Retinal Vessel Geometry and Retinal Abnormalities in Cerebral Autosomal Dominant Arteriopathy With Subcortical Infarcts and Leukoencephalopathy

Abstract

Purpose: To investigate the retinal vessel geometry and retinal abnormalities in cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL).

Methods: The data of 35 patients with CADASIL and 35 normal controls were retrospectively included in the study. Fundus photographs of all participants were obtained; signs of retinal abnormality were evaluated; an automated retinal vessel geometry evaluation program based on a deep learning model was used; and retinal vessel geometry parameters, including diameter, ratio, density, branching geometry, asymmetry, and tortuosity characteristics, were measured.

Results: Common signs of retinal abnormality in CADASIL patients included general arterial narrowing (n = 21), arteriovenous nicking (n = 10), and drusen (n = 8). In addition, one CADASIL patient experienced hemorrhage, and one patient had collateral vessels in their fundus. Compared with normal controls, CADASIL patients had a lower central retinal artery equivalent (P = 0.020), lower arteriovenous ratio (P < 0.001), lower number of first branches of the arterioles (P = 0.041), lower arteriolar junctional exponent deviation (P = 0.002), higher venular branching angle (P < 0.001), higher arteriolar branching coefficient (P = 0.001), and higher venular branching coefficient (P = 0.023).

Conclusions: The retinal vessel geometry is altered in CADASIL patients, indicating retinal vessel involvement and vascular remodeling during the pathological process of the disease. Arterial narrowing, arteriovenous nicking, and drusen are common, suggesting the presence of arteriosclerosis and possible premature aging in the fundus of these patients.

Translational Relevance: Our findings contribute to a deeper understanding of how CADASIL affects the retina.

Introduction

Cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL), first named in 1993, is a rare hereditary cerebrovascular disease caused by mutations in the NOTCH3 gene on chromosome 19. The prevalence is approximately one to 14 out of every 100,000 individuals worldwide, varying by region.¹ A genomic study has shown that pathogenic NOTCH3 variants have the highest prevalence in Asian populations.² The NOTCH3 gene encodes a transmembrane receptor protein expressed in vascular smooth muscle cells throughout the body. Genetic mutation of this gene alters the number of cysteine residues in the extracellular domain of the NOTCH3 protein, causing it to lose its receptor function and accumulate in arterioles, further leading to a series of lesions. The pathological features include the deposition of granular osmiophilic material (GOM) in the walls of cerebral arterioles and systemic arterioles, fibrosis and thickening of the vessel walls, and narrowing of the lumen. Affected patients may experience recurrent ischemic stroke, migraines, cognitive impairment, and emotional disorders.^3–6 

The retina is an extension of the central nervous system; the cerebral anterior circulation arteries and ophthalmic artery both originate from the internal carotid artery. Retinal blood vessels and cerebral blood vessels are similar in terms of embryonic development, tissue structure, and physiology; indeed, retinal microvascular abnormalities are common risk factors for cerebrovascular diseases and have demonstrated close associations with stroke and white matter disease.^7,8 CADASIL affects mainly small to medium-sized arteries with diameters ranging from 100 to 400 µm, which are comparable to the diameter of the central retinal artery.⁹ Therefore, CADASIL may also affect retinal arterioles. 

Previous studies have revealed changes in the retinal microcirculation in CADASIL patients, but the results remain controversial. Most studies have shown that compared with normal individuals, CADASIL patients have thinner retinal arterioles,^9–13 whereas another study has indicated no difference between them.¹⁴ One study has suggested that retinal microvascular abnormalities in CADASIL patients occur mainly in arterioles,⁹ but others have reported abnormalities in venules, as well.^14,15 One study has shown that CADASIL does not lead to retinal ischemia,¹⁴ whereas others have found that CADASIL patients may experience retinal hemorrhages, microinfarcts,⁹ cotton-wool spots,¹⁶ and even retinal vein occlusion.^10,15 

Previous studies on CADASIL have focused mainly on the subjective evaluation of abnormal signs of the retina^10,11,16 while neglecting parameters that can be used to measure the morphology of retinal vessels quantitatively. Retinal vessel geometry parameters, such as retinal arteriolar or venular diameters, branching angles, fractal dimensions, asymmetry, and tortuosity, which can be quantitatively derived from analysis of fundus photographs, can reflect the optimal state of retinal circulation. An optimal vessel geometry ensures effective blood flow transport and minimizes energy consumption, whereas changes in retinal vessel geometry often indicate pathological alterations in retinal vessels.¹⁷ Fundus photography provides a noninvasive method to observe and assess the retinal microcirculation. Over the years, many vascular measurement methods based on fundus images have been developed and employed to quantify various aspects of the retinal vessel geometry.¹⁸ Population research has shown that the retinal vessel geometry parameters obtained from these measurements have good predictive and diagnostic capabilities for cardiovascular and cerebrovascular diseases.¹⁸ 

In this study, we investigated retinal abnormalities and retinal vessel geometry changes in CADASIL patients. Any observable abnormal retinal signs, such as general arterial narrowing, arteriovenous nicking, microaneurysms, hemorrhage, cotton-wool spots, and exudations were recorded and compared. Furthermore, we conducted quantitative analyses on blood vessels in fundus photographs to obtain their diameters, ratios, densities, and quantifications of their branching geometry, asymmetry, and tortuosity to evaluate how CADASIL affects the retina. 

Materials and Methods

Study Design and Population

In this retrospective study, the data from 35 adult patients diagnosed with CADASIL at Peking University First Hospital between 2015 and 2023 were included. These patients were diagnosed based on genetic or histopathological testing. Patients with hypertension and diabetes; high myopia, glaucoma, or obvious cataracts; uveitis, endophthalmitis, epiretinal membrane, or retinal detachment; or a history of ophthalmic surgery were excluded. Thirty-five healthy individuals matched for sex and age with the patient group were selected as controls. This study adhered to the tenets of the Declaration of Helsinki and was approved by the National Unit of Clinical Trial Ethics Committee, Peking University First Hospital. 

Collection of Fundus Images and Assessment of Abnormal Retinal Signs

For all patients, fundus photographs of both eyes were obtained by an experienced ophthalmic technician using a nonmydriatic fundus camera (CR-2; Canon, Tokyo, Japan). The fundus photographs had to be clear enough to observe retinal blood vessels and abnormal lesions. Two trained and experienced ophthalmologists, masked to participant characteristics, independently and subjectively evaluated any retinal abnormalities on the fundus photographs. Evaluation criteria or scoring systems for CADASIL-related retinal abnormalities are not currently available. Previous studies often used signs such as general arterial narrowing and arteriovenous nicking to describe CADASIL-related retinal microvascular abnormalities.^9–11 General arterial narrowing and arteriovenous nicking were recorded; in addition, microaneurysms, hemorrhage, cotton-wool spots, exudations, and any other observable abnormalities were also recorded. When the retinal arteriovenous ratio was observed to be equal to or lower than 1:2, it was defined as general arterial narrowing. When a decrease in venular width on both sides of the crossing by an arteriole above the venule was observed at a distance of more than one disc diameter away from the optic disc, it was defined as arteriovenous nicking. The reliability between observers was excellent (kappa ≥ 0.81). When the two evaluators held different opinions on the results, a third evaluator joined the evaluation and they reached a consensus. 

Evaluation of Retinal Vessel Geometry Parameters

An automated retinal vessel geometry evaluation program based on a deep learning model was used to analyze the retinal vessel geometry parameters. The details have been described in a previous publication.¹⁹ In short, the measurement of retinal vessel geometry parameters includes two steps: (1) segmenting the arterioles or venules of the retinal image, and (2) calculating retinal vessel geometry parameters based on the segmented results. We used the U-Net network as the basic architecture for retinal arterioles or venule segmentation. On the basis of U-Net, the encoder part of U-Net was replaced with the Inception-V3 to improve the feature extraction ability of the network. At the same time, we added the Atrous Spatial Pyramid Pooling module to extract larger range features. Finally, the softmax function was used to distinguish arterioles or venules.¹⁹ 

The retinal images were segmented into zones, as shown in Figure 1. The zone marked by the innermost circle in the figure is the optic disc, defined as its radius r. Then, concentric circles with radii of 2r, 3r, and 5r were drawn around the center of the optic disc. Zone B was defined as the annular region remaining after the 2r circular zone was subtracted from the 3r circular zone. Zone C was defined as the annular zone remaining after the 2r circular zone was subtracted from the 5r circular zone. The retinal vessel geometry parameters were measured in regions B and C to ensure that the measured vessels possessed the proper geometric characteristics to be classified as arterioles and venules.¹⁸ During this image-processing step, a total of five retinal vessel geometry characteristics, including vessel diameter, density, asymmetry, branching geometry, and tortuosity, were measured. The specific measurement methods and parameter definitions are shown in Table 1. 

Figure 1.

 

Schematic diagram of retinal vessel geometry parameter measurement. (A) Measurement zone (Zones B and C) for retinal vessel geometry parameters. (B) Daughter angles (θ₁ and θ₂) and the width of the parent vessel (d₀) and its two daughter vessels (d₁ and d₂) at a bifurcation. (C) Formulas for calculating branching geometry and asymmetry parameters. BA, branching angle; BC, branching coefficient; AA, angle asymmetry; AR, asymmetry ratio; JED, junctional exponent deviation.

Table 1.

 

Description of the Retinal Vessel Geometry Parameters

 

Statistical Analysis

Statistical analysis was conducted with SPSS Statistics 21.0 (IBM, Chicago, IL, USA); only the parameters for the right eye were included in the analysis. The normality of the distributions of the variables was assessed with the Shapiro–Wilk method. The retinal vessel geometry parameters are presented as the mean ± standard deviation (SD) if they were normally distributed or as the median (minimum‒maximum) if they were non-normally distributed. Differences in retinal vessel geometry parameters between CADASIL patients and normal controls were assessed with the independent samples t-test (if they were normally distributed) or the Mann–Whitney U test (if they were non-normally distributed). P < 0.05 was considered statistically significant. In this study, 24 retinal vessel geometry parameters were compared. A Bonferroni correction was used for multiple comparison compensation, and, after correction, P < 0.002 was considered statistically significant. 

Results

The data of 35 CADASIL patients (16 males and 19 females; 46.97 ± 9.17 years old, ranging from 28 to 63 years) were included in the study. Among them, general arterial narrowing was found in 21 patients, arteriovenous nicking was found in 10 patients, and hemorrhage was found in one patient. Microaneurysms, exudation, and cotton-wool spots were not observed in any of the patients. In addition, eight patients had drusen, one patient had a collateral vessel, and one patient had myelinated nerve fibers in their fundus (Fig. 2). Thirty-five normal individuals (16 males and 19 females; age, 46.97 ± 9.13 years; range, 28–63 years) were included as controls. Among them, except for three who had general arterial narrowing in their fundus, no abnormalities were found in the fundus. There was no significant difference in age or sex between CADASIL patients and normal controls. 

Figure 2.

 

Fundus photographs of example CADASIL patients. (A) Fundus photograph of a 48-year-old male showing general arterial narrowing. (B) Fundus photograph of a 58-year-old male showing arteriovenous nicking (blue arrow) and collateral vessels (white arrow). (C) Fundus photograph of a 44-year-old female showing retinal hemorrhage (yellow arrow). (D) Fundus photograph of a 53-year-old female showing many drusen.

A comparison of retinal vessel geometry parameters between CADASIL patients and normal controls is shown in Table 2. There was no difference in the standard deviations of the arterioles or venules between the two groups. Compared with that in normal controls, the central retinal artery equivalent (CRAE) was decreased in CADASIL patients, whereas no difference in the central retinal vein equivalent (CRVE) was found between the two groups. Consequently, the arteriovenous ratio (AVR) of CADASIL patients was significantly lower than that of controls. There was no significant difference in the vessel area density (VAD) or vessel length density (VLD) between CADASIL patients and controls. 

Table 2.

 

Comparison of Retinal Vessel Geometry Parameters Between CADASIL Patients and Normal Controls

 

Regarding branching geometric parameters, compared with normal controls, CADASIL patients presented with a lower arteriolar number of first branches (NFB), greater venular branching angle (BA), greater arteriolar and venular branching coefficient (BC), and lower arteriolar junctional exponent deviation (JED) (all P < 0.05) (Fig. 3). After correcting for P values, venular BA, arteriolar BC, and arteriolar JED still showed significant differences between the two groups. None of the other branching geometric parameters, including venular NFB, arteriolar BA, venular JED, or fractal dimension (FD), differed between CADASIL patients and normal controls. There was no significant difference in the asymmetry or tortuosity parameters between the patient group and the control group. 

Figure 3.

 

Comparison of retinal vessel geometry parameters between CADASIL patients and normal controls. (A) Central retinal artery equivalent. (B) Arteriovenous ratio. (C) Number of first branches of the arterioles. (D) Venular branching angle. (E) Arteriolar branching coefficient. (F) Venular branching coefficient. (G) Arteriolar junctional exponent deviation.

Discussion

Blood vessels transport oxygen and nutrients to various organs and tissues to maintain their basic physiological functioning. A healthy retinal vascular network is crucial for the normal functioning of the retina. As the only vessels that can be directly observed in the body, retinal vessels can provide a variety of valuable information and consequently have been investigated extensively over the years. Numerous studies have confirmed that certain systemic diseases can result in abnormalities in the retinal microcirculation.^7,8,20 

In this study, we evaluated retinal abnormalities in CADASIL patients and, for the first time, assessed their retinal vessel geometry. A prominent finding was that the CRAE and AVR were significantly lower in CADASIL patients than in controls. In addition, 21 patients (60%) had general arterial narrowing, and 10 patients (29%) had arteriovenous nicking in their fundus photos, indicating the presence of retinal arteriosclerosis in CADASIL. Roine et al.⁹ reported a significantly reduced AVR in CADASIL patients and a high prevalence of general arterial narrowing and arteriovenous nicking. The study by Pretegiani et al.¹¹ revealed that 68% of CADASIL patients had arteriolar narrowing, and 29% of patients had arteriovenous nicking in their fundus, finds that are similar to our results. In fact, because the retinal vessel wall is transparent, the width of the blood vessels observed on fundus photographs is the width of the blood flow column. The pathological features of CADASIL include the deposition of GOM in the walls of cerebral arterioles, vessel wall fibrosis, and thickening and narrowing of the lumen.⁶ This disease affects not only the cerebral arterioles but also the arterioles of other organs, including the lungs, spleen, heart, liver, testes, and kidneys.²¹ GOM deposition, lumen thickening, and fibrosis have also been observed in the vascular walls of the central retinal artery and its branches in pathological autopsy eyeballs.²² Fang et al.¹³ used optical coherence tomography (OCT) to perform in vivo measurements of retinal blood vessel diameters, and the results revealed arterial wall thickening and a reduction in the inner diameter of the retinal artery in CADASIL patients. Therefore, we believe that the retinal arterioles are affected in CADASIL, and that thickening and fibrosis of the arterial wall lead to arterial stenosis. 

Previous studies found that CADASIL can also affect the diameter of retinal veins. For example, the study by Rufa et al.²³ revealed that five out of 17 CADASIL patients had retinal vein dilation, whereas Alten et al.¹⁴ reported that 14 CADASIL patients had retinal vein dilation in 24 out of 28 eyes. OCT-assisted in vivo vascular measurements have revealed that the inner diameter of the retinal veins and the thickness of the vein walls were greater in CADASIL patients than in controls.^13,14 The specific pathological mechanism by which CADASIL affects retinal veins is still unclear. In a family with CADASIL, Saiki et al.²⁴ reported that seven patients all had varicose veins. Pathological examination of varicosed lower limb veins revealed irregular vascular smooth muscle cells and deposition of GOM. However, in our study, although the CRVE of CADASIL patients was slightly greater than that of control individuals, the difference was not statistically significant. Similar results have been observed in another study.²⁵ The discrepancies may be attributed to differences in the populations included in the studies, the sample sizes, the inclusion criteria, the severity of the disease in the included patients, the vascular assessment methods and measurement areas employed, and so on. More research is necessary to explore whether CADASIL affects the diameters of retinal veins. 

In this study, we found that the branching geometry of the retinal vessels in CADASIL patients undergoes certain changes. The NFB of the arterioles was decreased in CADASIL patients. This parameter is an indicator of the complexity of vascular branching patterns. The identified decrease in the arteriolar NFB suggests a reduction in the complexity of the retinal arteriolar branches and reflects a decrease in vascular network efficiency and a reduction in the oxygen and nutrient supply.²⁶ A previous study has suggested that the reduced complexity of retinal arteriolar branching is associated with endothelial dysfunction, aging, hypertension, and premature mortality.²⁶ Changes in the arteriolar BA may be associated with arteriosclerosis.²⁷ However, we detected elevated venular BA in CADASIL patients, consistent with the results of some previous studies. For example, Huang et al.²⁸ reported that a larger venular BA was associated with greater difficulty in conceiving, and Cheung et al.²⁹ reported that an increased venular BA was related to an increased risk of diabetic retinopathy. JED and BC are indicators that reflect the optimality of the branching geometry. Taken together, the decreased arteriolar JED and increased arteriolar and venular BC identified in this study reflect reduced optimality of the branching geometry in CADASIL. The relationship between the diameters of the daughter vessels and the mother vessels at a bifurcation is related to fluid dynamics losses and can be quantified by calculating the optimal deviation.³⁰ Branching patterns in the retinal vascular network develop to minimize the energy required to maintain effective blood flow; deviation from the optimal retinal branch network is believed to lead to a decrease in microcirculatory transport efficiency and an increase in fluid flow energy loss, which is associated with vascular injury, arteriosclerosis, and aging.^29,30 The research by Frost et al.³¹ revealed that Alzheimer's disease patients had an increased BC of retinal blood vessels but a decreased JED. Choi et al.¹⁷ also reported that an increased arteriolar and venular BC and a decreased arteriolar JED were related to more serious retinopathy in patients with diabetes, which is consistent with our results. The physiological and pathological mechanisms underlying the association between changes in retinal vascular branching patterns and CADASIL are not fully understood and require further research. However, our results highly suggest that CADASIL causes retinal vessel remodeling. This vessel remodeling could reflect the presence of arteriosclerosis, aging, and reduced vascular network transport efficiency in the retinal vessels of CADASIL patients.^26,29,30 

In our study, a retinal hemorrhagic lesion was found in the fundus of a 44-year-old female CADASIL patient (Fig. 2C). Previous studies have also suggested that CADASIL patients may experience retinal microinfarcts, hemorrhage,⁹ and cotton-wool spots.¹⁶ Some studies have reported residual traces of hemorrhage near arteriovenous nicks in the fundus of CADASIL patients.¹⁰ Lacunar infarctions and cerebral microbleeds are also common cranial lesions in CADASIL. Retinal microvascular abnormalities are closely associated to these small infarctions and cerebral microbleeds.^13,32 We speculate that retinal hemorrhage in CADASIL originates from retinal arteriolar stenosis and sclerosis or from vascular injury related to vascular remodeling. We also discovered a collateral vessel connection between two branch veins near the optic disc in a 58-year-old male; however, as of this writing, we have been unable to determine whether this is a congenital formation or a pathological change related to CADASIL. Regular screening of the fundus to detect retinal lesions promptly is necessary. 

Interestingly, we detected varying degrees of fundus drusen in eight patients (approximately 23%). The youngest patient among them was only 38 years old. A large amount of drusen was found in the fundus of a 53-year-old man in this study (Fig. 2D). Some authors emphasize drusen and retinal pigment epithelial atrophy in the fundus of CADASIL patients. In the study by Roine et al.,⁹ among 32 CADASIL patients, two patients had drusen in their eyes, one patient was only 27 years old, and seven patients had pigmented clumps in the midperipheral retina. Modrzejewska et al.³³ reported on a 43-year-old female CADASIL patient and observed clusters of drusen along the retinal blood vessels in her fundus, which they believed may be related to GOM deposition. Pretegiani et al.¹¹ suggested that the presence of drusen may suggest that CADASIL patients are susceptible to premature retinal aging and degeneration. Some studies have found evidence that macular degeneration may be related to arteriosclerosis.³⁴ This may explain the high prevalence of drusen in CADASIL patients, but more research is necessary to confirm this relationship. 

This study has several limitations. Due to the low prevalence of CADASIL, the sample size of this study was small. In addition, the evaluation of abnormal retinal signs is subjective and may include errors. Furthermore, given the retrospective nature of the study, systematic data such as information on cognitive impairment and cranial lesions were limited. Prospective studies should be conducted in the future to explore the causal relationships among retinal vessel geometry parameters and systemic involvement, cranial lesions, and cognitive function in CADASIL patients. Moreover, we did not correct for magnification errors caused by refractive errors. The average refractive errors for CADASIL patients and normal controls were not calculated or compared. This may affect the accuracy of the results, especially for the measurement and comparison of CRAE and CRVE values, and should be taken into account when interpreting the study results. In this study, we also measured AVR. Research shows that refractive error affects the absolute value of retinal vessel diameter measured in fundus photographs but does not affect the AVR.³⁵ We also excluded subjects with high myopia, which may be a major component influencing branching geometry and asymmetry parameters.³⁶ In future studies, refractive errors for CADASIL patients and normal controls should be recorded and compared between the two groups. 

The strength of this study is that we quantitatively investigated the diameter, density, branching geometry, distortion, and asymmetry characteristics of retinal vessels in CADASIL patients, allowing us to better understand how CADASIL affects the retina. Both abnormal retinal signs and retinal vessel geometry are meaningful discoveries and reflect different aspects of the impact of CADASIL on the retina. General arterial narrowing and arteriovenous nicking suggest retinal arteriosclerosis, whereas drusen suggests retinal aging. The quantitative analysis of retinal vessel geometry provides supplementary information for the detection of abnormal retinal signs. For example, CRAE and AVR quantify general arterial narrowing; the tortuosity parameter helps evaluate the possibility and degree of retinal vein occlusion. In addition, branching geometry and asymmetry parameters reflect the optimal state of retinal circulation. Future research will explore the relationship between abnormal retinal signs and retinal vessel geometry, in order to further investigate the pathogenesis of retinal abnormalities in CADASIL patients. 

In conclusion, the retinal vessel geometry is altered in CADASIL patients, specifically manifesting as decreased CRAE, AVR, arteriolar NFB, and arteriolar JED and increased venular BA, arteriolar BC, and venular BC, indicating retinal vessel involvement and vascular remodeling in CADASIL. Arterial narrowing, arteriovenous nicking, and drusen are common manifestations, suggesting the presence of arteriosclerosis and possible premature aging in the fundus of these patients. 

Acknowledgments

Supported by a grant from the National Natural Science Foundation of China (82471112). 

Disclosure: W. Zhang, None; X. Zhou, None; J. Liu, None; T. Tian, None; Y. Guo, None; C. Ling, None; J. Xu, None; Q. Wei, None; Y. Liu, None; Y. Wu, None