Automated versus manual in vivo segmentation of carotid plaque MRI

A 3D framework for segmentation of carotid artery vessel wall and identification of plaque compositions in multi-sequence MR images

Accurately assessing carotid artery wall thickening and identifying risky plaque components are critical for early diagnosis and risk management of carotid atherosclerosis. In this paper, we present a 3D framework for automated segmentation of the carotid artery vessel wall and identification of the compositions of carotid plaque in multi-sequence magnetic resonance (MR) images under the challenge of imperfect manual labeling. Manual labeling is commonly done in 2D slices of these multi-sequence MR images and often lacks perfect alignment across 2D slices and the multiple MR sequences, leading to labeling inaccuracies. To address such challenges, our framework is split into two parts: a segmentation subnetwork and a plaque component identification subnetwork. Initially, a 2D localization network pinpoints the carotid artery's position, extracting the region of interest (ROI) from the input images. Following that, a signed-distance-map-enabled 3D U-net (Çiçek etal, 2016)an adaptation of the nnU-net (Ronneberger and Fischer, 2015) segments the carotid artery vessel wall. This method allows for the concurrent segmentation of the vessel wall area using the signed distance map (SDM) loss (Xue et al., 2020) which regularizes the segmentation surfaces in 3D and reduces erroneous segmentation caused by imperfect manual labels. Subsequently, the ROI of the input images and the obtained vessel wall masks are extracted and combined to obtain the identification results of plaque components in the identification subnetwork. Tailored data augmentation operations are introduced into the framework to reduce the false positive rate of calcification and hemorrhage identification. We trained and tested our proposed method on a dataset consisting of 115 patients, and it achieves an accurate segmentation result of carotid artery wall (0.8459 Dice), which is superior to the best result in published studies (0.7885 Dice). Our approach yielded accuracies of 0.82, 0.73 and 0.88 for the identification of calcification, lipid-rich core and hemorrhage components. Our proposed framework can be potentially used in clinical and research settings to help radiologists perform cumbersome reading tasks and evaluate the risk of carotid plaques.

Learning carotid vessel wall segmentation in black-blood MRI using sparsely sampled cross-sections from 3D data

Purpose

Atherosclerosis of the carotid artery is a major risk factor for stroke. Quantitative assessment of the carotid vessel wall can be based on cross-sections of three-dimensional (3D) black-blood magnetic resonance imaging (MRI). To increase reproducibility, a reliable automatic segmentation in these cross-sections is essential.

Approach

We propose an automatic segmentation of the carotid artery in cross-sections perpendicular to the centerline to make the segmentation invariant to the image plane orientation and allow a correct assessment of the vessel wall thickness (VWT). We trained a residual U-Net on eight sparsely sampled cross-sections per carotid artery and evaluated if the model can segment areas that are not represented in the training data. We used 218 MRI datasets of 121 subjects that show hypertension and plaque in the ICA or CCA measuring ≥ 1.5    mm in ultrasound.

Results

The model achieves a high mean Dice coefficient of 0.948/0.859 for the vessel's lumen/wall, a low mean Hausdorff distance of 0.417 / 0.660    mm , and a low mean average contour distance of 0.094 / 0.119    mm on the test set. The model reaches similar results for regions of the carotid artery that are not incorporated in the training set and on MRI of young, healthy subjects. The model also achieves a low median Hausdorff distance of 0.437 / 0.552    mm on the 2021 Carotid Artery Vessel Wall Segmentation Challenge test set.

Conclusions

The proposed method can reduce the effort for carotid artery vessel wall assessment. Together with human supervision, it can be used for clinical applications, as it allows a reliable measurement of the VWT for different patient demographics and MRI acquisition settings.

Carotid atherosclerotic plaque segmentation in multi-weighted MRI using a two-stage neural network: advantages of training with high-resolution imaging and histology

Introduction

A reliable and automated method to segment and classify carotid artery atherosclerotic plaque components is needed to efficiently analyze multi-weighted magnetic resonance (MR) images to allow their integration into patient risk assessment for ischemic stroke. Certain plaque components such as lipid-rich necrotic core (LRNC) with hemorrhage suggest a greater likelihood of plaque rupture and stroke event. Assessment for presence and extent of LRNC could assist in directing treatment with impact upon patient outcomes.

Methods

To address the need to accurately determine the presence and extent of plaque components on carotid plaque MRI, we proposed a two-staged deep-learning-based approach that consists of a convolutional neural network (CNN), followed by a Bayesian neural network (BNN). The rationale for the two-stage network approach is to account for the class imbalance of vessel wall and background by providing an attention mask to the BNN. A unique feature of the network training was to use ground truth defined by both high-resolution ex vivo MRI data and histopathology. More specifically, standard resolution 1.5 T in vivo MR image sets with corresponding high resolution 3.0 T ex vivo MR image sets and histopathology image sets were used to define ground-truth segmentations. Of these, data from 7 patients was used for training and from the remaining two was used for testing the proposed method. Next, to evaluate the generalizability of the method, we tested the method with an additional standard resolution 3.0 T in vivo data set of 23 patients obtained from a different scanner.

Results

Our results show that the proposed method yielded accurate segmentation of carotid atherosclerotic plaque and outperforms not only manual segmentation by trained readers, who did not have access to the ex vivo or histopathology data, but also three state-of-the-art deep-learning-based segmentation methods. Further, the proposed approach outperformed a strategy where the ground truth was generated without access to the high resolution ex vivo MRI and histopathology. The accurate performance of this method was also observed in the additional 23-patient dataset from a different scanner.

Conclusion

In conclusion, the proposed method provides a mechanism to perform accurate segmentation of the carotid atherosclerotic plaque in multi-weighted MRI. Further, our study shows the advantages of using high-resolution imaging and histology to define ground truth for training deep-learning-based segmentation methods.

The Differentiation in Image Post-processing and 3D Reconstruction During Evaluation of Carotid Plaques From MR and CT Data Sources

Background: Carotid plaque morphology and tissue composition help assess risk stratification of stroke events. Many post-processing image techniques based on CT and MR images have been widely used in related research, such as image segmentation, 3D reconstruction, and computer fluid dynamics. However, the criteria for the 3D numerical model of carotid plaque established by CT and MR angiographic image data remain open to questioning.

Method: We accurately duplicated the geometry and simulated it using computer software to make a 3D numerical model. The initial images were obtained by CTA and TOF-MRA. MIMICS (Materialize’s interactive medical image control system) software was used to process the images to generate three-dimensional solid models of blood vessels and plaques. The subsequent output was exported to the ANSYS software to generate finite element simulation results for the further hemodynamic study.

Results: The 3D models of carotid plaque of TOF-MRA and CTA were simulated by using computer software. CTA has a high-density resolution for carotid plaque, the boundary of the CTA image is obvious, and the main component of which is a calcified tissue. However, the density resolution of TOF-MRA for the carotid plaque and carotid artery was not as good as that of CTA. The results show that there is a large deviation between the TOF-MRA and CTA 3D model of plaque in the carotid artery due to the unclear recognition of plaque boundary during 3D reconstruction, and this can further affect the simulation results of hemodynamics.

Conclusion: In this study, two-dimensional images and three-dimensional models of carotid plaques obtained by two angiographic techniques were compared. The potential of these two imaging methods in clinical diagnosis and fluid dynamics of carotid plaque was evaluated, and the selectivity of image post-processing analysis to original medical image acquisition was revealed.

A deep learning oriented method for automated 3D reconstruction of carotid arterial trees from MR imaging

The scope of this paper is to present a new carotid vessel segmentation algorithm implementing the U-net based convolutional neural network architecture. With carotid atherosclerosis being the major cause of stroke in Europe, new methods that can provide more accurate image segmentation of the carotid arterial tree and plaque tissue can help improve early diagnosis, prevention and treatment of carotid disease. Herein, we present a novel methodology combining the U-net model and morphological active contours in an iterative framework that accurately segments the carotid lumen and outer wall. The method automatically produces a 3D meshed model of the carotid bifurcation and smaller branches, using multispectral MR image series obtained from two clinical centres of the TAXINOMISIS study. As indicated by a validation study, the algorithm succeeds high accuracy (99.1% for lumen area and 92.6% for the perimeter) for lumen segmentation. The proposed algorithm will be used in the TAXINOMISIS study to obtain more accurate 3D vessel models for improved computational fluid dynamics simulations and the development of models of atherosclerotic plaque progression.

Towards Automated Quantification of Vessel Wall Composition Using MRI

BACKGROUND

MRI can be used to generate fat fraction (FF) and R2* data, which have been previously shown to characterize the plaque compositional features lipid-rich necrotic core (LRNC) and intraplaque hemorrhage (IPH) in the carotid arteries (CAs). Previously, these data were extracted from CA plaques using time-consuming manual analyses.

PURPOSE

To design and demonstrate a method for segmenting the CA and extracting data describing the composition of the vessel wall.

STUDY TYPE

Prospective.

SUBJECTS

31 subjects from the Swedish CArdioPulmonary bioImage Study (SCAPIS).

FIELD STRENGTH/SEQUENCES

T₁ -weighted (T₁ W) quadruple inversion recovery, contrast-enhanced MR angiography (CE-MRA), and 4-point Dixon data were acquired at 3T.

ASSESSMENT

The vessel lumen of the CA was automatically segmented using support vector machines (SVM) with CE-MRA data, and the vessel wall region was subsequently delineated. Automatically generated segmentations were quantitatively measured and three observers visually compared the segmentations to manual segmentations performed on T₁ w images. Dixon data were used to generate FF and R2* maps. Both manually and automatically generated segmentations of the CA and vessel wall were used to extract compositional data.

STATISTICAL TESTS

Two-tailed t-tests were used to examine differences between results generated using manual and automated analyses, and among different configurations of the automated method. Interobserver agreement was assessed with Fleiss' kappa.

RESULTS

Automated segmentation of the CA using SVM had a Dice score of 0.89 ± 0.02 and true-positive ratio 0.93 ± 0.03 when compared against ground truth, and median qualitative score of 4/5 when assessed visually by multiple observers. Vessel wall regions of 0.5 and 1 mm yielded compositional information similar to that gained from manual analyses. Using the 0.5 mm vessel wall region, the mean difference was 0.1 ± 2.5% considering FF and 1.1 ± 5.7[1/s] for R2*.

LEVEL OF EVIDENCE

1.

TECHNICAL EFFICACY STAGE

1. J. Magn. Reson. Imaging 2020;52:710-719.

Non-invasive imaging techniques and assessment of carotid vasa vasorum neovascularization: Promises and pitfalls

Carotid adventitia vasa vasorum neovascularization (VVn) is associated with the initial stages of arteriosclerosis and with the formation of unstable plaque. However, techniques to accurately quantify that neovascularization in a standard, fast, non-invasive, and efficient way are still lacking. The development of such techniques holds the promise of enabling wide, inexpensive, and safe screening programs that could stratify patients and help in personalized preventive cardiovascular medicine. In this paper, we review the recent scientific literature pertaining to imaging techniques that could set the stage for the development of standard methods for quantitative assessment of atherosclerotic plaque and carotid VVn. We present and discuss the alternative imaging techniques being used in clinical practice and we review the computational developments that are contributing to speed up image analysis and interpretation. We conclude that one of the greatest upcoming challenges will be the use of machine learning techniques to develop automated methods that assist in the interpretation of images to stratify patients according to their risk.

Pierre S, Newman WP, Rae G, Perlman ES, Rosol M, Keith JC, Buckler AJ. Atherosclerotic Plaque Tissue: Noninvasive Quantitative Assessment of Characteristics with Software-aided Measurements from Conventional CT Angiography

Purpose To (a) evaluate whether plaque tissue characteristics determined with conventional computed tomographic (CT) angiography could be quantitated at higher levels of accuracy by using image processing algorithms that take characteristics of the image formation process coupled with biologic insights on tissue distributions into account by comparing in vivo results and ex vivo histologic findings and (b) assess reader variability. Materials and Methods Thirty-one consecutive patients aged 43-85 years (average age, 64 years) known to have or suspected of having atherosclerosis who underwent CT angiography and were referred for endarterectomy were enrolled. Surgical specimens were evaluated with histopathologic examination to serve as standard of reference. Two readers used lumen boundary to determine scanner blur and then optimized component densities and subvoxel boundaries to best fit the observed image by using semiautomatic software. The accuracy of the resulting in vivo quantitation of calcification, lipid-rich necrotic core (LRNC), and matrix was assessed with statistical estimates of bias and linearity relative to ex vivo histologic findings. Reader variability was assessed with statistical estimates of repeatability and reproducibility. Results A total of 239 cross sections obtained with CT angiography and histologic examination were matched. Performance on held-out data showed low levels of bias and high Pearson correlation coefficients for calcification (-0.096 mm2 and 0.973, respectively), LRNC (1.26 mm2 and 0.856), and matrix (-2.44 mm2 and 0.885). Intrareader variability was low (repeatability coefficient ranged from 1.50 mm2 to 1.83 mm2 among tissue characteristics), as was interreader variability (reproducibility coefficient ranged from 2.09 mm2 to 4.43 mm2). Conclusion There was high correlation and low bias between the in vivo software image analysis and ex vivo histopathologic quantitative measures of atherosclerotic plaque tissue characteristics, as well as low reader variability. Software algorithms can mitigate the blurring and partial volume effects of routine CT angiography acquisitions to produce accurate quantification to enhance current clinical practice. Clinical trial registration no. NCT02143102 © RSNA, 2017 Online supplemental material is available for this article. An earlier incorrect version of this article appeared online. This article was corrected on September 15, 2017.

Association of Progression of Carotid Artery Wall Volume and Recurrent Transient Ischemic Attack or Stroke: A Magnetic Resonance Imaging Study

BACKGROUND AND PURPOSE

This study aimed to investigate the association between carotid plaque progression and subsequent recurrent events using magnetic resonance imaging.

METHODS

Sixty-three symptomatic patients with ipsilateral carotid atherosclerotic stenosis (30%-69% stenosis) determined by ultrasound underwent first and second carotid artery magnetic resonance imaging for carotid artery at baseline and ≥6 months after the first scan, respectively. All the patients had clinical follow-up after the second magnetic resonance scan for ≤5 years until the onset of recurrent transient ischemic attack or stroke. Presence/absence of carotid plaque compositional features, particularly intraplaque hemorrhage and fibrous cap rupture was identified. The annual progression of carotid wall volume between 2 magnetic resonance scans was measured. Univariate and multivariate Cox regression was used to calculate the hazard ratio and corresponding 95% confidence interval of carotid plaque features in discriminating recurrent events. Receiver-operating-characteristic-curve analysis was conducted to determine the area-under-the-curve of carotid plaque features in predicting recurrent events.

RESULTS

Sixty-three patients (mean age: 66.5±10.0 years old; 54 males) were eligible for final statistics analysis. During a mean follow-up duration of 55.1±13.6 months, 14.3% of patients (n=9) experienced ipsilateral recurrent transient ischemic attack/stroke. The annual progression of carotid wall volume was significantly associated with recurrent events before (hazard ratio, 1.14 per 10 mm³; 95% confidence interval, 1.02-1.27; P=0.019) and after (hazard ratio, 1.19 per 10 mm3; 95% confidence interval, 1.03-1.37; P=0.022) adjusted for confounding factors. In discriminating the recurrence of transient ischemia attack/stroke, receiver-operator curve analysis indicated that combined with annual progression of wall volume, there was a significant incremental improvement in the area-under-the-curve of intraplaque hemorrhage (area-under-the-curve: 0.69-0.81) and fibrous cap rupture (area-under-the-curve: 0.73-0.84).

CONCLUSIONS

The annual progression of carotid wall volume is independently associated with recurrent ischemic cerebrovascular events, and this measurement has added value for intraplaque hemorrhage and fibrous cap rupture in predicting future events.

Unstable carotid artery plaque: new insights and controversies in diagnostics and treatment

Cardiovascular disease is estimated to be the leading cause of death, globally causing 14 million deaths each year. Stroke remains a massive public health problem and there is an increasing need for better strategies for the prevention and treatment of this disease. At least 20% of ischemic strokes are thromboembolic in nature, caused by a thromboembolism from an atherosclerotic plaque at the carotid bifurcation or the internal carotid artery. Current clinical guidelines for both primary and secondary prevention of stroke in patients with carotid stenosis caused by atherosclerotic plaques remain reliant on general patient characteristics (traditional risk factors for stroke) and static measures of the degree of artery stenosis. Patients with similar traditional risk factors, however, have been found to have different risk of stroke, and it has in recent years become increasingly clear that the degree of artery stenosis alone is not the best estimation of stroke risk. There is a need for new methods for the assessment of stroke risk to improve risk prediction for the individual patient. This review aims to give an overview of new methods available for the identification of carotid plaque instability and the assessment of stroke risk.

Manual versus Automated Carotid Artery Plaque Component Segmentation in High and Lower Quality 3.0 Tesla MRI Scans

PURPOSE

To study the interscan reproducibility of manual versus automated segmentation of carotid artery plaque components, and the agreement between both methods, in high and lower quality MRI scans.

METHODS

24 patients with 30-70% carotid artery stenosis were planned for 3T carotid MRI, followed by a rescan within 1 month. A multicontrast protocol (T1w,T2w, PDw and TOF sequences) was used. After co-registration and delineation of the lumen and outer wall, segmentation of plaque components (lipid-rich necrotic cores (LRNC) and calcifications) was performed both manually and automated. Scan quality was assessed using a visual quality scale.

RESULTS

Agreement for the detection of LRNC (Cohen's kappa (k) is 0.04) and calcification (k = 0.41) between both manual and automated segmentation methods was poor. In the high-quality scans (visual quality score ≥ 3), the agreement between manual and automated segmentation increased to k = 0.55 and k = 0.58 for, respectively, the detection of LRNC and calcification larger than 1 mm2. Both manual and automated analysis showed good interscan reproducibility for the quantification of LRNC (intraclass correlation coefficient (ICC) of 0.94 and 0.80 respectively) and calcified plaque area (ICC of 0.95 and 0.77, respectively).

CONCLUSION

Agreement between manual and automated segmentation of LRNC and calcifications was poor, despite a good interscan reproducibility of both methods. The agreement between both methods increased to moderate in high quality scans. These findings indicate that image quality is a critical determinant of the performance of both manual and automated segmentation of carotid artery plaque components.

A Convolutional Neural Network for Automatic Characterization of Plaque Composition in Carotid Ultrasound

Characterization of carotid plaque composition, more specifically the amount of lipid core, fibrous tissue, and calcified tissue, is an important task for the identification of plaques that are prone to rupture, and thus for early risk estimation of cardiovascular and cerebrovascular events. Due to its low costs and wide availability, carotid ultrasound has the potential to become the modality of choice for plaque characterization in clinical practice. However, its significant image noise, coupled with the small size of the plaques and their complex appearance, makes it difficult for automated techniques to discriminate between the different plaque constituents. In this paper, we propose to address this challenging problem by exploiting the unique capabilities of the emerging deep learning framework. More specifically, and unlike existing works which require a priori definition of specific imaging features or thresholding values, we propose to build a convolutional neural network (CNN) that will automatically extract from the images the information that is optimal for the identification of the different plaque constituents. We used approximately 90 000 patches extracted from a database of images and corresponding expert plaque characterizations to train and to validate the proposed CNN. The results of cross-validation experiments show a correlation of about 0.90 with the clinical assessment for the estimation of lipid core, fibrous cap, and calcified tissue areas, indicating the potential of deep learning for the challenging task of automatic characterization of plaque composition in carotid ultrasound.

Semiautomated Magnetic Resonance Imaging Assessment of Carotid Plaque Lipid Content

BACKGROUND

The composition of a carotid plaque is important for plaque vulnerability and stroke risk. The main aim of this study was to assess the potential of semiautomated segmentation of carotid plaque magnetic resonance imaging (MRI) in the assessment of the size of the lipid-rich necrotic core (LRNC).

METHODS

Thirty-four consecutive patients with carotid stenosis of 70% or higher, who were scheduled for carotid endarterectomy, underwent a clinical neurological examination, Color duplex ultrasound, 3-T MRI with an 8-channel carotid coil, and blood tests. All examinations were performed less than 24 hours prior to surgery and plaques were assessed histologically immediately following endarterectomy. Plaques were defined as symptomatic when associated with ipsilateral cerebral ischemic symptoms within 30 days prior to inclusion. The level of agreement between the size of the LRNC and calcification on MRI to the histological estimation of the same tissue components, plaque echolucency on ultrasound, and symptoms was assessed.

RESULTS

The size of the LRNC on MRI was significantly correlated to the percentage amount of lipid per plaque on histological assessment (P = .010, r = .5), and to echogenicity on ultrasound with echolucent plaques having larger LRNC than echogenic plaques (P = .001, r = -.7).

CONCLUSIONS

In this study, we found that semiautomated MRI assessments of the percentage LRNC in carotid plaques were significantly correlated to the percentage LRNC per plaque on histological assessment, and to echogenicity on ultrasound with echolucent plaques having larger LRNC than echogenic plaques.

Comparison of Gated and Ungated Black-Blood Fast Spin-echo MRI of Carotid Vessel Wall at 3T

PURPOSE

Multi-slice ungated double inversion recovery has been proposed as an alternative time-efficient and effective sequence for black-blood carotid imaging. The purpose of this study is to evaluate the comparative repeatability of this multi-contrast sequence with respect to a single slice double inversion recovery prepared gated sequence.

MATERIALS AND METHODS

Ten healthy volunteers and three patients with Doppler ultrasound defined carotid artery stenosis >30% were recruited. T1-weighted (T1W) and T2W fast spin-echo (FSE) images were acquired centered at the carotid bifurcation with and without cardiac gating. Repeat imaging was performed without patient repositioning to determine the variations in vessel wall measurement and signal intensity due to gating, while negating variations as a result of slice misalignment and anatomical displacement relative to the receiver coil. The distributions and the repeatability of lumen area, vessel wall area, signal and contrast-to-noise ratio (SNR/CNR) of the vessel wall and adjacent muscle were reported.

RESULTS

The T1W ungated sequence generally had comparable wall SNR/CNR with respect to the gated sequence, however the muscle SNR was lower (P = 0.013). The T2W ungated multi-slice sequence had lower SNR/CNR than the gated single slice sequence (P < 0.001), but with equivalent effective wall CNR (P = 0.735). Vessel area measurements using the gated/ungated sequences were equivalent. Ungated sequences had better repeatability in SNR/CNR than the gated sequences with borderline and statistically significant differences. The repeatability of T2W wall area measurement was better using the ungated sequences (P = 0.02), and the repeatability of the remaining vessel area measurements were equivalent.

CONCLUSIONS

Ungated sequences can achieve comparable SNR/CNR and equivalent carotid vessel area measurements than gated sequences with improved repeatability of SNR/CNR. Ungated sequences are good alternatives of gated sequences for vessel area measurement and plaque composition quantification.

In vivo semi-automatic segmentation of multicontrast cardiovascular magnetic resonance for prospective cohort studies on plaque tissue composition: initial experience

Automatic in vivo segmentation of multicontrast (multisequence) carotid magnetic resonance for plaque composition has been proposed as a substitute for manual review to save time and reduce inter-reader variability in large-scale or multicenter studies. Using serial images from a prospective longitudinal study, we sought to compare a semi-automatic approach versus expert human reading in analyzing carotid atherosclerosis progression. Baseline and 6-month follow-up multicontrast carotid images from 59 asymptomatic subjects with 16-79 % carotid stenosis were reviewed by both trained radiologists with 2-4 years of specialized experience in carotid plaque characterization with MRI and a previously reported automatic atherosclerotic plaque segmentation algorithm, referred to as morphology-enhanced probabilistic plaque segmentation (MEPPS). Agreement on measurements from individual time points, as well as on compositional changes, was assessed using the intraclass correlation coefficient (ICC). There was good agreement between manual and MEPPS reviews on individual time points for calcification (CA) (area: ICC; 0.85-0.91; volume: ICC; 0.92-0.95) and lipid-rich necrotic core (LRNC) (area: ICC; 0.78-0.82; volume: ICC; 0.84-0.86). For compositional changes, agreement was good for CA volume change (ICC; 0.78) and moderate for LRNC volume change (ICC; 0.49). Factors associated with LRNC progression as detected by MEPPS review included intraplaque hemorrhage (positive association) and reduction in low-density lipoprotein cholesterol (negative association), which were consistent with previous findings from manual review. Automatic classifier for plaque composition produced results similar to expert manual review in a prospective serial MRI study of carotid atherosclerosis progression. Such automatic classification tools may be beneficial in large-scale multicenter studies by reducing image analysis time and avoiding bias between human reviewers.

Repeatability of in vivo quantification of atherosclerotic carotid artery plaque components by supervised multispectral classification

Objective

To evaluate the agreement and scan–rescan repeatability of automated and manual plaque segmentation for the quantification of in vivo carotid artery plaque components from multi-contrast MRI.

Materials and methods

Twenty-three patients with 30–70 % stenosis underwent two 3T MR carotid vessel wall exams within a 1 month interval. T1w, T2w, PDw and TOF images were acquired around the region of maximum vessel narrowing. Manual delineation of the vessel wall and plaque components (lipid, calcification, loose matrix) by an experienced observer provided the reference standard for training and evaluation of an automated plaque classifier. Areas of different plaque components and fibrous tissue were quantified and compared between segmentation methods and scan sessions.

Results

In total, 304 slices from 23 patients were included in the segmentation experiment, in which 144 aligned slice pairs were available for repeatability analysis. The correlation between manual and automated segmented areas was 0.35 for lipid, 0.66 for calcification, 0.50 for loose matrix and 0.82 for fibrous tissue. For the comparison between scan sessions, the coefficient of repeatability of area measurement obtained by automated segmentation was lower than by manual delineation for lipid (9.9 vs. 17.1 mm²), loose matrix (13.8 vs. 21.2 mm²) and fibrous tissue (24.6 vs. 35.0 mm²), and was similar for calcification (20.0 vs. 17.6 mm²).

Conclusion

Application of an automated classifier for segmentation of carotid vessel wall plaque components from in vivo MRI results in improved scan–rescan repeatability compared to manual analysis.

Multi-Center MRI Carotid Plaque Component Segmentation Using Feature Normalization and Transfer Learning

Automated segmentation of plaque components in carotid artery magnetic resonance imaging (MRI) is important to enable large studies on plaque vulnerability, and for incorporating plaque composition as an imaging biomarker in clinical practice. Especially supervised classification techniques, which learn from labeled examples, have shown good performance. However, a disadvantage of supervised methods is their reduced performance on data different from the training data, for example on images acquired with different scanners. Reducing the amount of manual annotations required for each new dataset will facilitate widespread implementation of supervised methods. In this paper we segment carotid plaque components of clinical interest (fibrous tissue, lipid tissue, calcification and intraplaque hemorrhage) in a multi-center MRI study. We perform voxelwise tissue classification by traditional same-center training, and compare results with two approaches that use little or no annotated same-center data. These approaches additionally use an annotated set of different-center data. We evaluate 1) a nonlinear feature normalization approach, and 2) two transfer-learning algorithms that use same and different-center data with different weights. Results showed that the best results were obtained for a combination of feature normalization and transfer learning. While for the other approaches significant differences in voxelwise or mean volume errors were found compared with the reference same-center training, the proposed approach did not yield significant differences from that reference. We conclude that both extensive feature normalization and transfer learning can be valuable for the development of supervised methods that perform well on different types of datasets.

Atherosclerotic plaque component segmentation in combined carotid MRI and CTA data incorporating class label uncertainty

Atherosclerotic plaque composition can indicate plaque vulnerability. We segment atherosclerotic plaque components from the carotid artery on a combination of in vivo MRI and CT-angiography (CTA) data using supervised voxelwise classification. In contrast to previous studies the ground truth for training is directly obtained from 3D registration with histology for fibrous and lipid-rich necrotic tissue, and with μCT for calcification. This registration does, however, not provide accurate voxelwise correspondence. We therefore evaluate three approaches that incorporate uncertainty in the ground truth used for training: I) soft labels are created by Gaussian blurring of the original binary histology segmentations to reduce weights at the boundaries between components, and are weighted by the estimated registration accuracy of the histology and in vivo imaging data (measured by overlap), II) samples are weighted by the local contour distance of the lumen and outer wall between histology and in vivo data, and III) 10% of each class is rejected by Gaussian outlier rejection. Classification was evaluated on the relative volumes (% of tissue type in the vessel wall) for calcified, fibrous and lipid-rich necrotic tissue, using linear discriminant (LDC) and support vector machine (SVM) classification. In addition, the combination of MRI and CTA data was compared to using only one imaging modality. Best results were obtained by LDC and outlier rejection: the volume error per vessel was 0.9±1.0% for calcification, 12.7±7.6% for fibrous and 12.1±8.1% for necrotic tissue, with Spearman rank correlation coefficients of 0.91 (calcification), 0.80 (fibrous) and 0.81 (necrotic). While segmentation using only MRI features yielded low accuracy for calcification, and segmentation using only CTA features yielded low accuracy for necrotic tissue, the combination of features from MRI and CTA gave good results for all studied components.

Spatio-temporal texture (SpTeT) for distinguishing vulnerable from stable atherosclerotic plaque on dynamic contrast enhancement (DCE) MRI in a rabbit model

PURPOSE

To develop a new spatio-temporal texture (SpTeT) based method for distinguishing vulnerable versus stable atherosclerotic plaques on DCE-MRI using a rabbit model of atherothrombosis.

METHODS

Aortic atherosclerosis was induced in 20 New Zealand White rabbits by cholesterol diet and endothelial denudation. MRI was performed before (pretrigger) and after (posttrigger) inducing plaque disruption with Russell's-viper-venom and histamine. Of the 30 vascular targets (segments) under histology analysis, 16 contained thrombus (vulnerable) and 14 did not (stable). A total of 352 voxel-wise computerized SpTeT features, including 192 Gabor, 36 Kirsch, 12 Sobel, 52 Haralick, and 60 first-order textural features, were extracted on DCE-MRI to capture subtle texture changes in the plaques over the course of contrast uptake. Different combinations of SpTeT feature sets, in which the features were ranked by a minimum-redundancy-maximum-relevance feature selection technique, were evaluated via a random forest classifier. A 500 iterative 2-fold cross validation was performed for discriminating the vulnerable atherosclerotic plaque and stable atherosclerotic plaque on per voxel basis. Four quantitative metrics were utilized to measure the classification results in separating between vulnerable and stable plaques.

RESULTS

The quantitative results show that the combination of five classes of SpTeT features can distinguish between vulnerable (disrupted plaques with an overlying thrombus) and stable plaques with the best AUC values of 0.9631 ± 0.0088, accuracy of 89.98% ± 0.57%, sensitivity of 83.71% ± 1.71%, and specificity of 94.55% ± 0.48%.

CONCLUSIONS

Vulnerable and stable plaque can be distinguished by SpTeT based features. The SpTeT features, following validation on larger datasets, could be established as effective and reliable imaging biomarkers for noninvasively assessing atherosclerotic risk.

Can shape analysis differentiate free-floating internal carotid artery thrombus from atherosclerotic plaque in patients evaluated with CTA for stroke or transient ischemic attack?

RATIONALE AND OBJECTIVES

Patients presenting with transient ischemic attack or stroke may have symptom-related lesions on acute computed tomography angiography (CTA) such as free-floating intraluminal thrombus (FFT). It is difficult to distinguish FFT from carotid plaque, but the distinction is critical as management differs. By contouring the shape of these vascular lesions ("virtual endarterectomy"), advanced morphometric analysis can be performed. The objective of our study is to determine whether quantitative shape analysis can accurately differentiate FFT from atherosclerotic plaque.

MATERIALS AND METHODS

We collected 23 consecutive cases of suspected carotid FFT seen on CTA (13 men, 65 ± 10 years; 10 women, 65.5 ± 8.8 years). True-positive FFT cases (FFT+) were defined as filling defects resolving with anticoagulant therapy versus false-positives (FFT-), which remained unchanged. Lesion volumes were extracted from CTA images and quantitative shape descriptors were computed. The five most discriminative features were used to construct receiver operator characteristic (ROC) curves and to generate three machine-learning classifiers. Average classification accuracy was determined by cross-validation.

RESULTS

Follow-up imaging confirmed sixteen FFT+ and seven FFT- cases. Five shape descriptors delineated FFT+ from FFT- cases. The logistic regression model produced from combining all five shape features demonstrated a sensitivity of 87.5% and a specificity of 71.4% with an area under the ROC curve = 0.85 ± 0.09. Average accuracy for each classifier ranged from 65.2%-76.4%.

CONCLUSIONS

We identified five quantitative shape descriptors of carotid FFT. This shape "signature" shows potential for supplementing conventional lesion characterization in cases of suspected FFT.

Numerical simulations of carotid MRI quantify the accuracy in measuring atherosclerotic plaque components in vivo

PURPOSE

Atherosclerotic carotid plaques can be quantified in vivo by MRI. However, the accuracy in segmentation and quantification of components such as the thin fibrous cap (FC) and lipid-rich necrotic core (LRNC) remains unknown due to the lack of a submillimeter scale ground truth.

METHODS

A novel approach was taken by numerically simulating in vivo carotid MRI providing a ground truth comparison. Upon evaluation of a simulated clinical protocol, MR readers segmented simulated images of cross-sectional plaque geometries derived from histological data of 12 patients.

RESULTS

MR readers showed high correlation (R) and intraclass correlation (ICC) in measuring the luminal area (R = 0.996, ICC = 0.99), vessel wall area (R = 0.96, ICC = 0.94) and LRNC area (R = 0.95, ICC = 0.94). LRNC area was underestimated (mean error, -24%). Minimum FC thickness showed a mediocre correlation and intraclass correlation (R = 0.71, ICC = 0.69).

CONCLUSION

Current clinical MRI can quantify carotid plaques but shows limitations for thin FC thickness quantification. These limitations could influence the reliability of carotid MRI for assessing plaque rupture risk associated with FC thickness. Overall, MRI simulations provide a feasible methodology for assessing segmentation and quantification accuracy, as well as for improving scan protocol design.