Synopsis

Knowledge of intracranial vessel morphology may be important in predicting the risk of acute ischemic stroke. The three-dimensional nature of the vessels would make it challenging to measure vessels' segmental lengths, unless a software tool dedicated to the purpose is available. The goal of this study is to develop a customized graphical user interface that facilitates users' measurement of intracranial vessel tortuosity in an easy and interactive manner. Using the proposed tool, vessel branch lengths and vessel tortuosity data were collected from 11 proximal vessel segments (e.g., middle cerebral artery, anterior cerebral artery) of 532 subjects.

Introduction

Methods

Imaging data were acquired using a time-of-flight (TOF) MR angiography sequence on a 3 Tesla system (Achieva, Philips Healthcare, Best, The Netherlands). Sequence parameters were as follows: repetition time = 25 ms, echo time = 3.5 ms, slice thickness = 1.2 mm, number of slices = 180, spacing between slices = 0.6 mm, number of phase encoding steps = 331, flip angle = 20-deg, reconstruction matrix = 880 x 880, field-of-view = 250 mm x 250 mm, image voxel size = 0.2841 mm x 0.2841 mm.

TOF images underwent cubic interpolations in the slice direction to result in isotropic voxel size of 0.2841 mm in the interpolated TOF images. Three seed points were provided in a left middle cerebral artery (MCA), a right MCA, and a basilar artery, respectively, and 3D seeded region growing was performed to generate 3D binary masks of the major intracranial arteries.

A custom graphical user interface (GUI) was implemented in a MATLAB environment (see Figure 1). Intracranial vessels were segmented from MR angiography data. The centerlines of intracranial vessels were obtained after applying skeletonization to the binary format of the vessels. The GUI allowed the user to locate and click the two landmarks, which correspond to the start and end points of a vessel segment of interest. The user’s mouse-click of the “Compute vessel length” button resulted in automatic calculation of the branch length as well as Euclidean distance between the start and end points (shown in Figure 1). The centerline tracking from the start point to the end point was automatic, based on a directional cosine measure. The yellow arrow in Figure 1 indicates the visualization of a selected vessel segment. The vessel’s tortuosity was calculated as the ratio of the branch length to the Euclidean distance between the two points. Two modes existed for the extraction of vessel segments: (a) maximum mode = 64 vessel segments, and (b) minimum mode = 11 vessel segments. Figure 2 summarizes the image preprocessing and tortuosity calculation pipelines.

Results

Using the proposed user interface software, we obtained Euclidean vessel length, branch vessel length, and vessel tortuosity data of 11 vessel segments from 532 normal volunteers. The processing time for obtaining vessel tortuosity from all the vessel segments took 16-20 minutes per subject in the maximum mode.The vessel skeletonization procedure took approximately 4 minutes, requiring an increase of computational efficiency.

Table 1 lists the correlation coefficients between vessel segment’s tortuosity and age/Framingham_risk_score. Basilar artery, middle cerebral artery, and posterior cerebral artery P2 showed statistical significance (p-value < 0.01), with correlation coefficients ranging from 0.120 to 0.264. These features may be combined for machine learning to increase the performance in prediction of a vessel’s age and Framingham risk score.

Discussion

Conclusion

Acknowledgements

References

1. Bullitt, Elizabeth, et al. "Measuring tortuosity of the intracerebral vasculature from MRA images." IEEE transactions on medical imaging 22.9 (2003): 1163-1171.

2. Dunås, Tora, et al. "Automatic labeling of cerebral arteries in magnetic resonance angiography." Magnetic Resonance Materials in Physics, Biology and Medicine 29.1 (2016): 39-47.