Abstract
High arterial tortuosity may signify early arterial pathology which may precede development of intracranial aneurysms. We measured arterial tortuosity of intracranial vessels and reviewed the medical records of three groups of patients: with intracranial aneurysms, without aneurysms but at increased clinical risk, and controls without aneurysms or associated risk factors. There was significant but inconsistent evidence of increased arterial tortuosity in aneurysm cases and high-risk cases across different arteries. Medical records review identified that a subset of aneurysm cases carried a diagnosis of Loeys-Dietz syndrome that is often misdiagnosed as Marfan syndrome. We found increased arterial tortuosity in the Loeys-Dietz syndrome cases. A combination of medical record screening for Marfan syndrome or Loeys-Dietz symptoms such as aneurysms and evaluation of arterial tortuosity by a curve of scores from medical images may identify previously undiagnosed cases of Loeys-Dietz syndrome.
Keywords: Aneurysm, MRA, Loeys-Dietz, tortuosity
Introduction
There are autopsy reports of increased arterial and arteriolar tortuosity in aged subjects with hypertension and aneurysms1. While hypertension is considered a risk factor for developing intracranial aneurysms, the clinical importance of the greater arteriolar tortuosity noted on autopsy2,3, to the development of intracranial aneurysms is not clear. The small arterioles visible on dissection at autopsy are not readily visible using current medical imaging techniques but Time of Flight-Magnetic Resonance Angiography (TOF-MRA) imaging has been used to assess arterial tortuosity of larger vessels. Increased tortuosity of arteries visible in MRA has been shown to correlate with aging4, exercise level5, tumors6, retinal pathology7 and certain genetic syndromes8,9. The degree of arterial tortuosity can be quantitatively measured from MRA images with the Distance Factor Metric (DFM) tortuosity score that calculated by measuring the length (L) along the centerline of the artery divided by the straight line distance (d) from two points10–13. In the conventional use of DFM, only two points are selected per artery, producing a single tortuosity score or zero-dimensional measure of tortuosity (DFM0) for the artery. Whether used for inter-subject or intra-subject measurements, DFM0 is constrained by the underlying data: images may not consistently contain the same two well defined points along the artery, images may contain different lengths of the artery, and the tortuosity may be sensitively dependent upon the selection of the two points. This study expands the conventional use of DFM to create a one-dimensional tortuosity score curve (DFMc) displaying local tortuosity information along a vessel. The DFMc is then used to assess the relationship between arterial tortuosity of larger vessels seen in TOF-MRA with the development of intracranial aneurysms.
This study utilized medical imaging to assess the degree of arterial tortuosity noted in patients with a clinical history of aneurysm, or predisposition to aneurysm14–16, recorded in the medical record. We were specifically interested in determining whether or not patients with familial aneurysms, non-familial aneurysms or in high-risk subjects without any history of aneurysms as yet, have abnormally increased arterial tortuosity.
Materials and methods
Source Images
All TOF-MRA images were collected from the University of Utah Medical Center in Salt Lake City, UT, U.S.A. with approval from the University of Utah Institutional Review Board. A negative control population was collected retrospectively from clinical TOF-MRA head images taken within the last three years. The negative control population included patients with a diagnosis of headache or trigeminal neuralgia who underwent TOF-MRA head imaging but in whom no vascular disease (aneurysmal dilation or stenosis) was identified in the radiology report and in whom no risk factors for vascular disease were noted in the medical record (including: arterial disease, atrial fibrillation, diabetes, hypertension, and acquired heart disease). The control group also had cancer or genetic syndromes screened out. The high-risk group and aneurysm group were comprised of cases previously identified in a study on high (two-fold) familial risk of intracranial aneurysms14 and from patients treated for aneurysms at the University of Utah Medical Center. The images were clinical scans at a range of resolutions. Lower resolution images were interpolated to higher resolution with a sinc interpolation for comparison to higher resolution images.
Arterial tortuosity measurement summary
The arterial tortuosity measures were made by segmenting the arteries from the background, generating a centerline through the segmentation and selecting two end points along the centerline of the artery measured. The Distance Factor Metric (DFM) tortuosity score is calculated by measuring L along the centerline of the artery divided by d from the starting point10–13. Rather than compute a single DFM tortuosity score per artery (DFM0), the L/d tortuosity was calculated at every point on the centerline with respect to the starting point to create a one-dimensional tortuosity score curve (DFMc). A tortuosity measure based upon a smoothed centerline was also calculated by averaging the position of each centerline point with its two adjacent neighbors to compute a smoothed version of L (Ls) and the smoothed tortuosity score, DFMcs = Ls/d. After computing the DFMc curve, an optimal point along each artery was selected for reporting DFM. In this study, the final DFM tortuosity scores were taken as either the peak DFM value of the DFMc curve or the end DFM value of the DFMc curve when the arterial centerline left the image volume and no defined second end existed.
Tools
The segmentation and tortuosity measurement tools were implemented as ImageJ plugins17–19. The measured centerline positions and subject information were stored in a MySQL relational database available at http://www.mysql.com/. Plotting of tortuosity score curves, box and whiskers plot comparisons, and statistical tests were conducted with R20, connected directly to MySQL using MySQL Open Database Connectivity (ODBC), using statistical methods previously described21.
Statistical tests
Statistical significance was set at the α = 0.05 level and Bonferroni corrected to β = α/n, where n is the number of tests in a set. The significance level was adjusted instead of the P-value of statistical tests to show raw test results. The Wilcoxon rank sum test was used throughout the study as it does not require normality of the underlying populations and is resistant to outliers. F-test and T-test were used with larger sample set sizes where normality can be assumed by the central limit theorem. The F-test tested for differences in variation of tortuosity scores and the T-test was used to confirm the Wilcoxon rank sum test results.
Segmentation
The TOF-MRA images (Figure 1) were segmented using a Maximum Intensity Projection (MIP) Z-buffer segmentation22(Figure 2).
Figure 1.
Time of Flight-Magnetic Resonance Angiography (TOF-MRA) medical image shown in Maximum Intensity Projection (MIP).
Figure 2.
Segmented arteries from TOF-MRA with color changes at bifurcations in shaded surface display26.
Centerlines
The centerlines were generated from the segmentations using a centerline algorithm based on algorithms previously described23,24 with a cost function modification where the Center of Mass (COM) voxel costs were multiplied by the Distance From Edge (DFE) values of the voxels to give higher weights to voxels at the center of the segmented arteries. Due to the limitations of intensity based segmentation, the internal carotid artery often segments as a closed loop structure, presenting two apparent paths during centerline extraction. The ability of the centerline algorithm employed here to extract the proper geometry of the internal carotid artery was validated using the centerline stability metric previously developed24.
Artery selection and tortuosity measurement
Arteries were selected for tortuosity measurement by first selecting two end-points of a centerline through the segmentation of the desired artery. Unlike a traditional DFM0 measurement, in this case it was only necessary that one of the two end-points be a common anatomical location for each measured artery, generally a bifurcation shared by all subjects. The second end-point could either be another common anatomical location or the point at which the artery of interest exited the image volume. The three-dimensional segmented artery image was colored to assist the user in selecting centerline segments for tortuosity measurement. The red centerlines connect at the bifurcations which are indicated by green dots (Figure 3). A separate random color was assigned to each centerline to cause a color change at the bifurcations to aid the user when locating bifurcations.
Figure 3.
Selection in white of anterior cerebral artery (ACA) shown in Maximum Intensity Projection (MIP).
Visual correlation
The quantitative DFM tortuosity scores were correlated to visual tortuosity rankings. A total of 315 subjects including negative controls and vascular disease cases from multiple ongoing tortuosity studies were ordered highest to lowest by the DFMcs based tortuosity score at the end of the basilar artery where it bifurcates into the left and right vertebral arteries (Figure 4). Every 11th subject beginning with the subject with the highest tortuosity score was selected to obtain a subset of 25 subjects with a wide range of arterial tortuosities. For each of the 25 subjects, MIP images were computed in the transverse plane with 18 rotations taken every 10 degrees (MIP images at 180 degrees difference are the same) showing the entire brain vasculature imaged. These images were shown to a group of five volunteer medical imaging researchers who ranked the basilar arteries highest (rank 1) to lowest (rank 25) in tortuosity. The volunteers were advised to compare images pairwise and were given no time limit or consistency training to avoid bias25. The means of the human rankings were compared by Spearman rank correlation to the rank determined by the quantitative end of artery DFMc and DFMcs scores of the basilar artery.
Figure 4.
Histogram of basilar artery end DFMcs (labeled end DFM3) tortuosity scores.
Tortuosity was measured for multiple arteries in the image slabs. The arteries measured were the left and right anterior cerebral arteries (ACA), across the left ACA through the anterior communicating artery (Acomm) to the right ACA; the basilar artery, the left and right internal carotid arteries (ICA) from the ICA bifurcation with the middle cerebral artery (MCA) and ACA to the lower end of the image slab, and the left and right vertebral arteries from the bifurcation with the basilar artery to the lower end of the image slab. The tortuosity score was taken from the tortuosity score curve at the end of the curve for the basilar, anterior cerebral artery (ACA) and ACA-anterior communicating artery (AComm)-ACA measurements. The tortuosity score was taken at the peak of the curve for the internal carotid artery (ICA) and vertebral artery (VA) measurements because the end of these arteries varied depending on the depth of the image slabs.
Results
The end DFMc had higher correlation to the mean visual rankings than the smoothed end DFMcs tortuosity score. The end DFMc had a 0.72 Spearmen rank correlation coefficient (P < 0.0001) (Figure 5) with the mean visual ranking while the end DFMcs correlation was 0.67 (P = 0.00025). The mean of the correlation between all pairs of human visual ranks was 0.88±0.048. Both the end DFMc and end DFMcs quantitative tortuosity scores were calculated and used in statistical tests of differences between the test cases and negative controls. Only the end DFMc scores are reported due to better correlation with the mean visual ranks and due to the fact that the results of the tests differed little with the two measurements.
Figure 5.
Comparison of mean visual rank (x-axis) versus the rank of the end DFMc tortuosity score (y-axis) of the basilar artery with regression line (0.72 Spearmen rank correlation coefficient).
We measured the arterial tortuosity of eight arteries between the intracranial aneurysm group and the negative control population. The difference in tortuosity was tested at the β = α/n = 0.05/8 = 0.00625 level to account for testing eight arteries. Only the left ACA tortuosity measurement was noted to be significantly greater in the aneurysm cohort (indicated in bold in Table 1). While the aneurysm group had greater tortuosity in all eight arteries (indicated with a + in the “Difference of means” column in Table 1), the difference from the control group was not statistically significant for the other seven arteries. The aneurysm cases also had significantly higher variance in right ICA and left VA than the negative controls. Of note, the aneurysm population was approximately eight years older than the negative control group. As the data set was obtained from existing images taken for other purposes, the images often included different arteries and in some instances image quality prevented measurement of some arteries resulting in different number (N) of measurements for each artery recorded in Table 1.
Table 1.
Aneurysm cases versus negative control tortuosity comparisons.
| Artery | Measurement | Negative (N) | Mean resolution (mm) | Mean DFM ± SD | Mean age | Aneurysm (N) | Mean resolution (mm) | Mean DFM ± SD | Mean age | % Difference of means | 1-sided Wilcoxon Test | 2-sided F Test | 1-sided T test |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Left ACA | End | 43 | 0.263x0.263x0.606 | 1.375±0.110 | 46.674 | 64 | 0.208x0.208x0.432 | 1.462±0.141 | 54.822 | +6.3 | 0.00054 | 0.088 | 0.00028 |
| Right ACA | End | 39 | 0.268x0.268x0.594 | 1.450±0.135 | 45.641 | 65 | 0.207x0.207x0.430 | 1.497±0.159 | 54.099 | +3.2 | 0.079 | 0.289 | 0.058 |
| L-R ACA | End | 24 | 0.268x0.268x0.613 | 1.695±0.176 | 41.375 | 32 | 0.214x0.214x0.425 | 1.715±0.184 | 52.944 | +1.2 | 0.320 | 0.819 | 0.336 |
| Basilar | End | 42 | 0.261x0.261x0.613 | 1.196±0.093 | 46.333 | 56 | 0.204x0.204x0.441 | 1.216±0.105 | 53.900 | +1.7 | 0.157 | 0.408 | 0.161 |
| Left ICA | Peak | 35 | 0.260x0.260x0.615 | 3.157±0.737 | 48.343 | 55 | 0.207x0.207x0.423 | 3.522±1.053 | 55.831 | +11.5 | 0.097 | 0.028 | 0.028 |
| Right ICA | Peak | 36 | 0.271x0.271x0.593 | 2.941±0.521 | 47.056 | 51 | 0.204x0.204x0.421 | 3.341±1.075 | 55.940 | +13.6 | 0.078 | <0.0001 | 0.012 |
| Left VA | Peak | 36 | 0.265x0.265x0.607 | 1.350±0.155 | 46.000 | 49 | 0.208x0.208x0.465 | 1.448±0.318 | 54.105 | +7.3 | 0.043 | <0.0001 | 0.032 |
| Right VA | Peak | 35 | 0.259x0.259x0.610 | 1.340±0.142 | 46.400 | 50 | 0.209x0.209x0.465 | 1.352±0.176 | 53.850 | +0.85 | 0.431 | 0.196 | 0.372 |
The negative controls who were < 40 or > 55 years produced no statistically significant differences at β = 0.00625 with the 1-sided Wilcoxon signed rank test (used exclusively due to the small sample size) across the eight arteries. Three arteries had higher tortuosity in the > 55 population (+) and five arteries had higher tortuosity in the < 40 population (−) (Table 2).
Table 2.
Negative control age comparisons.
| Artery | Measurement | Under 40 (N) | Mean resolution (mm) | Mean DFM ± SD | Mean age | Over 55 (N) | Mean resolution (mm) | Mean DFM ± SD | Mean age | % Difference of means | 1-sided Wilcoxon Test |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Left ACA | End | 16 | 0.270x0.270x0.586 | 1.379±0.120 | 27.938 | 16 | 0.263x0.263x0.609 | 1.368±0.106 | 65.562 | −0.80 | 0.566 |
| Right ACA | End | 16 | 0.270x0.270x0.586 | 1.449±0.142 | 27.938 | 15 | 0.268x0.268x0.596 | 1.435±0.148 | 64.333 | −1.0 | 0.727 |
| L-R ACA | End | 12 | 0.266x0.266x0.605 | 1.699±0.150 | 26.000 | 6 | 0.286x0.286x0.620 | 1.708±0.223 | 68.000 | +0.51 | 0.625 |
| Basilar | End | 16 | 0.270x0.270x0.586 | 1.176±0.089 | 27.938 | 15 | 0.263x0.263x0.616 | 1.166±0.075 | 65.867 | −0.88 | 0.673 |
| Left ICA | Peak | 12 | 0.259x0.259x0.595 | 2.796±0.556 | 28.250 | 14 | 0.264x0.264x0.624 | 3.438±0.699 | 66.500 | +23.0 | 0.013 |
| Right ICA | Peak | 13 | 0.282x0.282x0.560 | 2.849±0.380 | 28.000 | 14 | 0.268x0.268x0.603 | 3.010±0.534 | 64.857 | +5.7 | 0.229 |
| Left VA | Peak | 14 | 0.276x0.276x0.577 | 1.339±0.154 | 28.357 | 13 | 0.269x0.269x0.611 | 1.318±0.104 | 64.615 | −0.46 | 0.215 |
| Right VA | Peak | 14 | 0.271x0.271x0.599 | 1.324±0.180 | 29.286 | 13 | 0.254x0.254x0.611 | 1.312±0.097 | 65.231 | −2.1 | 0.547 |
The familial aneurysm cases had significantly higher tortuosity of the left ACA than the negative controls at the β = 0.00625 level. Eight of the eight arteries had higher arterial tortuosity (+) in the familial aneurysm cases than the negative controls but seven were not significant (Table 3).
Table 3.
Familial aneurysm cases versus negative control tortuosity comparisons.
| Artery | Familial Aneurysm (N) | Mean resolution (mm) | Mean DFM ± SD | Mean age | % Difference of means | 1-sided Wilcoxon Test |
|---|---|---|---|---|---|---|
| Left ACA | 23 | 0.201x0.201x0.407 | 1.474±0.127 | 58.610 | +7.2 | 0.00063 |
| Right ACA | 22 | 0.202x0.202x0.406 | 1.535±0.136 | 56.250 | +5.8 | 0.019 |
| L-R ACA | 12 | 0.203x0.203x0.400 | 1.708±0.168 | 55.067 | +0.82 | 0.389 |
| Basilar | 18 | 0.195x0.195x0.403 | 1.217±0.096 | 57.571 | +1.8 | 0.183 |
| Left ICA | 20 | 0.202x0.202x0.401 | 3.345±0.685 | 59.654 | +6.0 | 0.220 |
| Right ICA | 20 | 0.202x0.202x0.403 | 3.171±0.843 | 58.568 | +7.8 | 0.268 |
| Left VA | 21 | 0.207x0.207x0.467 | 1.432±0.234 | 58.603 | +6.1 | 0.087 |
| Right VA | 22 | 0.207x0.207x0.467 | 1.366±0.184 | 58.485 | +2.0 | 0.357 |
A manual medical record chart review of the highest scoring tortuosity measures of intracranial aneurysms cases revealed one diagnosis of Marfan syndrome (without genetic confirmation), two of Loeys-Dietz syndrome (LDS) (with genetic confirmation) and eight high familial risk intracranial aneurysm (IA) cases. Before 2005 LDS cases were often diagnosed with Marfan syndrome making diagnosis without genetic confirmation ambiguous. Further chart review demonstrated that the one patient diagnosed with Marfan syndrome did not meet clinical criteria for this diagnosis raising the question of a misdiagnosis in a patient with a TGF-β LDS causing mutation.
Additional LDS cases were collected to test for an increase in arterial tortuosity in LDS patients. Six syndromic cases including five genetically confirmed Loeys-Dietz syndrome (LDS) and the one unconfirmed clinical Marfan diagnosis had significantly greater tortuosity of the basilar and the left VA at the β = 0.00625 level (Table 4). Examples of tortuous vertebral arteries of Loeys-Dietz patients are shown in Figure 6 and Figure 7. For comparison Figure 8 shows a low tortuosity VA and Figure 9 shows the tortuosity curves of those arteries. These patients had greater tortuosity of eight of eight arteries measured. Two of the confirmed LDS patients and the one unconfirmed clinical Marfan syndrome cases had intracranial aneurysms and the other three confirmed LDS cases did not have aneurysms.
Table 4.
Loeys-Dietz/Marfan syndrome cases versus negative control tortuosity comparisons.
| Artery | Syndrome (N) | Mean resolution (mm) | Mean DFM ± SD | Mean age | % Difference of means | 1-sided Wilcoxon Test |
|---|---|---|---|---|---|---|
| Left ACA | 5 | 0.229x0.229x0.380 | 1.407±0.168 | 19.200 | +2.3 | 0.474 |
| Right ACA | 4 | 0.229x0.229x0.387 | 1.704±0.448 | 20.000 | +17.5 | 0.131 |
| L-R ACA | 4 | 0.229x0.229x0.387 | 2.143±0.709 | 20.000 | +26.5 | 0.063 |
| Basilar | 5 | 0.211x0.211x0.375 | 1.443±0.300 | 25.300 | +20.6 | 0.0045 |
| Left ICA | 3 | 0.229x0.229x0.350 | 3.361±0.388 | 26.833 | +6.5 | 0.323 |
| Right ICA | 3 | 0.229x0.229x0.350 | 3.263±0.688 | 26.833 | +10.9 | 0.216 |
| Left VA | 6 | 0.220x0.220x0.371 | 1.931±0.652 | 23.750 | +43.1 | 0.00043 |
| Right VA | 6 | 0.220x0.220x0.371 | 1.511±0.297 | 23.750 | +12.7 | 0.051 |
Figure 6.
Distance Factor Metric (DFM) = Length (L) / distance (d) tortuosity scores of the left vertebral artery of a suspected Loeys-Dietz patient selected in white. Yellow lines show d and progressive steps. This subject had the maximum left vertebral artery tortuosity of the aneurysm subjects. Black line in Figure 9.
Figure 7.
Loeys-Dietz syndrome intracranial aneurysm subject with median tortuosity among aneurysm case subjects of the left vertebral artery and high tortuosity of the basilar artery. Green line in Figure 9.
Figure 8.
Non-familial intracranial aneurysm subject with low tortuosity left vertebral artery. Red line in Figure 9.
Figure 9.
Vertebral artery tortuosity score curves of the highest peak tortuosity score (black line), median tortuosity score (green line) and low tortuosity score (red line). The peak score is taken between the dotted lines to avoid small variations causing spikes when the lengths L and d are short and subject to noise and before vertebral arteries twist around the first cervical vertebrae.
Patients with non-familial aneurysms and without an underlying genetic syndrome had significantly greater left ACA arterial tortuosity by t-test than negative controls. Tortuosity measures were greater in seven of eight arteries. Non-familial aneurysm cases also had significantly higher variation in the left and right ICAs (Table 5).
Table 5.
Non-familial aneurysms versus negative control tortuosity comparisons.
| Artery | Non-familial aneurysm | Mean resolution (mm) | Mean DFM ± SD | Mean age | % Difference of means | 1-sided Wilcoxon Test | 2-sided F Test | 1-sided T Test |
|---|---|---|---|---|---|---|---|---|
| Left ACA | 42 | 0.214x0.214x0.454 | 1.454±0.150 | 54.751 | +5.7 | 0.0069 | 0.047 | 0.0036 |
| Right ACA | 44 | 0.213x0.213x0.454 | 1.471±0.167 | 54.953 | +1.4 | 0.330 | 0.193 | 0.270 |
| L-R ACA | 19 | 0.224x0.224x0.445 | 1.725±0.202 | 54.596 | +1.8 | 0.310 | 0.526 | 0.303 |
| Basilar | 38 | 0.212x0.212x0.475 | 1.207±0.107 | 55.789 | +0.89 | 0.328 | 0.360 | 0.318 |
| Left ICA | 34 | 0.211x0.211x0.447 | 3.577±1.262 | 54.710 | +13.3 | 0.165 | 0.0024 | 0.049 |
| Right ICA | 30 | 0.207x0.207x0.442 | 3.450±1.216 | 56.703 | +17.3 | 0.059 | <0.0001 | 0.020 |
| Left VA | 27 | 0.215x0.215x0.478 | 1.395±0.155 | 54.333 | +3.3 | 0.105 | 0.980 | 0.128 |
| Right VA | 27 | 0.217x0.217x0.476 | 1.317±0.097 | 53.790 | −1.7 | 0.562 | 0.050 | 0.773 |
High intracranial aneurysm risk family case subjects without aneurysms themselves had significantly higher arterial tortuosity in the left ICA and significantly higher variance in the left VA. These case subjects had higher tortuosity in seven of the eight arteries measured (Table 6).
Table 6.
High-risk relative cases versus negative control tortuosity comparisons.
| Artery | High-risk (N) | Mean resolution (mm) | Mean DFM ± SD | Mean age | % Difference of means | 1-sided Wilcoxon Test | 2-sided F Test | 1-sided T test |
|---|---|---|---|---|---|---|---|---|
| Left ACA | 53 | 0.217x0.217x0.486 | 1.423±0.111 | 46.604 | +3.5 | 0.012 | 0.921 | 0.019 |
| Right ACA | 52 | 0.218x0.218x0.486 | 1.494±0.131 | 46.577 | +3.0 | 0.076 | 0.805 | 0.061 |
| L-R ACA | 25 | 0.221x0.221x0.484 | 1.706±0.149 | 50.000 | +0.68 | 0.241 | 0.421 | 0.403 |
| Basilar | 31 | 0.215x0.215x0.481 | 1.182±0.082 | 47.387 | −1.2 | 0.724 | 0.490 | 0.724 |
| Left ICA | 37 | 0.215x0.215x0.488 | 3.850±1.053 | 46.405 | +21.9 | 0.0023 | 0.039 | 0.00092 |
| Right ICA | 36 | 0.215x0.215x0.482 | 3.330±0.781 | 45.639 | +13.2 | 0.028 | 0.019 | 0.0079 |
| Left VA | 47 | 0.215x0.215x0.482 | 1.423±0.254 | 48.234 | +5.4 | 0.087 | 0.0030 | 0.054 |
| Right VA | 43 | 0.215x0.215x0.479 | 1.404±0.189 | 47.302 | +4.8 | 0.086 | 0.087 | 0.046 |
Discussion
Measurement of arterial tortuosity is a newly developed technique that may prove to be of clinical utility in identifying diseased vasculature. The DFMc tortuosity score curve and associated peak and end measurements described herein appear to provide more information than the traditional single value DFM0 tortuosity score. By selecting the peak tortuosity score from a curve of values defined from a single end-point, we obtain a meaningful tortuosity value from arteries with only one well-defined end-point in a medical image. The original DFM0 method required selection of the same two defined end-points for all arteries to be compared, making it unusable when there was only one defined point as is often the case with the long ICA and vertebral arteries. Furthermore, analysis of the DFMc tortuosity score curves shows that the tortuosity score may vary significantly along the vessel as indicated in Figure 9. Thus, by considering only two particular end-points per artery, the traditional tortuosity analysis may both greatly underestimate the peak value and be sensitively dependent on end-point selection.
Using the methods described herein, we have been able to demonstrate a significantly greater degree of arterial tortuosity in patients with connective tissue syndromes who are known to be at risk for intracranial aneurysms8,27. Non-syndromic patients with intracranial aneurysms, subjects with high familial risk of intracranial aneurysms, relatives of high-risk aneurysm cases, and patients with non-familial aneurysms had inconsistently higher arterial tortuosity than negative controls. The overlap in tortuosity scores between high familial risk intracranial aneurysm cases with negative controls indicates high-risk subjects with normal tortuosity scores can develop intracranial aneurysms. There was also no significant difference between high-risk subjects with aneurysms compared to relatives without aneurysms. In contrast, arterial tortuosity in patients with Loeys-Dietz syndrome, a disorder associated with the presence of intracranial aneurysms, was significantly different than negative controls.
Age has been shown to mildly increase tortuosity in healthy populations4. The age comparisons conducted here showed no significant tortuosity increase due to age. It is thus unlikely that the age difference between the aneurysm and negative control populations accounted for the differences in tortuosity.
The human visual rankings correlation to each other was closer than to the quantitative tortuosity score based ranks. This phenomenon of humans correlating with each other better than a computer score has been previously described25. The human rankers may be using information seen in the surrounding image or alternatively, there could be a bias in the projections shown to the rankers causing their ranks to cluster together.
The results of this study provide evidence that tortuosity measurements may be able to assist in characterizing specific non-normal states and may even assist in distinguishing between patients with Loeys-Dietz syndrome and Marfan syndrome. Loeys-Dietz syndrome is a more aggressive disorder associated with visible arterial tortuosity and aneurysms throughout the arterial tree8,27 where many but not all affected patients will go on to develop cerebral aneurysms. In light of the fact that Loeys-Dietz syndrome was only determined to be a unique clinical entity apart from Marfan syndrome within the last decade28, many affected patients may still carry the diagnosis of Marfan syndrome. Marfan syndrome is caused by a mutation in the FBN1 gene that encodes for the glycoprotein fibrillin29. Patients are typically followed with only echocardiographic imaging of the ascending aorta as the remainder of the arterial vessels are not thought to be at significant risk of aneurysm formation. In contrast, Loeys-Dietz syndrome is caused by mutations in the TGFBRI and TGFBRII genes which encode for receptors for the cytokine TGF-β27,28. By collecting and measuring arterial tortuosity data in patients with either clinical diagnosis we hope to be able to distinguish between the two disorders and determine which patients with Loeys-Dietz syndrome are at greatest risk for cerebral aneurysm formation. Initial review of arterial tortuosity in Loeys-Dietz patients demonstrated that these patients may have the greatest increase in tortuosity in the extra-cranial vertebral arteries which are typically more caudal than the sections analyzed in this study. Assessment of both the cervicocephalic vessels and intracranial vessels may prove valuable 30.
This study demonstrates the potential to combine medical record screening with automated image analysis to screen patient data. The study started with familial and non-familial intracranial aneurysm cases and discovered the syndromic patients during the course of the research. The method for measuring arterial tortuosity is now semiautomated. Future development will further automate the tortuosity measurement. Automated medical record screening systems already exist. In this case the two methods of medical record and image screening could be combined to look for patients with diagnosis of Marfan syndrome or other Loeys-Dietz symptoms and high arterial tortuosity to identify undiagnosed Loeys-Dietz patients in electronic medical records.
Acknowledgments
We greatly appreciate the help of the staff at the Utah Center for Advanced Imaging Research in supporting this research. The research including database collection supported by grants from the Ben B. and Iris M. Margolis Foundation, NLM training grant T15LM007124, and NIH grants: R01-NS-37737 and R01-HL-48223.
References
- 1.Dobrin PB, Schwarcz TH, Baker WH. Mechanisms of arterial and aneurysmal tortuosity. Surgery. 1988 Sep;104(3):568–571. [PubMed] [Google Scholar]
- 2.Spangler KM, Challa VR, Moody DM, Bell MA. Arteriolar tortuosity of the white matter in aging and hypertension. A microradiographic study. J. Neuropathol. Exp. Neurol. 1994 Jan;53(1):22–26. doi: 10.1097/00005072-199401000-00003. [DOI] [PubMed] [Google Scholar]
- 3.Hiroki M, Miyashita K, Oda M. Tortuosity of the white matter medullary arterioles is related to the severity of hypertension. Cerebrovasc. Dis. 2002;13(4):242–250. doi: 10.1159/000057850. [DOI] [PubMed] [Google Scholar]
- 4.Bullitt E, Zeng D, Mortamet B, Ghosh A, Aylward SR, Lin W, et al. The effects of healthy aging on intracerebral blood vessels visualized by magnetic resonance angiography. Neurobiol. 2008. May 7, Aging [Internet] [cited 2009 Nov 20];Available from: http://www.ncbi.nlm.nih.gov/pubmed/18471935. [DOI] [PMC free article] [PubMed]
- 5.Bullitt E, Rahman FN, Smith JK, Kim E, Zeng D, Katz LM, et al. The Effect of Exercise on the Cerebral Vasculature of Healthy Aged Subjects as Visualized by MR Angiography. AJNR Am J Neuroradiol [Internet] 2009. Jul 9, [cited 2009 Oct 2];Available from: http://www.ncbi.nlm.nih.gov/pubmed/19589885. [DOI] [PMC free article] [PubMed]
- 6.Bullitt E, Reardon DA, Smith JK. A review of micro- and macrovascular analyses in the assessment of tumor-associated vasculature as visualized by MR. Neuroimage. 2007;37(Suppl 1):S116–9. doi: 10.1016/j.neuroimage.2007.03.067. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Dougherty G, Johnson MJ, Wiers MD. Measurement of retinal vascular tortuosity and its application to retinal pathologies. Med Biol Eng Comput. 2010 Jan;48(1):87–95. doi: 10.1007/s11517-009-0559-4. [DOI] [PubMed] [Google Scholar]
- 8.Johnson PT, Chen JK, Loeys BL, Dietz HC, Fishman EK. Loeys-Dietz syndrome: MDCT angiography findings. AJR Am J Roentgenol. 2007 Jul;189(1):W29–35. doi: 10.2214/AJR.06.1316. [DOI] [PubMed] [Google Scholar]
- 9.Ritelli M, Drera B, Vicchio M, Puppini G, Biban P, Pilati M, et al. Arterial tortuosity syndrome in two Italian paediatric patients. Orphanet J Rare Dis. 2009;4:20. doi: 10.1186/1750-1172-4-20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.O’Flynn P, O-Sullivan G, Pandit A. Methods for three-dimensional geometric characterization of the arterial vasculature. [cited 2009 Mar 9];Available from: http://www.ncbi.nlm.nih.gov/pubmed/17431787?ordinalpos=8&itool=EntrezSystem2.PEntrez.Pubmed.Pubmed_ResultsPanel.Pubmed_DefaultReportPanel.Pubmed_RVDocSum. [DOI] [PubMed]
- 11.Bullitt E, Gerig G, Pizer SM, Lin W, Aylward SR. Measuring tortuosity of the intracerebral vasculature from MRA images. IEEE Trans Med Imaging. 2003 Sep;22(9):1163–1171. doi: 10.1109/TMI.2003.816964. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Capowski JJ, Kylstra JA, Freedman SF. A numeric index based on spatial frequency for the tortuosity of retinal vessels and its application to plus disease in retinopathy of prematurity. Retina (Philadelphia, Pa.) 1995;15(6):490–500. doi: 10.1097/00006982-199515060-00006. [DOI] [PubMed] [Google Scholar]
- 13.Lotmar W, Freiburghaus A, Bracher D. Measurement of vessel tortuosity on fundus photographs. Albrecht Von Graefes Arch Klin Exp Ophthalmol. 1979 Jul 2;211(1):49–57. doi: 10.1007/BF00414653. [DOI] [PubMed] [Google Scholar]
- 14.Farnham JM, Camp NJ, Neuhausen SL, Tsuruda J, Parker D, MacDonald J, et al. Confirmation of chromosome 7q11 locus for predisposition to intracranial aneurysm. Hum Genet. 2004 Feb;114(3):250–5. doi: 10.1007/s00439-003-1044-z. [DOI] [PubMed] [Google Scholar]
- 15.Cannon Albright LA, Camp NJ, Farnham JM, MacDonald J, Abtin K, Rowe KG. A genealogical assessment of heritable predisposition to aneurysms. J Neurosurg. 2003 Oct;99(4):637–643. doi: 10.3171/jns.2003.99.4.0637. [DOI] [PubMed] [Google Scholar]
- 16.Berthelemy-Okazaki N, Zhao Y, Yang Z, Camp NJ, Farnham J, Parker D, et al. Examination of ELN as a candidate gene in the Utah intracranial aneurysm pedigrees. Stroke. 2005 Jun;36(6):1283–1284. doi: 10.1161/01.STR.0000166198.05439.f8. [DOI] [PubMed] [Google Scholar]
- 17.Burger W, Burge MJ. Digital Image Processing: An Algorithmic Introduction using Java. [Internet]. [cited 2009 Mar 7]. Available from: http://www.imagingbook.com/
- 18.Abramoff MD, Magelhaes PJ, Ram SJ. Image Processing with ImageJ. Biophotonics International. 2004;11(7):36–42. [Google Scholar]
- 19.Rasband W. Bethesda, Maryland, USA: 1997. ImageJ, U.S. National Institutes of Health. [Internet]. 2005 [cited 2009 Mar 8];Available from: http://rsb.info.nih.gov/ij/ [Google Scholar]
- 20.R Development Core Team . R: A language and environment for statistical computing. R Foundation for Statistical Computing; Vienna, Austria: 2009. [Internet] [cited 2010 Apr 23];Available from: http://www.R-project.org/ [Google Scholar]
- 21.Crawley MJ. The R Book. 1st ed. Wiley; 2007. [Google Scholar]
- 22.Parker DL, Chapman BE, Roberts JA, Alexander AL, Tsuruda JS. Enhanced image detail using continuity in the MIP Z-buffer: applications to magnetic resonance angiography. J Magn Reson Imaging. 2000 Apr;11(4):378–88. doi: 10.1002/(sici)1522-2586(200004)11:4<378::aid-jmri5>3.0.co;2-#. [DOI] [PubMed] [Google Scholar]
- 23.Zhang L, Chapman BE, Parker DL, Roberts JA, Guo J, Vemuri P, et al. Automatic detection of three-dimensional vascular tree centerlines and bifurcations in high-resolution magnetic resonance angiography. Invest Radiol. 2005 Oct;40(10):661–71. doi: 10.1097/01.rli.0000178433.32526.e0. [DOI] [PubMed] [Google Scholar]
- 24.Diedrich KT, Roberts JA, Schmidt RH, Parker DL. Comparing performance of centerline algorithms for quantitative assessment of brain vascular anatomy. Under review. [DOI] [PubMed]
- 25.Dougherty G, Johnson MJ. Clinical validation of three-dimensional tortuosity metrics based on the minimum curvature of approximating polynomial splines. Med Eng Phys. 2008 Mar;30(2):190–198. doi: 10.1016/j.medengphy.2007.02.009. [DOI] [PubMed] [Google Scholar]
- 26.Phong BT. Illumination for computer generated pictures. Commun. ACM. 1975;18(6):311–317. [Google Scholar]
- 27.Loeys BL, Schwarze U, Holm T, Callewaert BL, Thomas GH, Pannu H, et al. Aneurysm syndromes caused by mutations in the TGF-beta receptor. N Engl J Med. 2006 Aug 24;355(8):788–798. doi: 10.1056/NEJMoa055695. [DOI] [PubMed] [Google Scholar]
- 28.Loeys BL, Chen J, Neptune ER, Judge DP, Podowski M, Holm T, et al. A syndrome of altered cardiovascular, craniofacial, neurocognitive and skeletal development caused by mutations in TGFBR1 or TGFBR2. Nat Genet. 2005 Mar;37(3):275–281. doi: 10.1038/ng1511. [DOI] [PubMed] [Google Scholar]
- 29.Dietz HC, Cutting GR, Pyeritz RE, Maslen CL, Sakai LY, Corson GM, et al. Marfan syndrome caused by a recurrent de novo missense mutation in the fibrillin gene. Nature. 1991 Jul 25;352(6333):337–339. doi: 10.1038/352337a0. [DOI] [PubMed] [Google Scholar]
- 30.Yetman AT, Beroukhim RS, Ivy DD, Manchester D. Importance of the clinical recognition of Loeys-Dietz syndrome in the neonatal period. Pediatrics. 2007 May;119(5):e1199–1202. doi: 10.1542/peds.2006-2886. [DOI] [PMC free article] [PubMed] [Google Scholar]