| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Neuroradiology |
1 From the Max-Planck-Institut für Psychiatrie, Kraepelinstrasse 10, 80804 Munich, Germany (D.P.A., B.P., C.G., G.K.E.), and the Department of Neurology, Klinikum Grosshadern, Ludwig-Maximilians-Universität, Munich, Germany (T.G., M.D.). Received March 28, 2000; revision requested May 14; revision received June 28; accepted July 25. Address correspondence to D.P.A. (e-mail: auer@mpipsykl.mpg.de).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Magnetic resonance (MR; T2-weighted and fluid-attenuated inversion-recovery) images obtained in 28 patients with CADASIL were compared with images obtained in 24 patients with sSAE by using an automated pixel-based group comparison with statistical parametric mapping and regional semiquantitative rating.
RESULTS: Visual rating showed higher lesion scores for CADASIL in the temporal and temporopolar white matter (WM). Statistical parametric mapping group analysis independently revealed more extensive bilateral involvement of the anterior temporal and superior frontal WM in CADASIL. There were bilateral signal intensity reductions within the dentate nucleus, deep cerebellar WM, crus cerebri, and thalamus. Lesions extended remarkably more often into arcuate fibers in the temporopolar and paramedian superior frontal lobes in CADASIL. Linear discriminant analysis was used to classify 96% (50 of 52) of the cases correctly, with temporopolar WM and arcuate fiber involvement contributing most to the discrimination function.
CONCLUSION: The presented MR imaging criteria are useful in the diagnostic work-up in patients with leukoencephalopathy and help to differentiate CADASIL from sSAE. The observed pattern of vulnerability in CADASIL suggests future directions for research in the pathophysiology of this disorder. In addition, the study demonstrates the potential of automated image analysis to explore MR imaging lesion patterns.
Index terms: Brain, cortex, 10.184, 10.879 • Brain, diseases, 10.184, 10.879 • Brain, infarction, 10.78 • Brain, MR, 10.121411, 10.121413 • Chromosomes, abnormalities, 10.184
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Another hallmark of the disease is conspicuous magnetic resonance (MR) imaging findings with diffuse white matter (WM) signal intensity abnormalities and circumscribed subcortical lesions suggestive of small infarcts. These circumscribed lesions are located predominantly within the centrum semiovale, thalamus, basal ganglia, and pons (11–17). In most pathology and neuroimaging studies, the pattern of cerebral lesions in CADASIL is described as being similar to that of sporadic subcortical arteriosclerotic encephalopathy (sSAE), or Binswanger disease (15,18), whereas the authors of one report (14) mentioned possibly differentiating radiologic aspects. For clinical practice, it would be desirable to establish MR imaging criteria that support the differential diagnosis between CADASIL and sSAE because of the implications of genetic counseling and to avoid expensive and potentially harmful diagnostic tests (19).
The purpose of this study was to evaluate whether the MR imaging lesion pattern in CADASIL can be differentiated from that in sSAE and, if so, whether certain radiologic criteria may be established as more reliable for individual case classification. A secondary aim of the study was to compare a standard Scheltens (20) and modified visual rating scale with a fully automated statistical image analysis tool (21–23).
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
The second group consisted of 24 consecutive patients (nine men, 15 women; mean age, 65 years; age range, 40–82 years) with sSAE who were recruited from the authors’ institutions. The diagnosis of sSAE was based on the results of clinical and neuroradiologic examination according to ICD-10 criteria (24). These patients were selected on the basis of the following criteria: (a) marked subcortical microangiopathic lesions at MR imaging; (b) a negative family history for strokes, early cognitive impairment, or psychiatric disorders in first- and second-degree relatives; and (c) documented arterial hypertension, defined as systolic values higher than 160 mm Hg, diastolic values higher than 95 mm Hg, or both, measured at several occasions, as a commonly accepted risk factor for sSAE. Enrollment into the study was guided by no neuroimaging criterion other than the presence of marked subcortical vascular lesions. Age was not a selection criterion, even though patients in the sSAE group were significantly older than those in the CADASIL group (P < .05), which reflects the different ages of disease onset in the two conditions. All individuals gave informed consent according to institutional guidelines and the principles of Helsinki declaration.
MR Imaging Examination
MR images were obtained with a 1.5-T whole-body imager (Signa Echospeed; GE Medical Systems, Milwaukee, Wis) by using a standard quadrature head coil. The protocol included sagittal T2-weighted fast spin-echo images (3,000/105 [repetition time msec/echo time msec]; echo train length, eight; section thickness, 4 mm; gap, 2 mm; field of view, 24 cm; matrix, 256 x 192) and contiguous fast fluid-attenuated inversion-recovery (FLAIR) images aligned parallel to the anterior commissure–posterior commissure line and covering the whole brain (10,000/120; inversion time, 2,200 msec; echo train length, 10; section thickness, 5 mm; no gap; field of view, 22 cm; matrix, 256 x 192).
Visual Rating Scales
Semiquantitative regional scores were independently assessed by two investigators (D.P.A., G.K.E.) blinded to the diagnosis, symptoms, and age of the patients. The Scheltens scale (20) was used to rate areas of high signal intensity in the frontal, parietal, temporal, and occipital WM; basal ganglia; thalamus; cerebellum; and brainstem. In brief, the following scoring criteria were applied (20): 0, no lesions; 1, up to five lesions smaller than 3 mm; 2, more than six lesions smaller than 3 mm; 3, up to five lesions between 4 and 10 mm; 4, more than six 4-mm lesions; 5, more than one lesion larger than 11 mm; and 6, confluent lesions. Interrater consensus was reached in differing scores. A global lesion score was calculated as the sum of these eight regional scores (range, 0–64).
On the basis of the published data and our own observations of the distribution of MR imaging lesions in CADASIL, five additional regions were rated: the temporopolar WM, superior frontal WM, temporopolar arcuate fibers, paramedian superior frontal arcuate fibers, and external capsule. Temporopolar WM, defined as WM anterior to the temporal horns, and superior frontal WM, defined as WM 10 mm or more above the ventricular roof, were rated by using a modified Scheltens scale. We added one score level to better differentiate extensive confluent lesions (modification of number 6 of the Scheltens scale) such that score 7 indicated subtotal or total involvement of the WM.
The involvement of arcuate fibers was rated on both FLAIR images and sagittal T2-weighted fast spin-echo images by using the following scoring criteria: only areas of high signal intensity directly reaching the cortical rim in both imaging planes or at a distance of less than 3–4 mm from the external cortical surface were judged to involve the arcuate fibers. On each side, the degree was rated as follows: 0, no involvement; 1, one or two gyri involved; or 2, more than two gyri involved. Areas of high signal intensity in the external capsule were scored according to the spatial extent of involvement: 0, uninvolved; 1, unilateral involvement, up to half of the length of the external capsule; 2, bilateral involvement, up to half of the length; 3, more than half of the length on one side; 4, both sides more than half of the length; 5, total involvement. In addition, the number of lacunae, defined on FLAIR images as noncurvilinear fluid-filled spaces larger than 3 mm in diameter, were counted.
To validate the modified scale, the interobserver reliability was tested by using 
statistics. Multivariate analysis (Wilks 
test) was performed to test for significant group effects for the 13 regional semiquantitative scores, including the five subregional scores averaged between the two observers, and was followed by univariate F tests. To account for nonnormal distribution, all scores were z transformed prior to testing. Significance was defined as P less than .05 for multivariate and P less than .0038 for univariate tests to reduce multiple test error by use of the Bonferroni adjustment. Linear discriminant analysis was performed by using all MR imaging scores and age as possible discriminators. All statistical tests were performed with commercially available software (SPSS 7.5 for Windows; SPSS, Chicago, Ill).
Group Comparison with Statistical Parametric Mapping
Automated image analysis was performed by using FLAIR images in 27 patients with CADASIL and 20 patients with sSAE and discarding images in those patients with incomplete coverage of the head or quality impairment due to motion or other artifacts in any of the FLAIR images.
Postprocessing was performed by using statistical parametric mapping software (SPM96; Wellcome Department of Cognitive Neurology, Institute of Neurology at University College London, United Kingdom). All images were normalized to the T1 template structure provided by statistical parametric mapping, which resulted in a Talairach transformation, followed by spatial smoothing with 4- and 8-mm full-width-at-half-maximum (FWHM) Gaussian kernels that corresponded to the size of about 2 and 4 pixels, respectively. Statistical maps were generated for group contrasts (CADASIL-sSAE, sSAE-CADASIL) after global normalization (analysis of covariance) of pixel intensities to correct for differences in the absolute signal intensities between subjects. To be not too restrictive in data reduction performed by the statistical parametric mapping omnibus F test, the corresponding threshold was set to P less than .05. We performed thresholding of maps at the level of P less than .01; maps were projected onto glass brain templates and, for better depiction, onto a transparent three-dimensional brain surface display provided by the statistical parametric mapping software. Corrected probability values for individual clusters, with adjustment for the number of pixels tested (ie, whole volume, cluster size, peak effect, and smoothness), were computed for both kernels; we performed thresholding of the respective maps at P less than .05 for extent and P less than .01 for height.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
|
|
|
|
= 0.310, P < .0005 and = 0.300, P < .0005, respectively). The following scores were significantly higher in patients with CADASIL compared with those with sSAE (univariate test, P < .0005 with and without global lesion score as a covariate): paramedian superior frontal and temporopolar arcuate fibers and temporopolar and temporal WM. The external capsule reached significance (P = .008) only without the global lesion score as a covariate and not surviving Bonferroni correction. The global lesion score showed significant covariance not only with seven of the eight Scheltens scores (P < .0005) but also with the paramedian superior frontal WM (P < .0005) and external capsule (P < .0005) scores, only marginally with the temporopolar WM scores (P < .004), and not with the arcuate fiber scores.
Table 1 summarizes the score distribution between the two diseases. The interobserver reliability was reasonable to excellent, with values ranging from 0.524 for the external capsule to 0.825 for the right paramedian superior frontal arcuate fibers (Table 2).
|
On the basis of these classification results and the score distributions (Fig 3) in the two conditions, three radiologic criteria with excellent interrater agreement were retrospectively defined and tested for their predictive power and interrater reliability: (a) presence or absence of temporopolar arcuate fiber involvement ( = 0.922); (b) severity of temporopolar WM involvement (score greater than 3, = 0.920); and (c) paramedian superior frontal arcuate fiber involvement (score of 2 or greater, = 0.822). In 52 cases, the first criterion led to correct diagnoses in 48 and 46 cases; the second criterion, in 46 and 44 cases; and the third criterion, in only 39 and 37 cases by observers 1 and 2, respectively. Accepting any of the three criteria for the diagnosis of CADASIL improved the number of correct diagnoses to 49 cases for observer 1 and 48 cases for observer 2, with an interrater agreement of 0.961.
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
The different methods applied in this study concordantly demonstrated that temporal WM involvement is the major abnormality differentiating CADASIL from sSAE. Interestingly, morphometric analyses of vessel wall diameters in medullary arteries in patients undergoing normal aging and in patients with Binswanger disease have demonstrated lobar differences, with the predominant effect on the frontal lobe arteries and the least frequent effect on the temporal lobe arteries (25). This is consistent with the low temporal scores found in sSAE patients. It is conceivable that a dissimilar lobar distribution of small-vessel pathology in CADASIL may in part account for the differential MR pattern found in this study. Systematic regional vessel analyses in CADASIL are required to clarify this issue.
Our finding of frequent arcuate fiber involvement in CADASIL was somewhat unexpected, since previous radiologic reports and autopsy studies have stressed relative sparing or partial preservation of arcuate fibers (8–11). Sparing of the arcuate fibers in sSAE, as confirmed by the results of the present study, has been explained by a particular arterial supply of the immediate subcortical layer arising from cortical rather than perforating medullary arteries that are preferentially affected in sSAE (26). Conversely, in CADASIL, cortical arteries in the anterior temporal and superior frontal regions may also become affected and could thus explain the observed common arcuate fiber involvement. Previous pathology studies (8–12) of CADASIL probably did not focus on arcuate fiber or cortical arterial involvement in temporopolar and superior frontal sites.
The differential anatomic vulnerabilities of subcortical structures in CADASIL and sSAE further substantiate the distinctions that can be drawn between the two conditions and that possibly reflect differential pathophysiologic mechanisms: (a) Ultrastructural examination of small arteries and capillaries in CADASIL reveals characteristic granular deposits that are not found in sSAE and that have been suggested as interfering with the integrity of the blood-brain barrier (7,8). (b) CADASIL is a systemic vasculopathy that demonstrates similar ultrastructural abnormalities in almost all parts of the body (8).
To our knowledge, this is the first statistical parametric mapping–based group comparison of the whole-brain gray-value distribution, aiming to differentiate lesion patterns in vascular encephalopathies. The statistical analysis within statistical parametric mapping directly tests the signal intensity distribution of each pixel by using a group mean comparison equivalent to a t test. Only recently has statistical parametric mapping been introduced as a tool for structural analysis of differences in gray matter distribution (27–29) by using three-dimensional T1-weighted data sets. For our purpose, FLAIR images were chosen owing to the known advantage over other techniques to delineate diffuse WM abnormalities from cystic lesions and ventricles (30). As opposed to visual rating scales, automated image analysis has the advantage of being observer-independent and independent from prior hypotheses that are necessarily contained in any rating scale. Therefore, it is not surprising that statistical parametric mapping could be used to detect additional regional differences (eg, the involvement of the paramedian superior frontal lobe) that were not detected by means of visual rating.
The group comparison within statistical parametric mapping tests for relative areas of high signal intensity between groups without specifying which of the two groups carries the signal intensity abnormality—areas of high signal intensity in one group or areas of low signal intensity in the other group. The finding of between-group areas of low signal intensity within the cerebellum, mainly the dentate nucleus, crus cerebri, and thalamus, in the CADASIL group cannot be explained by WM abnormalities in sSAE, as rating scores were not increased. Also, lacunar lesions that would give rise to areas of low signal intensity on FLAIR images were not more frequent in the CADASIL group. However, at visual inspection, some patients with CADASIL displayed signal intensity decreases within the dentate nuclei, thalami, and basal ganglia (Fig 1b).
One possible explanation of this finding may come from two MR studies in patients with long-standing multiple sclerosis (31) and in children with a history of cerebral infarction (32). Results from these studies showed similar areas of low signal intensity on T2-weighted images within the deep gray matter nuclei; these areas of low signal intensity were interpreted as increased iron deposition that possibly resulted from disturbed axonal iron transport. Neuropathologic studies in CADASIL have shown increased iron or other pigment deposition within the brain (9,10). Additional studies are required to clarify whether the MR imaging areas of low signal intensity found in this study are due to pigment deposition in CADASIL.
A natural limitation of this study is the marked age difference that reflects the known difference in the clinical and radiologic course of the two conditions. As this study aimed to evaluate lesion patterns and the respective radiologic criteria that may prove helpful in the differential diagnosis of the two vascular leukoencephalopathies, the emphasis had to be on group comparability for lesion extent rather than age. Typical age-associated changes such as brain atrophy may be expected to result in systematic lower signal intensity values in the sSAE group in the ventricular and sulcal regions. However, no relative intensity differences were observed in these regions, which argues against indirect age effects.
There is a small possibility that individual cases within the sSAE group may have been labeled erroneously as non-CADASIL cases. We decided to withdraw from systematic biopsy studies and mutational screening of the Notch3 gene in the sporadic cases, since these methods are either invasive or subject to a considerable number of false-negative results (3). From looking at the data, the particularly strong discrimination obtained by means of the linear discriminant analysis seems to confirm the distinct nature of the two populations investigated in this study. In conclusion, this study shows that areas of high T2 signal intensity within the anterior temporal and paramedian superior frontal WM regions and arcuate fiber involvement are useful radiologic signs that help to differentiate CADASIL from sSAE.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantor of integrity of entire study, D.P.A.; study concepts, D.P.A., M.D.; study design, D.P.A.; definition of intellectual content, D.P.A., M.D.; literature research, D.P.A., M.D., B.P.; clinical studies, M.D., T.G.; data acquisition, D.P.A., G.K.E.; data analysis, D.P.A., G.K.E., B.P., C.G.; statistical analysis, D.P.A., C.G., B.P.; manuscript preparation, D.P.A., M.D., B.P.; manuscript editing, D.P.A., B.P., G.K.E.; manuscript review, all authors; manuscript final version approval, D.P.A., M.D.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  K. Yamada, Y. Nagakane, H. Sasajima, M. Nakagawa, K. Mineura, T. Masunami, K. Akazawa, and T. Nishimura Incidental Acute Infarcts Identified on Diffusion-Weighted Images: A University Hospital-Based Study AJNR Am. J. Neuroradiol., May 1, 2008; 29(5): 937 - 940. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Viswanathan, J.-P. Guichard, A. Gschwendtner, F. Buffon, R. Cumurcuic, C. Boutron, E. Vicaut, M. Holtmannspotter, C. Pachai, M.-G. Bousser, et al. Blood pressure and haemoglobin A1c are associated with microhaemorrhage in CADASIL: a two-centre cohort study Brain, September 1, 2006; 129(9): 2375 - 2383. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Melberg, L. Hallberg, H. Kalimo, and R. Raininko MR characteristics and neuropathology in adult-onset autosomal dominant leukodystrophy with autonomic symptoms. AJNR Am. J. Neuroradiol., April 1, 2006; 27(4): 904 - 911. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Holtmannspotter, N. Peters, C. Opherk, D. Martin, J. Herzog, H. Bruckmann, P. Samann, A. Gschwendtner, and M. Dichgans Diffusion Magnetic Resonance Histograms as a Surrogate Marker and Predictor of Disease Progression in CADASIL: A Two-Year Follow-Up Study Stroke, December 1, 2005; 36(12): 2559 - 2565. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Singhal, P. Rich, and H. S. Markus The Spatial Distribution of MR Imaging Abnormalities in Cerebral Autosomal Dominant Arteriopathy with Subcortical Infarcts and Leukoencephalopathy and Their Relationship to Age and Clinical Features AJNR Am. J. Neuroradiol., November 1, 2005; 26(10): 2481 - 2487. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T Haritunians, T Chow, R P J De Lange, J T Nichols, D Ghavimi, N Dorrani, D M St Clair, G Weinmaster, and C Schanen Functional analysis of a recurrent missense mutation in Notch3 in CADASIL J. Neurol. Neurosurg. Psychiatry, September 1, 2005; 76(9): 1242 - 1248. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Tomimoto, R. Ohtani, M. Ihara, and E.K. Perry Absence of cholinergic deficits in "pure" vascular dementia Neurology, July 12, 2005; 65(1): 179 - 179. [Full Text] [PDF]
|
|
|
|
|
|
H Markus Genes for stroke J. Neurol. Neurosurg. Psychiatry, September 1, 2004; 75(9): 1229 - 1231. [Full Text] [PDF]
|
|
|
|
|
|
S. Singhal, S. Bevan, T. Barrick, P. Rich, and H. S. Markus The influence of genetic and cardiovascular risk factors on the CADASIL phenotype Brain, September 1, 2004; 127(9): 2031 - 2038. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 W. S. Millar, A. Lignelli, and M. Hirano MRI of Five Patients with Mitochondrial Neurogastrointestinal Encephalomyopathy Am. J. Roentgenol., June 1, 2004; 182(6): 1537 - 1541. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. van den Boom, S. A. J. Lesnik Oberstein, M. D. Ferrari, J. Haan, and M. A. van Buchem Cerebral Autosomal Dominant Arteriopathy with Subcortical Infarcts and Leukoencephalopathy: MR Imaging Findings at Different Ages--3rd-6th Decades Radiology, December 1, 2003; 229(3): 683 - 690. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Dichgans, B. Putz, D. Boos, and D. P. Auer Role of Subvoxel Free Fluid on Diffusion Parameters in Brain Tissue with Cerebral Autosomal Dominant Arteriopathy with Subcortical Infarcts and Leukoencephalopathy and Its Correlation with Physical Disability: Histogram Analysis of Standard and Fluid-Attenuated MR Diffusion AJNR Am. J. Neuroradiol., June 1, 2003; 24(6): 1083 - 1089. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Tatsch, W. Koch, R. Linke, G. Poepperl, N. Peters, M. Holtmannspoetter, and M. Dichgans Cortical Hypometabolism and Crossed Cerebellar Diaschisis Suggest Subcortically Induced Disconnection in CADASIL: An 18F-FDG PET Study J. Nucl. Med., June 1, 2003; 44(6): 862 - 869. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F Schon, R J Martin, M Prevett, C Clough, T P Enevoldson, and H S Markus "CADASIL coma": an underdiagnosed acute encephalopathy J. Neurol. Neurosurg. Psychiatry, February 1, 2003; 74(2): 249 - 252. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Dong, A. Hassan, Z. Zhang, D. Huber, C. Dalageorgou, H. S. Markus, D. Inzitari, and C. Sarti Yield of Screening for CADASIL Mutations in Lacunar Stroke and Leukoaraiosis Stroke, January 1, 2003; 34(1): 203 - 206. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. S. Markus, R. J. Martin, M. A. Simpson, Y. B. Dong, N. Ali, A. H. Crosby, and J. F. Powell Diagnostic strategies in CADASIL Neurology, October 22, 2002; 59(8): 1134 - 1138. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. van den Boom, S. A. J. Lesnik Oberstein, S. G. van Duinen, M. Bornebroek, M. D. Ferrari, J. Haan, and M. A. van Buchem Subcortical Lacunar Lesions: An MR Imaging Finding in Patients with Cerebral Autosomal Dominant Arteriopathy with Subcortical Infarcts and Leukoencephalopathy Radiology, September 1, 2002; 224(3): 791 - 796. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Liebetrau, J. Herzog, C. U.A. Kloss, G. F. Hamann, and M. Dichgans Prolonged Cerebral Transit Time in CADASIL: A Transcranial Ultrasound Study Stroke, February 1, 2002; 33(2): 509 - 512. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Dichgans, M. Holtmannspotter, J. Herzog, N. Peters, M. Bergmann, and T. A. Yousry Cerebral Microbleeds in CADASIL: A Gradient-Echo Magnetic Resonance Imaging and Autopsy Study Stroke, January 1, 2002; 33(1): 67 - 71. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||
= observer 1, 
= observer 2; closed symbols denote CADASIL cases: 
= observer 1, 
= observer 2.
