Retinal vessel diameter reflects altered resting-state fMRI connectivity and cognitive performance: A community-based study

Abstract

This study investigated the relationship between fundus microvascular characteristics and the nodal local efficiency (Nle) of brain functional connectivity (FC), as well as their association with cognitive performance in a community-based cohort. A total of 1532 participants from Lishui City, China, were enrolled between May 2017 and September 2019 as part of the polyvascular evaluation for cognitive impairment and vascular events (PRECISE) study. Cognitive performance was assessed using the Montreal Cognitive Assessment-Beijing (MoCA-Beijing), and Nle was derived from resting-state functional magnetic resonance imaging (rs-fMRI). Fundus photography of the left eye was performed to measure microvascular features, including the central retinal arterial equivalent (CRAE), central retinal vein equivalent (CRVE), and their ratio (AVR). Correlations between fundus microvascular indices, cognitive function scores, and brain FC were analyzed. Notably, a wider CRVE was significantly associated with poorer naming scores on cognitive assessments. Several key brain regions, including the left orbital gyrus, right inferior temporal gyrus, left parahippocampal gyrus, bilateral posterior hippocampus, left fusiform gyrus, and left inferior parietal lobule, demonstrated significant correlations between fundus microvascular indices and brain FC. These regions played a crucial role in cognitive function and neural network connectivity. Overall, fundus microvascular characteristics were correlated with the indicators of brain FC related to cognitive function. Our findings suggest that fundus microvascular characteristics may serve as a potential non-invasive biomarker for detecting brain functional alterations linked to cognitive dysfunction in elderly populations.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11357-025-01667-w.

Introduction

With the rapid growth of the aging population, the number of individuals aged 65 years and older is projected to increase from 0.7 billion (9%) in 2019 to 1.5 billion (16%) by 2050 [1]. This demographic shift has heightened concerns about elderly health issues, particularly cognitive dysfunction. According to incomplete statistics, cognitive impairment affects tens of millions of older adults [2], reducing the quality of life and imposing significant burdens on families and society. Consequently, early screening and intervention are critical for preventing the progression of cognitive decline.

Resting-state functional magnetic resonance imaging (rs-fMRI) is a high-spatial-resolution imaging technique that aids in the early detection of cognitive dysfunction [3–5]. Graph theory provides a framework for understanding cognitive aging, emphasizing the importance of maintaining a balance between integration, segregation, and “small-world” properties within brain networks [6]. Nodal local efficiency (Nle), a key graph-theoretical metric, reflects the ability of spatially proximate nodes to form tightly interconnected clusters, thereby optimizing signal transmission speed and regional processing [7]. Notably, Nle is highly sensitive to aging-related network alterations, making it a robust marker for detecting age-related changes in brain function. Consistently, studies have reported reduced local efficiency in older adults compared to younger individuals [8]. However, its reliance on specialized sequencing and post-processing tools limits its utility as a routine screening method.

The retina shares pathological and physiological properties with the brain [9]. Previous studies have reported that individuals with cognitive dysfunction exhibit retinal microvascular changes, such as widened small veins, narrowed retinal small arteries, and increased vascular tortuosity in fundus images [10, 11]. These changes suggest that retinal microvascular characteristics could be a non-invasive screening indicator of cognitive dysfunction. However, the connections between fundus microvascular features, brain connectivity, and cognitive performance remain unclear.

This study explores the correlations between fundus microvascular characteristics and brain functional connectivity in relation to cognitive domains within an elderly community-based population. Our findings aim to establish a theoretical foundation for using fundus imaging as a screening tool for cognitive decline in older adults, thereby providing a foundation for future community-based cognitive health assessments.

Methods

Study participants

Participants were selected from the PRECISE study (Clinicaltrials.gov registration number: NCT03178448). This study aimed to investigate the relationship between cardiovascular risk factors and the prevalence of atherosclerotic plaque progression in older adults.

The inclusion and exclusion criteria for PRECISE have been detailed in previous publications [12]. Briefly, PRECISE enrolled community-dwelling adults aged 50–75 years, excluding individuals with contraindications to magnetic resonance imaging (MRI) or computed tomography (CT) for whole-body vascular assessment, those with life expectancy ≤ 4 years due to advanced malignancies or severe comorbidities, and patients with psychiatric disorders. For the present study, participants without baseline data on left-eye retinal microvascular features, cognitive function, and rs-fMRI were further excluded.

The study protocol was approved by the ethics committees at Beijing Tiantan Hospital (IRB approval number: KY2017 - 010–01) and Lishui Hospital (IRB approval number: 2016–42). Written informed consent was obtained from all participants.

Baseline data

Baseline data were collected through face-to-face interviews conducted by trained research coordinators [13]. Information gathered included demographics (age, gender, and education level), body mass index (BMI), systolic blood pressure (SBP), fasting blood glucose (FBG), homocysteine (Hcy), glycated hemoglobin (HbA1c), total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), high-sensitivity C-reactive protein (Hs-CRP), and medical history (including smoking history, alcohol consumption history, hypertension history, diabetes history, coronary heart disease history, dyslipidemia history, and cerebrovascular disease history).

Fundus imaging acquisition and calculation

Each participant underwent fundus photography using a non-mydriatic digital fundus camera. Two 45° fundus photographs were taken: one centered on the optic disc and the other on the macula). Using MATLAB (http://cn.mathworks.com/products/matlab/), the central retinal arterial equivalent (CRAE), central retinal vein equivalent (CRVE), and arteriole-to-venule ratio (AVR) were calculated based on the revised Knudtson–Parr–Hubbard formula [14, 15].

Cognitive measures

Cognitive assessments were conducted through face-to-face interviews by trained neurologists at each visit. To ensure consistency and minimize inter-rater variability, all investigators underwent standardized training at Beijing Tiantan Hospital. Cognitive function was evaluated using the Montreal Cognitive Assessment (MoCA) at baseline [12]. This 30-item scale assesses seven cognitive domains, including: visuospatial and executive function, naming, attention, abstraction, language, delayed recall, and orientation. The total MoCA score ranges from 0 to 30, with higher scores indicating better cognitive function. The MoCA scale has been reported as superior or at least equivalent to other cognitive screening tools for evaluating global cognitive function in older adults [16, 17].

MRI image acquisition

All participants underwent scanning with a 3.0 T MRI scanner (Ingenia 3.0 T, Philips, Best, The Netherlands) while minimizing body movement during the procedure. Brain structural imaging was conducted using a three-dimensional T1-weighted fast field echo sequence, with the following main scanning parameters: repetition time (TR) = 6.7 ms, echo time (TE) = 3.0 ms, voxel size = 1 mm × 1 mm × 1 mm. Resting state functional MRI (rs-MRI) images were acquired using echo-planar imaging (EPI) with the following main scanning parameters: TR = 2100 ms, TE = 30 ms, 38 scanning layers, slice thickness = 2.5 mm, and voxel size = 3 mm × 3 mm × 4 mm.

Rs-fMRI image pre-processing

Pre-processing of raw rs-fMRI images was conducted using SPM software (www.fil.ion.ucl.ac.uk/spm) and the DPARSF toolbox [18]. The brain was divided into 246 regions using a standardized human brain network atlas (Supplementary Table 1) [19]. Each brain region was defined as a node, with its average time correlation sequence extracted as the time correlation sequence for that node. Pearson correlation analysis was performed between the time correlation sequences of any two nodes, forming a 246 × 246 correlation coefficient matrix. Correlation coefficients (r values) were then transformed into Z values using Fisher r–z transformation to normalize the data and construct a regional whole-brain resting-state functional network.

Graph theoretical analysis of nodal attributes in rs-fMRI

Rs-fMRI graph theory analyses model the brain as hierarchical networks of nodes (neurons, populations, and regions) linked by structural–functional edges. These networks exhibit small-worldness, a topology balancing dense local clustering with efficient global integration, reflecting the brain’s optimized organization [20]. We compute nodal local efficiency (Nle) to assess nodal characteristics in brain functional networks. Defined as the reciprocal of the shortest average path length within a node’s immediate neighborhood subgraph, Nle quantifies information transfer robustness among adjacent nodes upon focal disruption, and can serve as a network fault-tolerance metric. Detailed mathematical formulations follow the established methodologies in prior studies [21]. Brain network topological properties were analyzed using graph theoretical network analysis (GRETNA), a MATLAB-based toolbox specifically designed for quantifying functional connectivity and characterizing network architecture in neuroscience research [22].

Statistical analysis

All statistical analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA). Continuous variables were presented as mean ± standard deviation for normally distributed data or as median (interquartile range, IQR) for non-normally distributed data. Categorical variables were represented as percentages. For group comparisons, two-sample T-tests were used for normally distributed continuous variables, while Mann–Whitney U tests were applied to non-normally distributed variables. Categorical variables were compared using χ² tests.

Linear regression analyses were conducted to examine the correlations between fundus microvascular characteristics (CRAE, CRVE, and AVR) and cognitive function. Pearson correlation analyses were conducted to evaluate the relationship between Nle and MoCA scale scores. Additionally, linear regression analyses assessed the associations between CRAE, CRVE, AVR, and Nle indices across various brain regions. Statistical significance was set at P ≤ 0.05.

Results

Demographic and baseline characteristics

From the 3067 participants in the PRECISE cohort, we excluded 32 individuals lacking MoCA scores, 796 individuals without baseline left-eye retinal microvascular measurements and 707 individuals with incomplete rs-fMRI data. A total of 1532 participants were included in the final analysis (Table 1). The cohort had a mean age of 59.4 years (64.8 ~ 55.2), with 45.6% (n = 698) being male. The detailed characteristics of the study population are summarized in Table 2.

Table 1.

Study flow chart

Table 2.

Characteristics of the study participants

Characteristics	Patients included (n = 1532)
Age, years,median (IQR)	59.4 (64.8 ~ 55.2)
Male, n (%)	698 (45.56)
BMI (kg/m2, mean ± SD)	23.9 ± 2.95
Education level (%)
Elementary or below	566 (36.95)
Middle school	485 (31.66)
High school or above	481 (31.4)
Current or ever smoking, n (%)	485 (31.66)
Current or ever drinking, n (%)	356 (23.24)
Medical history, n (%)
Hypertension	630 (41.12)
Diabetes	315 (20.56)
Dyslipidemia	654 (42.69)
Coronary heart disease	213 (13.90)
Ischemic stroke	31 (2.02)
TIA	6 (0.39)
SBP, mmHg, median (IQR)	128 (138 ~ 117)
Hcy, μmol/L, median (IQR)	10.4 (12.7 ~ 8.7)
FBG (mmol/L, mean ± SD)	5.93 ± 1.44
HbA1c, %, median (IQR)	5.70 (6.1 ~ 5.5)
TC, mmol/L,median (IQR)	5.31 (6.01 ~ 4.71)
TG (mmol/L,mean±SD)	1.81 ± 1.3
HDL-C, mmol/L, median (IQR)	1.34 (1.58 ~ 1.15)
LDL-C, mmol/L,median (IQR)	2.78 (3.31 ~ 2.31)
Hs-CRP, mg/L, median (IQR)	1.00 (1.80 ~ 1.00)
MoCA score
Total score, median (IQR)	22 (25 ~ 18)
Visuospatial and executive function, median (IQR)	3 (4 ~ 2)
Naming, mean±SD	2.52 ± 0.76
Attention, median (IQR)	6 (6 ~ 5)
Language function, median (IQR)	2 (2 ~ 1)
Abstraction, median (IQR)	1 (1 ~ 0)
Delayed recall, median (IQR)	3 (4 ~ 1)
Orientation, median (IQR)	6 (6 ~ 6)
Retinal microvascular characteristics
CRAE, mean±SD	19.85 ± 2.49
CRVE, median (IQR)	30.19 (32.19 ~ 28.29)
AVR, mean±SD	0.66 ± 0.10

BMI body mass index, TIA transient ischemic attack, Hcy homocysteine, HbA1c glycated hemoglobin, TC total cholesterol, TG triglycerides, HDL-C high-density lipoprotein, LDL-C low-density lipoprotein, Hs-CRP hypersensitive C-reactive protein, SBP systolic blood pressure, FBG fasting blood glucose, MoCA Montreal Cognitive Assessment, CRAE central retinal arterialy equivalent, CRVE central retinal vein equivalent, AVR arteriole-to-venule ratio

Association between rs-fMRI Nle indicators and cognitive performance

Correlation analysis demonstrated that FC in specific brain regions was significantly associated with the total MoCA score. These regions include bilateral dorsolateral area of superior frontal gyrus, bilateral middle frontal gyrus, left dorsal area of inferior frontal gyrus, left lateral area of orbital gyrus, left caudal and rostral area of superior temporal gyrus, right caudoventral of inferior temporal gyrus, left medioventral of fusiform gyrus, bilateral parahippocampal gyrus, left rostrodorsal area of inferior parietal lobule, left precuneus, bilateral cingulate gyrus, bilateral lateral occipital cortex, right lateral amygdala, bilateral hippocampus, bilateral thalamus and bilateral superior parietal lobule (P < 0.05) (Table 3; Fig. 1). In addition, Nle values in these brain regions showed significant correlations with scores of specific MoCA cognitive subdomains (Supplementary Table 2).

Table 3.

Correlation between nodal local efficiency (Nle) and MoCA total scores

246 Brain region	MNI coordinates			Brain label	r	P
	X	Y	Z
SFG_L_7_4	− 18	− 1	65	7	− 0.082	0.001
SFG_R_7_4	20	4	64	8	− 0.060	0.020
MFG_L_7_1	− 27	43	31	15	− 0.055	0.031
MFG_L_7_3	− 28	56	12	19	− 0.058	0.022
MFG_R_7_3	28	55	17	20	− 0.054	0.035
MFG_L_7_4	− 41	41	16	21	− 0.059	0.022
IFG_L_6_1	− 46	13	24	29	0.057	0.025
OrG_L_6_6	− 41	32	− 9	51	0.093	< 0.001
STG_L_6_4	− 62	− 33	7	75	0.058	0.023
STG_L_6_6	− 55	− 3	− 10	79	0.056	0.029
ITG_R_7_7	54	− 31	− 26	102	− 0.058	0.022
FuG_L_3_2	− 31	− 64	− 14	105	− 0.083	0.001
PhG_L_6_3	− 28	− 32	− 18	113	− 0.053	0.037
PhG_R_6_3	30	− 30	− 18	114	− 0.077	0.003
PhG_R_6_5	22	1	− 36	118	− 0.056	0.029
SPL_L_5_4	− 22	− 47	65	131	− 0.073	0.004
SPL_R_5_4	23	− 23	67	132	− 0.055	0.031
IPL_L_6_2	− 38	− 61	46	137	0.071	0.005
PCun_L_4_3	− 12	− 67	25	151	− 0.062	0.015
PCun_L_4_4	− 6	− 55	34	153	− 0.065	0.012
CG_L_7_2	− 3	8	25	177	0.071	0.005
CG_R_7_2	5	22	12	178	0.058	0.024
CG_L_7_4	− 8	− 47	10	181	− 0.071	0.006
CG_R_7_4	9	− 44	11	182	− 0.060	0.018
CG_L_7_6	− 7	− 23	41	185	0.054	0.034
LOcC _L_4_2	− 47	− 74	3	201	− 0.090	< 0.001
LOcC _R_4_2	48	− 70	− 1	202	− 0.084	0.001
Amyg_R_2_2	28	− 3	− 20	214	− 0.051	0.048
Hipp_R_2_1	22	− 12	− 20	216	− 0.069	0.007
Hipp_L_2_2	− 28	− 30	− 10	217	− 0.086	< 0.001
Hipp_R_2_2	29	− 27	− 10	218	− 0.098	< 0.001
Tha_L_8_2	− 18	− 13	3	233	− 0.052	0.041
Tha_L_8_3	− 18	− 23	4	235	− 0.063	0.013
Tha_R_8_3	18	− 22	3	236	− 0.052	0.044
Tha_L_8_4	− 7	− 14	7	237	− 0.076	0.003
Tha_L_8_5	− 16	− 24	6	239	− 0.070	0.006
Tha_R_8_5	15	− 25	6	240	− 0.080	0.002
Tha_L_8_6	− 15	− 28	4	241	− 0.076	0.003
Tha_R_8_6	13	− 27	8	242	− 0.074	0.004
Tha_L_8_7	− 12	− 22	13	243	− 0.094	< 0.001
Tha_R_8_7	10	− 14	14	244	− 0.089	< 0.001
Tha_L_8_8	− 11	− 14	2	245	− 0.053	0.038
Tha_R_8_8	13	− 16	7	246	− 0.056	0.028

SFG superior frontal gyrus, MFG middle frontal gyrus, IFG inferior frontal gyrus, OrG orbital gyrus, STG superior temporal gyrus, ITG inferior temporal gyrus, FuG fusiform gyrus, PhG parahippocampal gyrus, SPL superior parietal lobule, IPL inferior parietal lobule, Pcun precuneus, CG cingulate gyrus, LOcC lateral occipital cortex, Amyg amygdala, Hipp hippocampus, Tha thalamus, MNI Montreal Neurological Institute

Fig. 1.

A schematic representation of brain functional network is shown below, where the spheres represent locations of brain regions which closely related to cognitive function

Correlation between fundus microvascular features and cognitive function

Analysis of fundus microvascular features and cognitive scores revealed a negative correlation between CRVE and the naming cognitive subdomain score of MoCA. After adjusting for age, gender, BMI, education level, medical history (e.g., hypertension, and diabetes), SBP, Hcy, FBG, HbA1c, lipid profiles (including TC, TG, HDL-C, and LDL-C), and Hs-CRP, multiple linear regression analysis confirmed that wider CRVE was associated with poorer naming scores (β = − 0.016, P = 0.009, 95% CI = − 0.027 ~ − 0.001) (Table 4).

Table 4.

Correlation between fundus microvascular features and cognitive function

	Total score		Visuospatial and executive function		Naming
	Beta (95% CI)	p	Beta (95% CI)	p	Beta (95% CI)	p
CRAE	0.018 (− 0.081 ~ 0.117)	0.724	0.012 (− 0.017 ~ 0.041)	0.414	0.001 (− 0.014 ~ 0.016)	0.932
CRVE	− 0.052 (− 0.117 ~ 0.013)	0.120	− 0.008 (− 0.027 ~ 0.011)	0.401	− 0.016 (− 0.027 ~ − 0.001)	0.009*
AVR	1.510 (− 1.027 ~ 4.047)	0.243	0.459 (− 0.282 ~ 1.202)	0.224	0.295 (− 0.158 ~ 0.749)	0.202
	Attention			Language function
	Beta (95% CI)		p	Beta (95% CI)		p
CRAE	− 0.013 (− 0.039 ~ 0.014)		0.344	− 0.002 (− 0.0229 ~ 0.0181)		0.818
CRVE	− 0.003 (− 0.020 ~ 0.014)		0.704	− 0.009 (− 0.0223 ~ 0.0038)		0.166
AVR	− 0.379 (− 1.053 ~ 0.295)		0.270	0.283 (− 0.2438 ~ 0.8092)		0.292
	Abstraction		Delayed recall		Orientation
	Beta (95% CI)	p	Beta (95% CI)	p	Beta (95% CI)	p
CRAE	0.004 (− 0.012 ~ 0.021)	0.606	0.018 (− 0.021 ~ 0.057)	0.370	0.0001 (− 0.012 ~ 0.012)	0.987
CRVE	− 0.003 (− 0.014 ~ 0.008)	0.607	− 0.007 (− 0.033 ~ 0.018)	0.580	− 0.006 (− 0.014 ~ 0.003)	0.189
AVR	0.116 (− 0.316 ~ 0.547)	0.599	0.571 (− 0.429 ~ 1.571)	0.262	0.165 (− 0.153 ~ 0.483)	0.309

Adjusted for age, gender, BMI, education level, medical history, SBP, Hcy, FBG, HbA1c, TC, TG, HDL-C, LDL-C, and Hs-CRP.

CRAE central retinal arterial equivalent, CRVE central retinal vein equivalent, AVR arteriole-to-venule ratio.

*P < 0.05.

Association between fundus microvascular characteristics and the cognitive-related brain region Nle.

CRAE correlated negatively with Nle in several brain regions including the left medial and lateral areas of the orbital gyrus (medial area: r = − 0.068, P = 0.030; lateral area: r = − 0.070; P = 0.027), the right caudoventral area of inferior temporal gyrus (r = − 0.087, P = 0.005), the left posterior parahippocampal gyrus (r = − 0.072, P = 0.023), and the bilateral posterior hippocampus (left: r = − 0.062, P = 0.049; right: r = − 0.064, P = 0.043). CRVE was negatively correlated with Nle in the left posterior parahippocampal gyrus (r = − 0.065, P = 0.011) and left medioventral area of fusiform gyrus (r = − 0.101, P = 0.004). Additionally, AVR demonstrated a negative correlation with Nle in the left medial area of orbital gyrus (r = − 0.091, P = 0.004) and left rostrodorsal area of the inferior parietal lobule (r = − 0.087, P = 0.006) (Table 5, Figs. 2, 3, and 4).

Table 5.

The locations of brain regions where fundus microvascular features showed significant correlation between Nle

	246 Brain region	MNI coordinates			Brain label	r	p
		X	Y	Z
CRAE	OrG_L_6_1	− 7	54	− 7	41	− 0.068	0.030
	OrG_L_6_6	− 41	32	− 9	51	− 0.070	0.027
	ITG_R_7_7	54	− 31	− 26	102	− 0.087	0.005
	PhG_L_6_3	− 28	− 32	− 18	113	− 0.072	0.023
	Hipp_L_2_2	− 28	− 30	− 10	217	− 0.062	0.049
	Hipp_R_2_2	29	− 27	− 10	218	− 0.064	0.043
CRVE	FuG_L_3_2	− 31	− 64	− 14	105	− 0.101	0.004
	PhG_L_6_3	− 28	− 32	− 18	113	− 0.065	0.011
AVR	OrG_L_6_1	− 7	54	− 7	41	− 0.091	0.004
	IPL_L_6_2	− 38	− 61	46	137	− 0.087	0.006

CRAE central retinal arterial equivalent, CRVE central retinal vein equivalent, AVR arteriole-to-venule ratio, Nle nodal local efficiency, OrG orbital gyrus, ITG inferior temporal gyrus, PhG parahippocampal gyrus, Hipp hippocampus, FuG, fusiform gyrus, IPL, inferior parietal lobule, MNI Montreal Neurological Institute

Fig. 2.

Correlation between central retinal arterial equivalent (CRAE) and Nle. Note: Nle: nodal local efficiency; OrG, Orbital Gyrus; ITG, Inferior Temporal Gyrus; PhG: parahippocampal gyrus; Hipp: hippocampus. For detailed spatial information on these regions, please refer to Supplementary Table 1

Fig. 3.

Correlation between central retinal vein equivalent (CRVE) and Nle. Note: Nle: nodal local efficiency; FuG, Fusiform Gyrus; PhG: parahippocampal gyrus. For detailed spatial information on these regions, please refer to Supplementary Table 1

Fig. 4.

Correlation between arteriole-to-venule ratio (AVR) and Nle. Note: Nle: nodal local efficiency; OrG, Orbital Gyrus; IPL, Inferior Parietal Lobule. For detailed spatial information on these regions, please refer to Supplementary Table 1

Discussion

This study investigates the relationships between retinal vessel caliber, brain functional connectivity, and cognitive performance, identifying key brain regions of significant interest associated with retinal characteristics: the left orbitofrontal cortex (OFC), the right caudoventral region of the inferior temporal gyrus (ITG), the left medioventral fusiform gyrus (FuG), the left posterior parahippocampal gyrus (PhG), the left rostrodorsal area of the inferior parietal lobule (IPL), and the bilateral hippocampus (Hipp). These regions are critically involved in advanced cognitive functions and are closely associated with the progression of cognitive decline [23–27].

The hippocampus and the surrounding PhG are integral to higher-order neural functions through their extensive fiber connections with other brain regions. These connections underlie cognitive control, memory storage, and emotional regulation. Located between the thalamus and the medial temporal lobe, the hippocampus plays a central role in the limbic system and is essential for memory consolidation and neuroplasticity [28], processes that are key to understanding Alzheimer’s disease (AD) progression [27, 29, 30]. Similarly, the PhG is instrumental in memory encoding, information storage, and retrieval [31, 32]. Structural or functional abnormalities in the PhG such as hypoperfusion, atrophy, reduced gray matter volume, decreased fractional anisotropy, and lowered functional connectivity values are commonly observed in individuals with mild cognitive impairment (MCI) or AD [33–35].

The orbitofrontal cortex (OFC), through its connection with the PhG, contributes to visuospatial learning and episodic memory. Evidence suggests that the integrity of PhG-OFC connecting fibers is compromised in individuals with subjective cognitive decline, amnestic MCI, or AD, which correlates with episodic memory deficits [36]. The lateral OFC also maintains strong connections with regions critical to higher-level cognition, including the inferior parietal lobule and the dorsolateral prefrontal cortex [37].

Interestingly, our findings revealed a significant correlation between central retinal arterial equivalent (CRVE) and naming abilities. A significant linear correlation was observed between CRVE and nodal local efficiency (Nle) in the fusiform gyrus (FG), suggesting that fundus features may reflect brain regions associated with naming-related cognitive functions. While cognitive assessments for AD risk often focus on episodic memory [38], emerging evidence suggests that semantic knowledge and word search ability may be among the first cognitive functions to decline during the preclinical stages of AD [39]. The FG is crucial for integrating visual semantic information and aiding object naming [40]. This implies that alterations in FG activation during naming tasks could serve as an early biomarker for AD risk, even before noticeable cognitive decline [41]. In addition, the FG connectivity with ITG and IPL further emphasizes its importance. A recent study has also provided converging evidence highlighting the crucial role of FG and ITG in visual semantic tasks. A positive correlation was observed between the functional connectivity strength between FG and ITG and MMSE scores [42]. Alterations in FG-IPL functional connectivity between IPL and FG have also been observed in individuals with MCI, suggesting its potential as an effective biomarker for early brain functional changes [43].

The novelty of our study resides in its exploration of the association between retinal microvascular characteristics and brain functional connectivity using rs-fMRI. Previous research has demonstrated strong correlations between retinal microvascular characteristics and structural brain imaging features, as well as structural network connectivity [44, 45]. However, rs-fMRI provides a more sensitive measure of functional connectivity changes that often precede structural damage in early disease stages [4]. To our knowledge, this is the first study to establish a link between retinal microvascular characteristics, brain functional connectivity assessed through rs-fMRI, and cognitive performance. The features of the fundus may be associated with the functionality of specific brain regions involved in naming tasks. Our findings underscore the potential of retinal microvascular characteristics as predictive biomarkers for brain functional alterations related to cognitive function, particularly in naming tasks.

This study has several limitations. Firstly, the assessment of retinal microvascular characteristics was limited to arteriovenous diameter indices. Future studies should incorporate additional microvascular morphological features, such as vessel tortuosity and branching complexity, to provide deeper insights into microvascular health and pathology. Secondly, while Nle was used to evaluate brain FC, it may not fully capture the complexity of brain networks. Integrating global network metrics, such as the clustering coefficient or characteristic path length, could enhance our understanding of the brain’s functional architecture and its relationship with retinal microvascular changes. Thirdly, our study did not perform a detailed classification of cognitive impairment etiologies in the elderly population. Future research should further explore the relationship between fundus microvascular characteristics and brain functional features in individuals with different types of cognitive decline, to better establish the potential of fundus imaging as a screening tool for cognitive impairment. Finally, the cross-sectional design of this study limits our ability to infer causal relationships between retinal microvascular characteristics and indicators of brain FC. Therefore, longitudinal cohort studies are essential to further elucidate these associations and determine their clinical relevance.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contribution

Yan Min conceptualized and designed the study, drafted and revised the manuscript, analyzed and interpreted the data, accepted responsibility for research conduct and approved data acquisition and final version. Chang Liu analyzed and interpreted the data and performed the validation. Yanli Zhang and Yuesong Pan performed statistical analysis and revised the manuscript for intellectual content.

Tao Liu supervised the research activities and put forward constructive suggestions during the manuscript revision process. Hongyu Zhou revised the manuscript and supervised the research activities. Yongjun Wang verified the research quality and accuracy, and clinical data. Zixiao Li interpreted the data and critically revised the manuscript for intellectual content.

Funding

National Key R&D Programme of China (2022YFC2504900 and 2022YFC2504902), Beijing Natural Science Foundation (Z200016) funded this work.

Data availability

All datasets generated or analyzed in the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval

The protocol was approved by the ethics committee at Beijing Tiantan Hospital (IRB approval number: KY2017 - 010–01) and the ethics committee at Lishui Hospital (IRB approval number: 2016–42). Participants gave informed consent to participate in the study before taking part.

Conflict of interest

The authors declare no competing interests.