Effect of black blood MR image quality on vessel wall segmentation

Three-dimensional carotid ultrasound segmentation variability dependence on signal difference and boundary orientation

Quantitative measurements of the progression (or regression) of carotid plaque burden are important in monitoring patients and evaluating new treatment options. We previously developed a quantitative metric to analyze changes in carotid plaque morphology from 3-D ultrasound (US) on a point-by-point basis. This method requires multiple segmentations of the arterial wall and lumen boundaries to obtain the local standard deviation (SD) of vessel-wall-plus-plaque thickness (VWT) so that t-tests could be used to determine whether a change in VWT is statistically significant. However, the requirement for multiple segmentations makes clinical trials laborious and time-consuming. Therefore, this study was designed to establish the relationship between local segmentation SD and local signal difference on the arterial wall and lumen boundaries. We propose metrics to quantify segmentation SD and signal difference on a point-by-point basis, and studied whether the signal difference at arterial wall or lumen boundaries could be used to predict local segmentation SD. The ability to predict the local segmentation SD could eliminate the need of repeated segmentations of a 2-D transverse image to obtain the local segmentation standard deviation, thereby making clinical trials less laborious and saving time. Six subjects involved in this study were associated with different degrees of atherosclerosis: three carotid stenosis subjects with mean plaque area >3 cm(2) and >60% carotid stenosis were involved in a clinical study evaluating the effect of atorvastatin, a cholesterol-lowering and plaque-stabilizing drug; and three subjects with carotid plaque area >0.5 cm(2) were subjects with moderate atherosclerosis. Our results suggest that when local signal difference is higher than 8 greyscale value (GSV), the local segmentation SD stabilizes at 0.05 mm and is thus predictable. This information provides a target value of local signal difference on the arterial boundaries that should be achieved to obtain an accurate prediction of local segmentation SD. (E-mail: bcychiu@alumni.uwo.ca).

Dependence of vessel area accuracy and precision as a function of MR imaging parameters and boundary detection algorithm

PURPOSE

To determine the appropriate image acquisition parameters for an accurate measurement of vessel cross-sectional area from MR angiography (MRA) images.

MATERIALS AND METHODS

A series of images with different vessel cross-sectional areas, resolutions, and signal-to-noise ratios (SNRs) were simulated and validated experimentally. Dynamic programming (DP) was employed to determine the accuracy and precision of the vessel cross-sectional area as a function of vessel size, sampling matrix, acquisition time, and postprocessing parameters such as zooming and bias correction.

RESULTS

We show that there is an optimal value of lambda (the ratio of vessel diameter to resolution) for a given intrinsic SNR that yields the most accurate and precise area measurement. Specifically, when the SNR is > or =10:1, this value of lambda is 8 and yields a cross-sectional area error of <5% with a zoom factor of > or =2.

CONCLUSION

The predicted ideal result of lambda = 8 is within reach with current technology to image vessels such as the carotid artery or aorta. It is possible to determine the ideal resolution that minimizes errors in the measurement of the vessel cross-sectional area for a given SNR, processing algorithm, and vessel of interest.

Automated classification of atherosclerotic plaque from magnetic resonance images using predictive models

The information contained within multicontrast magnetic resonance images (MRI) promises to improve tissue classification accuracy, once appropriately analyzed. Predictive models capture relationships empirically, from known outcomes thereby combining pattern classification with experience. In this study, we examine the applicability of predictive modeling for atherosclerotic plaque component classification of multicontrast ex vivo MR images using stained, histopathological sections as ground truth. Ten multicontrast images from seven human coronary artery specimens were obtained on a 9.4 T imaging system using multicontrast-weighted fast spin-echo (T1-, proton density-, and T2-weighted) imaging with 39-mum isotropic voxel size. Following initial data transformations, predictive modeling focused on automating the identification of specimen's plaque, lipid, and media. The outputs of these three models were used to calculate statistics such as total plaque burden and the ratio of hard plaque (fibrous tissue) to lipid. Both logistic regression and an artificial neural network model (Relevant Input Processor Network-RIPNet) were used for predictive modeling. When compared against segmentation resulting from cluster analysis, the RIPNet models performed between 25 and 30% better in absolute terms. This translates to a 50% higher true positive rate over given levels of false positives. This work indicates that it is feasible to build an automated system of plaque detection using MRI and data mining.

3D geometric reconstruction of thoracic aortic aneurysms

Background

The thoracic aortic aneurysm (TAA) is a pathology that involves an expansion of the aortic diameter in the thoracic aorta, leading to risk of rupture. Recent studies have suggested that internal wall stress, which is affected by TAA geometry and the presence or absence of thrombus, is a more reliable predictor of rupture than the maximum diameter, the current clinical criterion. Accurate reconstruction of TAA geometry is a crucial step in patient-specific stress calculations.

Methods

In this work, a novel methodology was developed, which combines data from several sets of magnetic resonance (MR) images with different levels of detail and different resolutions. Two sets of images were employed to create the final model, which has the highest level of detail for each component of the aneurysm (lumen, thrombus, and wall). A reference model was built by using a single set of images for comparison. This approach was applied to two patient-specific TAAs in the descending thoracic aorta.

Results

The results of finite element simulations showed differences in stress pattern between the coarse and fine models: higher stress values were found with the coarse model and the differences in predicted maximum wall stress were 30% for patient A and 11% for patient B.

Conclusion

This paper presents a new approach to the reconstruction of an aneurysm model based on the use of several sets of MR images. This enables more accurate representation of not only the lumen but also the wall surface of a TAA taking account of intraluminal thrombus.

Avaliação da aterosclerose carotídea por intermédio de ultra-sonografia e ressonância magnética

OBJETIVO: Identificar ateromas carotídeos em coronariopatas com indicação de terapia cirúrgica. Avaliar o grau de estenose das artérias carótidas internas por meio de ultra-sonografia com Doppler colorido (UDC) e angiografia por ressonância magnética (ARM). Comparar a ecogenicidade das placas visualizadas pela ultra-sonografia (US), com a intensidade de sinal nos exames de ressonância magnética (RM). Avaliar a qualidade de imagens e confiabilidade interobservadores nos exames de RM. MATERIAIS E MÉTODOS: Estudo prospectivo em 50 pacientes. Utilizamos US e RM em seqüências ponderadas em T1 e T2, ambas pelas técnicas "black-blood" (BB) e "fat sat black-blood" (FSBB), e ARM 3D TOF ("time-of-flight'') com e sem contraste paramagnético. RESULTADOS: Do total de 100 segmentos, 81% apresentaram estenose pela US. Em 72 placas com ecogenicidade tipo 4 houve aumento da intensidade de sinal em 59,7% em T1-BB, 65,3% em T1-FSBB, 62,5% em T2-BB e 66,7% em T2-FSBB. Nas placas tipo 2 houve aumento da intensidade de sinal em 71,4% em T1-BB e T1-FSBB, 85,7% em T2-BB e 71,4% em T2-FSBB. Nas placas tipo 1 houve aumento da intensidade de sinal em 50,0% em T1 e T2. Em 19 segmentos a US foi considerada normal. Quando estes segmentos foram avaliados pela RM, houve aumento da intensidade de sinal em 47,4% em T1-BB, 57,9% em T1-FSBB e 52,6% em T2-BB e T2-FSBB. CONCLUSÃO: Houve alta incidência de aterosclerose carotídea. Houve reprodutibilidade marginal na associação entre o grau de estenose visualizado pela UDC e ARM. Não houve correlação entre os tipos de ecogenicidade das placas visualizadas pela US com as alterações de intensidade de sinal pela RM. A qualidade de imagens dos exames de RM foi considerada ótima em T1 e T2 e muito boa em 3D TOF (axial). A qualidade de imagem dos exames de ARM foi excelente. Notou-se ótima reprodutibilidade interobservadores, com valores de índice kappa acima de 0,71.

Rapid extended coverage simultaneous multisection black-blood vessel wall MR imaging

A two-dimensional rapid extended coverage (REX) rapid acquisition with relaxation enhancement (RARE) pulse sequence for simultaneous multisection double inversion-recovery (DIR) black-blood vessel wall magnetic resonance (MR) imaging was developed. Aortic vessel wall MR imaging was performed in five healthy subjects (mean age, 33 years +/- 4 [SD]) and five patients with atherosclerotic disease (mean age, 67 years +/- 11.7). Shortening of blood inversion time and imaging of multiple sections after single DIR block resulted in simultaneous acquisition of up to 20 aortic wall sections in less than 1 minute (spatial resolution, 0.97 x 0.97 x 3 mm(3)). Higher signal-to-noise ratios per unit time per section (16.0 +/- 2.45 vs 7.5 +/- 1.10, P <.05), no significant changes in contrast-to-noise ratios (15.0 +/- 5.3 vs 20.1 +/- 3.9, P >.05), and 17-fold improvement in acquisition time compared with those at conventional single-section DIR RARE imaging was achieved. Use of the REX method significantly shortened aortic imaging acquisition times without degrading image quality.

Quantification of multicontrast vascular MR images with NLSnake, an active contour model: In vitro validation and in vivo evaluation

Vessel-wall measurements from multicontrast MRI provide information on plaque structure and evolution. This requires the extraction of numerous contours. In this work a contour-extraction method is proposed that uses an active contour model (NLSnake) adapted for a wide range of MR vascular images. This new method employs length normalization for the purpose of deformation computation and offers the advantages of simplified parameter tuning, fast convergence, and minimal user interaction. The model can be initialized far from the boundaries of the region to be segmented, even by only one pixel. The accuracy and reproducibility of NLSnake endoluminal contours were assessed on vascular phantom MR angiography (MRA) and high-resolution in vitro MR images of rabbit aorta. An in vivo evaluation was performed on rabbit and clinical data for both internal and external vessel-wall contours. In phantoms with 95% stenoses, NLSnake measured 94.3% +/- 3.8%, and the accuracy was even better for milder stenoses. In the images of rabbit aorta, variability between NLSnake and experts was less than interobserver variability, while the maximum intravariability of NLSnake was equal to 1.25%. In conclusion, the NLSnake technique successfully quantified the vessel lumen in multicontrast MR images using constant parameters.

Triple contrast technique for black blood imaging with double inversion preparation

This work reports on the development of a pulse sequence to simultaneously acquire proton density, T(1), and T(2) weighted images in a single magnetization prepared fast spin echo acquisition. The technique is based upon the application of a magnetization preparation consisting of a global inversion followed by slice-selective 180 degrees and 90 degrees pulses to prepare the signal of specific slices. Slices are acquired in an interleaved manner with time delays appropriate for the desired image contrasts. Data acquisition is repeated for all combinations of slice interleaving covering the region of interest until images from all slice locations have been acquired with all desired image contrasts. The multiple image contrasts obtained with this technique should be useful in applications where discrimination between different types of tissue components is desired, such as in the analysis of plaque in cervical carotid artery disease.

Reproducibility study of 3D geometrical reconstruction of the human carotid bifurcation from magnetic resonance images

The combined magnetic resonance imaging (MRI) and computational fluid dynamics (CFD) modeling approach is playing an increasingly important role in advancing our understanding of the relationship between hemodynamics and arterial disease. Nevertheless, such a modeling approach involves a number of uncertainties associated with various stages of the process. The present study is concerned with the reproducibility of geometry reconstruction, one of the most crucial steps in the modeling process. The reproducibility test was conducted on the right carotid bifurcation of eight normal human subjects, each of whom were scanned twice using the same MR protocol with an in-plane resolution of 0.625 mm. Models constructed from different scans of the same subject were compared and assessed using four quantitative measures: centerline distance, cross-sectional area, contour shape factors, and mean radius difference. The difference in the maximum carotid bulb area between the two scans was found to be <8.1% for all subjects. Shape factors (measuring the dissimilarity between two contours) of <10% were achieved in most of the common carotid arteries (CCAs) and internal carotid arteries (ICAs). The mean radius difference between the two scans was <0.4 mm for all subjects. Among the three vessels, the geometry of CCA was well reproduced by the reconstruction procedure in most of the cases, while the external carotid artery (ECA) showed the worst reproducibility. The impact of geometrical differences on CFD-predicted flow patterns and wall shear stress (WSS) will be investigated and discussed in a separate paper.

Reconstruction of carotid bifurcation hemodynamics and wall thickness using computational fluid dynamics and MRI

A thorough understanding of the relationship between local hemodynamics and plaque progression has been hindered by an inability to prospectively monitor these factors in vivo in humans. In this study a novel approach for noninvasively reconstructing artery wall thickness and local hemodynamics at the human carotid bifurcation is presented. Three-dimensional (3D) models of the lumen and wall boundaries, from which wall thickness can be measured, were reconstructed from black-blood magnetic resonance imaging (MRI). Along with time-varying inlet/outlet flow rates measured via phase contrast (PC) MRI, the lumen boundary was used as input for computational fluid dynamic (CFD) simulation of the subject-specific flow patterns and wall shear stresses (WSSs). Results from a 59-year-old subject with early, asymptomatic carotid artery disease show good agreement between simulated and measured velocities, and demonstrate a correspondence between wall thickening and low and oscillating shear at the carotid bulb. High shear at the distal internal carotid artery (ICA) was also colocalized with higher WSS; however, a quantitative general relationship between WSS and wall thickness was not found. Similar results were obtained from a 23-year-old normal subject. These findings represent the first direct comparison of hemodynamic variables and wall thickness at the carotid bifurcation of human subjects. The noninvasive nature of this image-based modeling approach makes it ideal for carrying out future prospective studies of hemodynamics and plaque development or progression in otherwise healthy subjects.