultrasonography. Br J Surg 1993; 80: References 23 El-Barghouty, Geroulakos G, Nicolaides AN et al. Computerassisted carotid plaque characterization. Eur J Vasc Endovasc Surg 1Lusby RJ, Ferrell LD, Ehrenfeld WK et al. Carotid plaque 1995; 9: hemorrhage: its role in production of cerebral ischemia. Arch 24 El-Barghouty, Nicolaides NA, Bahal V et al. The identification Surg 1982; 117: of the high carotid plaque. Eur J Vasc Endovasc Surg 1996; 11: 2Imparato AM, Riles TS, Mintzer R et al. The importance of hemorrhage in the relationship between gross morphologic 25 Elatrozy T, Nicolaides A, Tegos Th et al. The effect of B- characteristics and cerebral symptoms in 376 carotid artery mode ultrasonic image standardisation on the echodensity of plaques. Ann Surg 1983; 197: symptomatic and asymptomatic carotid bifurcation plaques. Int 3Zukowski AJ, Nicolaides AN, Lewis RT et al. The correlation Angiol 1998; 17: between carotid plaque ulceration and cerebral infarction on CT 26 Zweibel WJ. Spectrum analysis in carotid sonography. Ultrasound scan. J Vasc Surg 1984; 1: Med Biol 1987; 13: Persson AV. Intraplaque hemorrhage. Surg Clin North Am 1988; 27 Moniz E, Lima A, Lacerda R. Hémiplégies par thrombose de 66: la carotide interne. Presse Med 1937; 45:

6 Computer-assisted Carotid Plaque Analysis Miller Fisher C. Occlusion of the internal carotid artery. Arch ultrasound appearance in symptomatic and asymptomatic vessels. Neurol Psychiatry 1951; 5: Aust NZ J Surg 1989; 59: Estol CJ. Dr C Miller Fisher and the history of carotid disease. 34 El-Barghouty, Nicolaides AN, Tegos T et al. The relative effect Stroke 1996; 27: of carotid plaque heterogeneity and echogenicity on ipsilateral 30 Gronholdt ML, Nordestgaard BG, Wiebe BM. Echolucency cerebral infarction and symptoms of cerebrovascular disease. Int of computerized ultrasound images of carotid atherosclerotic Angiol 1996; 15: plaques are associated with increased levels of triglyceride-rich 35 Biasi GM, Mingazzini P, Baronio L et al. Carotid plaque lipoproteins as well as increased plaque lipid content. Circulation characterization using digital image processing and its potential 1998; 97: in future studies of carotid endarterectomy and angioplasty. J 31 Gronholdt ML, Wiebe BM, Laursen H et al. Lipid-rich artery Endovasc Surg 1998; 5: plaques appear echolucent on ultrasound B-mode images and 36 Sitzer M, Muller W, Rademacher J et al. Colour-flow Doppler- may be associated with intraplaque haemorrhage. Eur J Vasc assisted duplex imaging fails to detect ulceration in high-grade Endovasc Surg 1997; 14: carotid artery stenosis. J Vasc Surg 1996; 23: Gray-Weale AC, Graham JC, Burnett JR et al. Carotid athersonographic 37 Kardoulas DG, Katsamouris AN, Gallis A et al. Ultraoma Comparison of B-mode ultrasound appearance with and histologic characteristics of symptom-free and carotid endarterectomy pathology. J Cardiovasc Surg 1988; 29: symptomatic carotid plaque. Cardiovasc Surg 1996; 4: Stephen CM, Gray-Weal AC, Byrne KE et al. Carotid atheroma: Accepted 17 May 1999