Three-dimensional ultrasound evaluation of the effects of pomegranate therapy on carotid plaque texture using locality preserving projection
Introduction
Stroke is the second leading cause of death worldwide, with 5.5 million stroke-related deaths in 2016 [1], [2]. China, accounting for one-fifth of the population, has a disproportionate share of stroke mortality, which stood at 1.8 million in 2016 and accounted for one-third of the total number of deaths from stroke worldwide [1], [2]. Carotid atherosclerosis is a major source of atherosclerotic emboli (platelet aggregates and plaque debris) that would block cerebral arteries, leading to ischemic strokes. Fortunately, for patients with high stroke risk, 75–80% of stroke can be prevented by lifestyle/dietary changes and medical therapies [3]. With recent advances in the pathogenesis of atherosclerosis, the numbers of new interventions are expected to rise rapidly. Effects of these treatment or management strategies are required to be thoroughly validated in clinical trials. Therefore, in parallel to the development of novel therapies, there is a critical requirement of measurement tools for cost-effective serial monitoring of carotid atherosclerosis.
Carotid intima-media thickness (IMT) measured from 2D B-mode ultrasound was an early imaging biomarker introduced to increase the sensitivity of carotid disease monitoring, but recent investigations suggest two major weaknesses of IMT. First, the annual change of IMT is small ( ~ 0.15 mm, below the spatial resolution of carotid ultrasound) and does not permit the measurement of change in individuals within a clinically affordable timeframe [4]. Second, IMT measures vascular wall thickening, which is not directly related to atherosclerosis [5], rendering it a weak predictor of cardiovascular risk [6], [7]. Although total plaque area (TPA) measurement has emerged as a more accurate metric for stroke risk prediction [8], both IMT and TPA are measured from 2D ultrasound images, acquisition of which requires an operator to choose an imaging plane to be scanned, making an image difficult to reproduce, even for the same operator. Therefore, 2D ultrasound measurements are sub-optimal in serial monitoring of plaque [9].
Development of 3D ultrasound has allowed for more sensitive and reproducible quantification of carotid atherosclerosis. Total plaque volume (TPV) measured from 3D ultrasound has been shown to be reproducible [10], able to predict vascular events [11] and sensitive to disease progression and regression in a clinical trial involving high-dose atorvastatin [12]. The risk of cardiovascular events is related to the presence of vulnerable plaques and the vulnerability of a plaque is determined by its composition. Changes in plaque textural features have served as a surrogate for plaque composition assessment. Textural analysis was shown to be more accurate than TPV in showing plaque regression with high-dose atorvastatin versus placebo subjects [13]. Van Engelen et al. [14] showed that textural change was more accurate than TPV change in predicting cardiovascular events. Zhou et al. [15] showed that fractal dimension texture analysis was more able than TPV to discriminate subjects receiving atorvastatin and placebo. A number of investigations performed in 2D ultrasound have also shown that textural features were useful in identifying symptomatic patients [16], [17]. Roy-Cardinal et al. [18] showed that quantitative tissue information extracted from ultrasound imaging was useful in classifying carotid artery plaque components, including lipid, calcification and ruptured fibrous cap. These results suggest that a biomarker based on plaque textural features would be more able than TPV to highlight treatment effects. The increased discriminative power conferred by such a texture-based biomarker could substantially reduce the sample size, duration and therefore cost required to establish the efficacy of novel dietary/lifestyle interventions.
Many textural features have been extracted from carotid ultrasound images in investigations involving the stratification of stroke risk [16], [19], [20]. Christodoulou et al. [16] extracted 61 texture features and shape parameters from 2D longitudinal carotid ultrasound images to classify symptomatic statuses using a combination of classifiers. Kyriacou et al. [19] extracted 43 texture features from 2D longitudinal ultrasound images in a stroke risk assessment study. The probabilistic neural network (PNN) and support vector machine (SVM) were used to classify the symptomatic statuses of patients. Araki et al. [20] extracted 16 textural features from 2D longitudinal images and classified the stroke risk into high and low using SVM. The ground truth classification was determined by lumen diameter. Huang et al. [21] extracted 300 texture features in classifying plaques into 3 categories (i.e., hyperechoic, intermediate and anechoic) using the K-nearest neighbour classifier. The ground truth labels were manually assigned by expert observers. These techniques were evaluated in cross-sectional analyses based on 2D carotid ultrasound images, but not in longitudinal studies of plaque change. The increased reproducibility afforded by 3D carotid ultrasound imaging techniques has allowed for 3D textural analysis of plaque change. Awad et al. [13] extracted 270 textural features to detect statin-related changes in carotid atherosclerosis and showed that textural features were more accurate than TPV in classifying patients who received high-dose atorvastatin and placebo. Van Engelen et al. [14] extracted 376 features to predict cardiovascular events in high-risk patients and showed that textural change was more accurate than TPV change in predicting cardiovascular events. The 376 features extracted in this study were generated by 9 textural extraction techniques and encompass most features used in the 2D and 3D studies discussed above.
In the current study, we aim to develop a cost-effective texture-based biomarker capable of quantifying the degree of plaque change. The 2D and 3D textural analyses discussed above primarily involved feature-based classification. The classifiers generated discrete class labels but did not allow the quantification of how much change has occurred in a plaque. To address this issue, we developed a scalar, easy-to-interpret and discriminative biomarker integrating the 376 plaque features extracted in Ref. [14], which encompass textural features used in previous studies. In addition, a physical understanding of the biomarker should be established for it to be accepted by clinicians. One way of characterizing the biomarker is to identify important textural features that contribute to the biomarker’s ability for discriminating treatment and placebo subjects. A metric was proposed in this study to determine the weight of each of the 376 features in the proposed biomarker (Eq. (7)).
The proposed biomarker was validated in a clinical trial aiming to evaluate the effect of pomegranate juice and tablets. Lipid oxidation in arterial macrophages and lipoproteins is a major contributor to carotid atherosclerosis. As a rich source of potent anti-oxidants [22], pomegranate juice has been shown to inhibit LDL oxidation and attenuate atherosclerosis development in animal studies [23], [24]. However, a recent 18-month randomized placebo-controlled trial aiming at assessing the effect of pomegranate juice in 289 subjects at risk of coronary disease and stroke showed that no difference in carotid IMT change was found between the pomegranate and the control group (p = 0.65) [25], thereby casting a doubt on whether pomegranate juice is effective in slowing the progression of atherosclerosis.
As pomegranate is a dietary supplement expected to confer smaller beneficial effect than high-dose atorvastatin, the requirement on the discriminative power for the proposed biomarker is higher than that required for detecting the effect of atorvastatin in Ref. [13]. We hypothesized that the proposed biomarker has enough discriminative power in detecting the effects of pomegranate juice/tablets in a placebo-controlled study. This hypothesis was extensively evaluated in the current study.
Section snippets
Study subjects and ultrasound image acquisition and preprocessing
Subjects were recruited at the Stroke Prevention & Atherosclerosis Research Centre at Robarts Research Institute (London, Ontario, Canada) for a registered clinical trial (ISRCTN30768139). The 171 subjects involved in this study were randomized into two groups. 96 subjects received pomegranate extract in a tablet and juice form once daily, and 75 subjects received a placebo-matching tablet and juice lacking the active ingredients once daily. A total of 629 carotid plaques were identified from
Results
Table 2 shows the background medications taken by pomegranate and placebo subjects. No statistically significant difference was observed between the background medications taken by the two groups. Plaque textural and volume measurements were obtained at baseline and 376 ± 23 days (range: 283 - 579 days) later. Since the time interval between baseline and follow-up varies with subjects, TPV and textural change per year were calculated to allow comparisons among subjects. Nx for PCA and K for
Discussion
We developed a cost-effective and computationally efficient texture-based biomarker that detected changes in atherosclerotic plaques due to dietary supplements such as pomegranate. A previous study of the effect of pomegranate juice on carotid IMT [25] showed no significant effect, but the finding may be attributed to the insensitivity of IMT in detecting treatment effect. TPV, as a three-dimensional measurement, has shown to be more able to detect change [51], [52] and more correlated to
Declaration of Competing Interest
The patient data were acquired in a clinical study sponsored by POM Wonderful, LLC (USA). Professor Spence is the Scientific Officer of Vascularis Inc.
Acknowledgments
Dr. Chiu is grateful for funding support from the Research Grant Council of the HKSAR, China (Project nos. CityU 11205917, CityU 11203218) and the City University of Hong Kong Strategic Research Grants (nos. 7004617, 7005226).
References (79)
- et al.
Stroke in China: advances and challenges in epidemiology, prevention, and management
Lancet Neurol.
(2019) - et al.
Validation of 3D ultrasound vessel wall volume: an imaging phenotype of carotid atherosclerosis
Ultrasound Med. Biol.
(2007) - et al.
Stroke risk stratification and its validation using ultrasonic echolucent carotid wall plaque morphology: a machine learning paradigm
Comput. Biol. Med.
(2017) - et al.
Effects of consumption of pomegranate juice on carotid intima–media thickness in men and women at moderate risk for coronary heart disease
Am. J. Cardiol.
(2009) - et al.
Quantification of carotid plaque volume measurements using 3D ultrasound imaging
Ultrasound Med. Biol.
(2005) - et al.
Image processing
Digit. Image Process.
(2007) - et al.
Atherosclerotic risk stratification strategy for carotid arteries using texture-based features
Ultrasound Med. Biol.
(2012) Texture analysis using gray level run lengths
Comput. Graph. Image Process.
(1975)- et al.
Shape-adapted smoothing in estimation of 3-D shape cues from affine deformations of local 2-D brightness structure
Image Vis. Comput.
(1997) Time course of atherosclerosis regression
Atherosclerosis
(2014)
Carotid plaque burden as a measure of subclinical atherosclerosis: comparison with other tests for subclinical arterial disease in the high risk plaque bioimage study
JACC
Relationship between ultrasound texture classification images and histology of atherosclerotic plaque
Ultrasound Med. Biol.
Duplex ultrasonography and selection of patients for carotid endarterectomy: plaque morphology or luminal narrowing?
J. Vasc. Surg.
The role of plaque morphology and diameter reduction in the development of new symptoms in asymptomatic carotid arteries
J. Vasc. Surg.
Minor asymptomatic carotid stenosis contralateral to carotid endarterectomy (CEA): our experience
Eur. J. Vasc. Surg.
Symptoms, stenosis and carotid plaque morphology is plaque morphology relevant?
Eur. J. Vasc. Endovasc. Surg.
Natural history of asymptomatic carotid plaque: five year follow-up study
Am. J. Surg.
Carotid artery duplex scanning: does plaque echogenicity correlate with patient symptoms?
Eur. J. Vasc. Endovasc. Surg.
Carotid plaque histology using real-time ultrasonography: clinical and therapeutic implications
Am. J. Surg.
Lipid-rich carotid artery plaques appear echolucent on ultrasound b-mode images and may be associated with intraplaque haemorrhage
Eur. J. Vasc. Endovasc. Surg.
The use of speckle reduction imaging (SRI) ultrasound in the characterization of carotid artery plaques
Eur. J. Radiol.
Global, regional, and national burden of stroke, 1990–2016: a systematic analysis for the global burden of disease study 2016
Lancet Neurol.
Intensive management of risk factors for accelerated atherosclerosis: the role of multiple interventions
Curr. Neurol. Neurosci. Rep.
Recent advances in pathogenesis, assessment, and treatment of atherosclerosis
F1000Research
Correlation between carotid intimal/medial thickness and atherosclerosis: a point of view from pathology
Arterioscler. Thrombosis Vasc. Biol.
Common carotid intima-media thickness measurements in cardiovascular risk prediction: a meta-analysis
Jama
Baseline values but not treatment-induced changes in carotid intima-media thickness predict incident cardiovascular events in treated hypertensive patients: findings in the European Lacidipine Study on Atherosclerosis (ELSA)
Circulation
Ultrasound measurement of atherosclerosis
Stroke
Development of 3D ultrasound techniques for carotid artery disease assessment and monitoring
Int. J. Comput. Assist. Radiol. Surg.
Progression of carotid plaque volume predicts cardiovascular events
Stroke
3D ultrasound measurement of change in carotid plaque volume: a tool for rapid evaluation of new therapies
Stroke
Texture analysis of carotid artery atherosclerosis from three-dimensional ultrasound images
Med. Phys.
Three-dimensional carotid ultrasound plaque texture predicts vascular events
Stroke
Fractal dimension based carotid plaque characterization from three-dimensional ultrasound images
Med. Biol. Eng. Comput.
Texture-based classification of atherosclerotic carotid plaques
IEEE Trans. Med. Imaging
Atherosclerotic plaque tissue characterization in 2D ultrasound longitudinal carotid scans for automated classification: a paradigm for stroke risk assessment
Med. Biol. Eng. Comput.
Assessment of carotid artery plaque components with machine learning classification using homodyned-K parametric maps and elastograms
IEEE Trans. Ultrason. Ferroelectr. Freq. Control
An integrated system for assessing stroke risk
IEEE Eng. Med. Biol. Mag.
Classification of carotid plaque echogenicity by combining texture features and morphologic characteristics
J. Ultrasound Med.
Cited by (11)
Quality assessment of stroke radiomics studies: Promoting clinical application
2023, European Journal of RadiologyQuantification of Local Vessel Wall and Plaque Volume Change for Assessment of Effects of Therapies on Carotid Atherosclerosis Based on 3-D Ultrasound Imaging
2023, Ultrasound in Medicine and BiologyThree-dimensional ultrasound assessment of effects of therapies on carotid atherosclerosis using vessel wall thickness maps
2021, Ultrasound in Medicine and BiologyCitation Excerpt :However, this may be due to the limited sensitivity of IMT to the effect of therapy. We have previously developed two plaque texture biomarkers (Chen et al., 2020; Lin et al., 2020) that established the effect of the pomegranate juice and extract in the same subject cohort investigated in the current paper. However, measurements of plaque texture require reproducible plaque segmentation.
Multiclass and Multilabel Classifications by Consensus and Complementarity-Based Multiview Latent Space Projection
2024, IEEE Transactions on Systems, Man, and Cybernetics: SystemsLocal Vessel Wall and Plaque Volume Evaluation of Three-Dimensional Carotid Ultrasound Images for Sensitive Assessment of the Effect of Therapies on Atherosclerosis
2023, 3D Ultrasound: Devices, Applications, and AlgorithmsInterpretable Deep Biomarker for Serial Monitoring of Carotid Atherosclerosis Based on Three-Dimensional Ultrasound Imaging
2023, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)