Multi-sequence MRI Carotid Artery Wall Analysis

Image segmentation and registration of multi-sequence MR vessel wall images of the carotid artery in dynamic and longitudinal studies in order to assess atherosclerosis

This research is part of the Plaque at Risk project (ParisK) which is financed by the Center for Translational Molecular Medicine (CTMM) and the Dutch Heart Foundation

Ronald van ’t Klooster, M.Sc.
Rob J. van der Geest, PhD

Background

Atherosclerosis is a progressive disease which, at an early stage, is characterized by vessel wall thickening causing outward remodeling, then narrowing of the lumen, and at a later stage by the formation of plaque lesions inside the vessel wall. In patients with unstable plaques, the thin fibrous cap can rupture causing the plaque contents to enter the vessel lumen causing a stroke. Therefore, accurate assessment of the vessel wall dimensions and composition of the vessel wall is essential for identifying patients at risk.

Magnetic Resonance Imaging (MRI) offers high-resolution non-invasive imaging of the vessel wall of the carotid artery. However, a typical MRI examination generates many images and the manual segmentation of these images is a labor-intensive, highly observer dependent and time-consuming task. Computerized segmentation techniques have been developed to overcome these limitations. These methods need further improvements to improve the accuracy, robustness and speed of automated quantitative vessel wall analysis. Also, methods have to be developed to compare studies of different points in time (e.g. baseline and two years later) and to analyze dynamic studies (e.g. to measure the uptake of contrast agent).

Goals

  • Method and validation for intra-scan registration for carotid MRI studies.
  • Technique for accurate bifurcation segmentation in carotid MR studies.
  • Validated registration technique for longitudinal carotid MRI studies.
  • Validation automated plaque characterization in carotid MRI studies.
  • Optimization of segmentation and registration techniques for 7T MRI studies.

Approach

A typical MR protocol to study the status of Atherosclerosis in the carotid artery consists of the application of multiple MR sequences to obtain information about the arterial structure of the carotid and detailed images of its vessel wall. The duration of such an exam is between 30 and 60 minutes. During that time, motion artifacts will occur by small movements of the subject inside the scanner. In order to classify and quantify atherosclerotic plaques in the vessel wall correctly aligned MR vessel wall images are required. The manual registration of these images is time-consuming.

Manual segmentations were obtained for 37 MR vessel wall studies. Each study consisted of five vessel wall sequences. The manual observer used a dedicated software package to align the images to each other by applying a translation to each image slice. Using these manual registrations the amount of movement within one MR study was investigated. The average movement of a subject in the scanner per image slice is 1.27 mm (SD 0.85 mm). Subject movement was increasing over time as shown in Figure 1.

Movement of subject within one MR study increases over time.

An example of misalignment within one study is shown in Figure 2. Automatic image registration can be applied to correct for patient movement within one study. Various registration techniques were investigated and quantitatively validated with a gold standard. The different registration methods were applied using publicly available elastix software.

Figure 2. left) T1W TSE image including lumen segmentation, middle) T1W TFE showing misalignment between T1W TSE and T1W TFE lumen segmentations, right) registered image

Next, techniques for the registration, matching, and correction of motion at the anatomical level (to support the analysis of dynamic studies) and detection and quantification of change at the anatomical level (to support the analysis of longitudinal studies) will be developed. Non-rigid registration techniques will be optimized to correct for the complex deformations that occur in dynamic and longitudinal studies. Landmarks, such as the bifurcation point, will be used to aid the registration process.

Status

Project started in 2010.

Contact

Ronald van ‘t Klooster, M.Sc.
Division of Image Processing
Department of Radiology, 1-C2S
Leiden University Medical Center
P.O. Box 9600
2300 RC Leiden
The Netherlands
Tel. +31 (0)71 526 6206
e-mail: r.van_t_klooster@lumc.nl