Arterial Tortuosity: An Imaging Biomarker of Childhood Stroke Pathogenesis?
Background and Purpose—Arteriopathy is the leading cause of childhood arterial ischemic stroke. Mechanisms are poorly understood but may include inherent abnormalities of arterial structure. Extracranial dissection is associated with connective tissue disorders in adult stroke. Focal cerebral arteriopathy is a common syndrome where pathophysiology is unknown but may include intracranial dissection or transient cerebral arteriopathy. We aimed to quantify cerebral arterial tortuosity in childhood arterial ischemic stroke, hypothesizing increased tortuosity in dissection.
Methods—Children (1 month to 18 years) with arterial ischemic stroke were recruited within the Vascular Effects of Infection in Pediatric Stroke (VIPS) study with controls from the Calgary Pediatric Stroke Program. Objective, multi-investigator review defined diagnostic categories. A validated imaging software method calculated the mean arterial tortuosity of the major cerebral arteries using 3-dimensional time-of-flight magnetic resonance angiographic source images. Tortuosity of unaffected vessels was compared between children with dissection, transient cerebral arteriopathy, meningitis, moyamoya, cardioembolic strokes, and controls (ANOVA and post hoc Tukey). Trauma-related versus spontaneous dissection was compared (Student t test).
Results—One hundred fifteen children were studied (median, 6.8 years; 43% women). Age and sex were similar across groups. Tortuosity means and variances were consistent with validation studies. Tortuosity in controls (1.346±0.074; n=15) was comparable with moyamoya (1.324±0.038; n=15; P=0.998), meningitis (1.348±0.052; n=11; P=0.989), and cardioembolic (1.379±0.056; n=27; P=0.190) cases. Tortuosity was higher in both extracranial dissection (1.404±0.084; n=22; P=0.021) and transient cerebral arteriopathy (1.390±0.040; n=27; P=0.001) children. Tortuosity was not different between traumatic versus spontaneous dissections (P=0.70).
Conclusions—In children with dissection and transient cerebral arteriopathy, cerebral arteries demonstrate increased tortuosity. Quantified arterial tortuosity may represent a clinically relevant imaging biomarker of vascular biology in pediatric stroke.
Arteriopathy is the leading cause of childhood arterial ischemic stroke (AIS) and its recurrence.1–3 Outcomes are poor with most survivors having lifelong disability. Mechanisms are poorly understood, limiting treatment and prevention strategies. The most common syndrome is a unilateral stenotic arteriopathy of the internal, middle, and anterior cerebral artery trifurcation; the term focal cerebral arteriopathy (FCA) of childhood has been coined to describe this imaging appearance in children. The differential diagnosis for FCA includes transient cerebral arteriopathy (TCA), a presumed inflammatory arteriopathy that can have distinct angiographic features (like arterial banding) and by definition has a monophasic natural history.3 Intracranial dissection has also been suggested as a mechanism for FCA with supportive evidence, including possible associations between childhood AIS and trauma, a lack of inflammatory biomarkers, and small pathological series demonstrating dissection in FCA cases.4
An improved understanding of the vascular biology underlying childhood cerebral arteriopathy is essential to develop treatment strategies and to improve outcomes. Large and medium cerebral arteries are inaccessible to pathological examination; however, radiographic imaging of arteriopathy is an alternative, rapidly evolving approach to assessing arterial properties in vivo.5 A growing number of associations between childhood arteriopathies and congenital, genetic syndromes further support a role for inherent abnormalities of the cerebral arteries in childhood AIS pathogenesis.6,7 Abnormal arterial structure marked by having more kinks, twists, and loops can be described as more tortuous. Arterial tortuosity is highly variable and known to be increased in a variety of genetic connective tissue disorders (eg, Menke disease and Loey–Dietz syndrome).8 A recent adult stroke study using standardized visual categorization of cervical arterial tortuosity found an association with extracranial dissection.5 However, computer-assisted analysis of magnetic resonance angiograms (MRAs) may afford more sensitive and objective quantifications of arterial tortuosity and has been used to demonstrate associations with hypertension and other adult cerebrovascular conditions.9,10
Arterial tortuosity has not been investigated in childhood AIS and may represent a window into inherent vascular structure and biology. We hypothesized that arterial tortuosity (of vessels that seem unaffected on standard vascular imaging) is increased in children with stroke because of arterial dissection compared with those with stroke because of other causes or control children.
Materials and Methods
This was a substudy of the Vascular Effects of Infection in the Pediatric Stroke (VIPS) study, the complete methodology of which is described elsewhere.11 VIPS was a prospective, multicenter study of childhood AIS. Children recruited were aged 1 month to 18 years with magnetic resonance imaging–confirmed acute AIS. VIPS collected extensive infectious histories obtained through parental interview, blood and serum samples (and cerebrospinal fluid, when clinically obtained), and standardized brain and cerebrovascular imaging. Importantly, all imaging was reviewed and classified by both the site investigator and additional centralized, multiple, expert, blinded raters. Using standardized criteria,11 each case was first classified by the central review committee into 1 of 3 mutually exclusive diagnostic categories: definite, possible, or no arteriopathy. Those with arteriopathy were then further classified as having secondary diagnoses, including arterial dissection (spontaneous and traumatic), TCA, moyamoya, and secondary vasculitis (including meningitis). The level of certainty on the secondary diagnosis was also assigned. Cases with no arteriopathy were further classified as cardioembolic, other known cause, or idiopathic. For this substudy, we included only those subjects with the highest level of certainty on their diagnostic category: those classified as definite arteriopathy and with a secondary diagnosis classified with high certainty, as well as a subgroup of children with no arteriopathy and cardioembolic stroke. The original, anonymized DICOM files of all eligible subjects were obtained directly from the central VIPS imaging repository for analysis.
To determine normative values for childhood craniocervical arterial tortuosity, MRA studies completed on children from the same age range were obtained from the Alberta Children’s Hospital Pediatric Neuroimaging Database in accordance with previously approved methods. Criteria were (1) age 29 days to 18 years, (2) cerebral time-of-flight MRA completed between 2005 and 2013 (same scanner and protocol requirements as VIPS sites) and reported as normal, and (3) no history of stroke, cerebral or systemic arterial or connective tissue disease, or recent trauma. All control scans were completed on a 1.5-T Siemens Avanto magnetic resonance imaging scanner (Siemens Medical Systems, Erlangen, Germany). Both the VIPS study and this substudy were approved by the institutional Research Ethics Board.
Arterial Tortuosity Quantification
We used a previously validated methodology using ImageJ software to analyze and quantify arterial tortuosity.12 Our technique was similar to that previously described with slight modifications as follows. First, each subject’s cerebral arteries were isolated from their 3-dimensional (3D) time-of-flight MR angiographic source images in DICOM format. The imaging study of top quality closest to stroke diagnosis was used. Segments with focal disease (eg, TCA and dissection) were not included. The algorithm iterates through each 2D source image slice in the 3D space, calculates the center of mass point (single voxel) for each cross section of an arterial lumen, and crops the rest of the local area. These center points are connected to form centerlines that make up an isolated skeleton structure of the arteries. Local and global arterial structure is maintained, including bifurcations (Figure 1).
Tortuosity was then calculated for each individual artery by dividing the path length by the Euclidean (shortest) distance between its end points; this value is referred to as the distance factor metric (DFM). The software does not distinguish arteries from one another, so each arterial segment was manually defined by selecting 2 end points. A limitation in previous studies was analyzing the internal carotid and vertebral arteries as they descend down the neck lacking clearly definable end points. Selection of the end point to define the artery of interest may bias the DFM calculation (Figure 1). To address this, our methodology is designed to only require 1 definite end point, such as the convergence of the vertebral arteries or bifurcation point of the internal carotid artery at the circle of Willis.12 The second end point must still be placed in roughly the same area for comparable results, but the margin for error is much greater. The software then iterates through each voxel along the centerline. At each voxel, the path length and Euclidian distance are calculated between it and the first end point generating a local DFM. After iterating through all the voxels in 3D, the final tortuosity score assigned to an artery is the maximum DFM generated. This choice of using maximum DFM was made based on previously validated methods.12
This process was repeated for each of the following major cerebral arteries: basilar, left and right vertebral, left and right internal carotids, and the M1 segments of the left and right middle cerebral arteries. Anterior cerebral and further order branches were beyond the resolution of the method. The most caudal slices available were used, resulting in vertebral and internal carotid artery imaging to the midcervical level. In subjects with diagnosed arteriopathy, the affected arterial segments were not included in the tortuosity measurements. Primary outcome was the tortuosity score, calculated as the mean maximum DFM of the 7 arteries in each subject.
Analysis and Sample Size
After confirmation of a normal distribution, the relative tortuosity of each major artery was compared using ANOVA with post hoc Tukey test. A paired t test compared relative symmetry between left and right for all paired vessels within subjects. Differences in mean tortuosity across control and disease groups were compared using ANOVA (post hoc Tukey). Tortuosity of traumatic versus atraumatic dissection cases was compared with a Student t test (means) and a Levene test (variance). A blinded intrarater analysis before study initiation confirmed highly reproducible mean and segmental tortuosity measurements (all intraclass correlations, >0.96). On the basis of typical means and variances from previous adult data using similar measures,12 a significant increase of 1SD in dissection subjects, and α = 0.05, our sample of convenience from the VIPS study was 94% powered to address the primary hypothesis.
Of the 480 subjects enrolled in the VIPS study, 100 (21%) satisfied inclusion criteria for this substudy. Excluded case demographics did not differ from the study sample. The characteristics of the study population (including 15 controls) divided by the group are summarized in the Table. Age and sex were comparable across groups.
Representative examples across the spectrum of tortuosity observed are shown in Figure 2. Differences in tortuosity were not readily apparent on visual inspection of the original MRA images.9 Tortuosity scores were normally distributed in all groups. Controls (93% imaged for headaches) demonstrated an average tortuosity score of 1.333 (median, 1.331) with a range of 1.283 to 1.443. Average values, ranges, and variance seemed comparable with previously published values in adults.9
Across all subjects, average tortuosity varied among the different arterial segments (P<0.0001; Figure 3). Consistent with expected anatomic differences, the internal carotid had the highest values, whereas basilar scores were lower. Tortuosity scores were symmetrical with comparable values between left and right measures of paired arteries. Tortuosity scores were not associated with age or sex (Figure 4).
Differences in mean tortuosity were observed across disease groups (P<0.001; Figure 5). Variability around this number was low with an SD of 0.039. On the basis of control measures, the fifth and 95th percentiles for tortuosity were 1.28 to 1.44. Variance of tortuosity was also greater in dissection (P=0.017) and TCA (P=0.042) groups compared with controls but not compared with the other disease groups.
Compared with controls, tortuosity was higher in both dissection (1.398±0.072; P=0.021) and TCA (1.421±0.076; P=0.001) groups. Tortuosity scores were not different from controls for the remaining stroke disease groups: moyamoya (1.324±0.038; P=0.998), meningitis (1.348±0.052; P=0.989), and cardioembolic (1.379±0.056; P=0.190). Within the dissection group, mean tortuosity between traumatic (1.391±0.036) and spontaneous (1.403±0.090; P=0.671) were not different although variance was higher in the spontaneous group (P=0.018).
Our findings suggest that arterial tortuosity is different in children with certain forms of arteriopathic stroke, specifically dissection and TCA. Tortuosity seems to be accurately measurable from clinically obtained MRA in children. Arterial tortuosity may represent an imaging biomarker of inherent vascular biology with implications for understanding the pathophysiology of childhood stroke.
Inherent arterial structure plays a role in specific cerebrovascular diseases at all ages. The number of genetic connective tissue diseases responsible for cerebral arteriopathies continues to grow, such as collagen 4A1 and A2, Majewski Osteodysplastic Primordial Dwarfism Type 2 (MOPD2), and smooth muscle actin (ACTA2).6,13,14 That many of these begin early in life and are accompanied by complications throughout the arterial tree and other organs attests to the importance of inherent arterial stability in long-term health. In adult stroke caused by dissection, evidence of connective tissue alterations is well established, including a large proportion of otherwise asymptomatic patients with evidence of disordered collagen, elastin, or other connective tissue elements visible on skin electron microscopy.15,16 A recent adult stroke study described an association between visually classified tortuosity and dissection.5 Linking these pathological and genetic findings with such readily recognizable imaging biomarkers, such as arterial tortuosity, could facilitate the earlier assignment of likely mechanism and appropriate management in children with stroke.
The TCA syndrome is a well-established imaging syndrome, but its pathophysiology has emerged as one of the most perplexing and controversial issues in childhood stroke.17 Its clinical radiographic characteristics are often indistinguishable from other forms of FCA although we used the best available consensus imaging criteria for classification. Observations of limited, weak epidemiological associations with remote infections and lack of laboratory or imaging biomarkers of inflammation have reasonably questioned the grounds for a primary infectious or inflammatory mechanism. Our finding that the mean tortuosity is different in children with TCA brings a fundamental new consideration to trying to understanding the biological mechanisms of the disease. That the inherent structural properties of the cerebral arteries should predispose one specific section to an acquired infectious or inflammatory process seems unlikely.
Could TCA be mainly because of intracranial dissection? Despite much interest and reasonable theory for an inflammatory, possibly parainfectious mechanism to TCA, definitive proof has been lacking. Transient, abnormal serum biomarkers of disordered inflammation have been described in a small case series of children with TCA when compared with those with cardioembolic stroke.18 Another small case series described 3 children with clinically diagnosed TCA/FCA who died and went to autopsy where pathological evidence of intracranial dissection (and no evidence of inflammation) was described.4 It should also be noted that these 2 possibilities are also not mutually exclusive (eg, an artery damaged by acute inflammation might well be vulnerable to dissection). Our findings that TCA and dissection share a similar degree of increased tortuosity at regional/distant sites to the pathology that differentiates them from both controls and other childhood AIS subtypes do not prove that TCA is intracranial dissection. They do raise serious consideration that the inherent structure of the artery itself may be a key component of the mechanism that underlies the disease.
Our technique provides a straightforward method of objectively quantifying abnormality in arterial structure. However, several methodological issues are identified. Because this was a multicenter study where different MR scanners were used, not all imaging was standardized. Some imaging data from sites were unusable or incompatible with the software. The software method might also be improved when calculating the centerline for an artery. The 3D time-of-flight MRA source images still contained voxel information from the skull, which, in some cases, added noise possibly interfering with the centerline calculations. Signals from the anterior cerebral artery imaging were too weak to be analyzed. Increasing computational power available and improvements in the algorithm may increase our ability to capture smaller vascular structures. In our study, tortuosity scores were assigned by averaging the tortuosity score of each major artery. However, it is possible that specific arteriopathies affect specific arteries differently.
Arterial tortuosity is measureable in children with stroke and may represent a clinically relevant imaging biomarker of vascular biology in pediatric stroke. Children with dissection have increased arterial tortuosity, and no difference was found in traumatic and spontaneous dissection. Whether this reflects inherent abnormalities of arterial structure requires further study. Children with the TCA syndrome also seem to have higher tortuosity. This provides indirect support of previous suggestions that some TCA cases are intracranial dissections.
Appendix: Vascular Effects of Infection in Pediatric Stroke Investigators
Dowling MM (University of Texas Southwestern Medical Center, Dallas), Benedict SL (Primary Children’s Medical Center, Salt Lake City, UT), Bernard TJ (Children’s Hospital Colorado, Aurora), Fox CK (University of California San Francisco), deVeber GA (The Hospital for Sick Children, Toronto, ON), Friedman NR (Cleveland Clinic Children’s Hospital, OH), Lo WD (The Ohio State University and Nationwide Children’s Hospital, Columbus), Ichord RN (Children’s Hospital of Philadelphia, PA), Tan MA (University of the Philippines-Philippine General Hospital, Manila, Philippines), Mackay MT (Royal Children’s Hospital Melbourne, Melbourne, Victoria, Australia), Kirton A (Alberta Children’s Hospital, Calgary, Alberta, Canada), Hernandez Chavez MI (Pontificia Universidad Catolica de Chile, Santiago, Chile), Humphreys P (Children’s Hospital of Eastern Ontario, Ottawa, Ontario, Canada), Jordan LC (Vanderbilt University Medical Center, Nashville, TN), Sultan SM (Columbia University Medical Center, New York, NY), Rivkin MJ (Boston Children’s Hospital, MA), Rafay MF (Children’s Hospital, Winnipeg, University of Manitoba, Winnipeg, Manitoba, Canada), Titomanlio L (Hôpital Robert Debré, Paris, France), Kovacevic GS (Mother and Child Healthcare Institute, Beograd, Serbia), Yager JY (Stollery Children’s Hospital, Edmonton, Alberta, Canada), Amlie-Lefond C (Seattle Children’s Hospital, WA), Dlamini N (Evelina London Children’s Hospital, London, United Kingdom), Condie J (Phoenix Children’s Hospital, AZ), Yeh EA (Children’s Hospital of Buffalo, NY), Kneen R (Alder Hey Children’s Hospital, Liverpool, United Kingdom), Bjornson BH (British Columbia Children’s Hospital, Vancouver, British Columbia, Canada), Pergami P (West Virginia University, Morgantown), Zou LP (Chinese PLA General Hospital, Beijing, China), Elbers J (Stanford Children’s Health, Palo Alto, CA), Abdalla A (Akron Children’s Hospital, OH), Chan AK (McMaster University Medical Center, Hamilton, Hamilton, Ontario, Canada), Farooq O (Women & Children’s Hospital of Buffalo, NY), Lim MJ (Evelina London Children’s Hospital, London, United Kingdom), Carpenter JL (Children’s National Medical Center, Washington, DC), Pavlakis S (Maimonides Medical Center, Brooklyn, NY), Wong VCN (Queen Mary Hospital, the University of Hong Kong, Hong Kong), Forsyth R (Institute of Neuroscience, Newcastle University, Newcastle, United Kingdom).
Sources of Funding
The Vascular Effects of Infection in Pediatric Stroke (VIPS) study (R01 NS062820) was funded by National Institutes of Health. F. Wei was funded by Alberta Innovates–Health Solutions. A. Kirton was funded by Alberta Innovates–Health Solutions and Heart and Stroke Foundation of Canada. The other authors report no conflicts.
↵* A list of all VIPS Investigators is given in the Appendix.
- Received October 30, 2015.
- Revision received February 24, 2016.
- Accepted February 26, 2016.
- © 2016 American Heart Association, Inc.
- Fullerton HJ,
- Wu YW,
- Sidney S,
- Johnston SC
- Amlie-Lefond C,
- Bernard TJ,
- Sébire G,
- Friedman NR,
- Heyer GL,
- Lerner NB,
- et al
- Dlamini N,
- Freeman JL,
- Mackay MT,
- Hawkins C,
- Shroff M,
- Fullerton HJ,
- et al
- Kirton A,
- Crone M,
- Benseler S,
- Mineyko A,
- Armstrong D,
- Wade A,
- et al
- Wintermark M,
- Hills NK,
- deVeber GA,
- Barkovich AJ,
- Elkind MS,
- Sear K,
- et al
- Guo DC,
- Papke CL,
- Tran-Fadulu V,
- Regalado ES,
- Avidan N,
- Johnson RJ,
- et al