In Vivo Macro–Micro Neurovascular Mapping and Visual Field Correspondence in Non-Arteritic Anterior Ischemic Optic Neuropathy

Abstract

Introduction

This study aimed to establish an in vivo neurovascular staging framework for non-arteritic anterior ischemic optic neuropathy (NAION) by integrating wide-field swept-source optical coherence tomography angiography (SS-OCTA) and adaptive optics scanning laser ophthalmoscopy (AOSLO), enabling precise mapping of visual field (VF) defects.

Methods

This prospective study enrolled 22 acute NAION eyes (≤ 3 weeks), 27 chronic NAION eyes (≥ 3 months), and 20 age-/sex-matched healthy controls. SS-OCTA (12 × 12 mm) quantified peripapillary and pan-retinal nerve fiber layer (RNFL) thickness and vessel density (VD) corresponding to VF defect location. AOSLO assessed axonal grading and photoreceptor integrity. Multivariable linear regression was performed to identify independent predictors of visual function.

Results

Within VF defect regions, acute cases showing widespread RNFL swelling, vascular dilation, and VD increase in the temporal sector and intermediate capillary plexus (ICP), chronic cases exhibiting axonal thinning and a nasal-to-temporal capillary attenuation (adj-P < 0.05). Deep capillary plexus (DCP) VD was an independent predictor of acute visual function, whereas chronic visual function was primarily determined by macular ganglion cell thickness and optic disc perfusion. AOSLO confirmed acute axonal edema with hyperreflective dot-like foci (HDF) and chronic atrophy with vessel narrowing (adj-P < 0.05), while photoreceptors remained preserved (P > 0.05).

Conclusions

Wide-field SS-OCTA and AOSLO delineate the stage-specific neurovascular cascade of NAION. Acute NAION is characterized by DCP VD serving as an independent biomarker of visual functional impairment, alongside axonal swelling and early imaging features (e.g., HDF). In the chronic stage, visual outcome is primarily determined by macular neuro-axonal loss and progressive capillary attenuation.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40123-025-01298-6.

Key Summary Points

Why carry out this study?

NAION is a major cause of sudden vision loss, but high-resolution, in vivo mapping of its stage-specific neurovascular pathology remains limited.

This study aimed to establish an in vivo neurovascular staging framework for NAION by integrating wide-field SS-OCTA and AOSLO.

What was learned from the study?

Acute NAION lesions showed axonal swelling with capillary dilation and hyperreflective dot-like foci, while chronic lesions exhibited axonal atrophy and capillary rarefaction; vascular metrics correlated more strongly with visual function than RNFL thickness.

The findings provide objective, stage-specific neurovascular biomarkers that can serve as practical surrogates for visual function, potentially improving disease monitoring and patient stratification in future clinical trials.

Introduction

Non-arteritic anterior ischemic optic neuropathy (NAION) is a major cause of sudden vision loss in adults over 50 and remains a significant therapeutic challenge in neuro-ophthalmology [1]. NAION typically presents with sudden, altitudinal or arcuate visual field (VF) loss due to impaired perfusion of the short posterior ciliary arteries (PCAs), often in patients with systemic vasculopathies [2, 3]. Despite various efforts to improve perfusion or provide neuroprotection, visual outcomes remain poor because of irreversible axonal injury [4, 5].

In recent years, the neurovascular unit (NVU) has emerged as a critical regulator of microcirculatory integrity and neural homeostasis in neuro-ophthalmic diseases. The NVU is a dynamic complex comprising endothelial cells, pericytes, glia, and neurons, and its dysfunction is implicated in conditions such as glaucoma, optic neuritis, and neuromyelitis optica [6, 7]. Although direct evidence of NVU dysfunction in NAION is still lacking, its hallmark pathological features, such as axonal ischemia, blood–optic nerve barrier disruption, and neuroinflammation, closely align with NVU breakdown in other conditions [8, 9]. This convergence suggests a plausible role for NVU dysregulation in NAION, warranting in vivo investigations capable of resolving structural and microvascular alterations at cellular resolution.

In NAION, the typical pattern of altitudinal VF loss reflects sectoral retinal nerve fiber layer (RNFL) damage following the fan-shaped vascular distribution of the PCAs. Standard 30–2 VF (≈ 18 × 18 mm retinal region) covers the papillomacular bundle and arcuate fibers, although standard automated perimetry detects functional deficits, its low spatial resolution and poor reproducibility and reliability hinder accurate localization of lesion boundaries [2, 10]. These limitations highlight the need for high-resolution structural imaging capable of correlating visual dysfunction with underlying anatomical damage. While spectral domain/swept-source optical coherence tomography (SD-/SS-OCT) quantifies RNFL and ganglion cell inner plexiform layer (GCIPL) loss, their limited scan area misses peripheral involvement. Swept-source optical coherence tomography angiography (SS-OCTA) improves vascular visualization but remains constrained to small fields (3–6 mm), insufficient for full NAION mapping [11–16]. Wide-field SS-OCTA (12 × 12 mm) extends coverage but still lacks resolution for cellular features such as microglial activation or axonal disarray. To bridge these scales, we integrated wide-field SS-OCTA with confocal adaptive optics scanning light ophthalmoscopy (AOSLO), enabling concurrent macrovascular and cellular-level assessment.

To overcome these limitations, we combined wide-field SS-OCTA (12 × 12 mm) with confocal AOSLO (approx. 2 μm resolution), establishing a dual-modality imaging platform that bridges macrovascular mapping and cellular-level assessment in vivo [17–20]. This approach enables the comprehensive characterization of NVU dysfunction in NAION across spatial scales by capturing its lesion extent, perfusion deficits, axonal atrophy, and inflammatory cell responses. By integrating structural, vascular, and cellular information, this strategy refines disease phenotyping, supports development of in vivo biomarkers, and provides a foundation for stage-specific, mechanism-driven therapeutic strategies.

Therefore, the purpose of this study was to integrate wide-field SS-OCTA and confocal AOSLO to establish a comprehensive in vivo neurovascular staging framework for NAION, enabling precise topographic mapping of VF defects and characterization of stage-specific structural, vascular, and cellular alterations across the NVU.

Methods

Participants

This observational pilot study was conducted at Tianjin Medical University between January 2024 and June 2025. A total of 45 patients with NAION and 10 healthy controls (HC) were enrolled. Among the NAION cases, 22 patients (22 eyes) were in the acute stage, and one patient had sequential bilateral involvement, with both eyes in the chronic stage at enrollment. No acute precipitating events were identified, and all cases met criteria for idiopathic NAION. The chronic group consisted of 24 patients (27 eyes), 19 with unilateral and 4 with bilateral chronic involvement (enrolled 7 eyes). One patient had acute NAION in one eye and chronic in the fellow eye, counted in the acute group. All HC were age- and sex-matched and had no history of ocular or systemic disease. All subjects gave written informed consent. The study protocol was approved by the Ethics Committee of Tianjin Medical University Eye Hospital (Approval No. 2023KY(L)-46) and followed the principles of the Declaration of Helsinki.

Diagnostic Criteria

NAION diagnosis was based on the following criteria: (1) sudden decrease in best-corrected visual acuity (BCVA), with BCVA > 0.05 logarithm of the minimum angle of resolution (logMAR); (2) relative afferent pupillary defect in the monocularly affected patients; (3) optic disc edema observed during fundus examination; (4) VF defects with typical altitudinal or arcuate pattern, mainly affecting the inferior field; (5) normal erythrocyte sedimentation rate and C-reactive protein; and (6) absence of giant cell arteritis symptoms.

Exclusion Criteria

Exclusion criteria were (1) BCVA worse than 0.05 logMAR ; (2) refractive errors greater than ± 5.0 D or astigmatism exceeding 3.0 D; (3) significant media opacity; (4) history of glaucoma or other neurological diseases; (5) presence of retinal or choroidal lesions; (6) ocular surgery within 6 months (post-cataract); (7) intraocular pressure (IOP) greater than 21 mmHg or an inter-eye difference exceeding 5 mmHg; (8) unreliable VF, defined as fixation losses > 20%, or false-positive/false-negative rates > 15%.

Grouping Criteria

Participants were stratified into two groups: (1) acute group (symptom onset ≤ 3 weeks with optic disc edema), and (2) chronic group (symptom duration ≥ 3 months with optic disc pallor). Retinal regions in each NAION eye were classified as VF defect or VF unaffected based on standard automated perimetry. For case–control analyses, both VF defect and VF unaffected regions in NAION eyes were compared with topographically matched areas in age- and sex-matched HC. As all patients with NAION exhibited inferior visual field defects corresponding to the superior retina, the HC dataset was specifically acquired from the superior retina to enable precise topographic matching for the primary case–control analysis. Consequently, both the superior (VF-defect) and inferior (VF-unaffected) retinal regions from NAION eyes were referenced to the superior HC regions. Within-eye comparisons were performed between VF defect and unaffected regions in the same eye. All analyses were conducted separately for the acute and chronic groups to account for stage-specific alterations.

Ocular Examinations

All participants underwent a comprehensive neuro-ophthalmic examination, including BCVA measurement using a logMAR scale, IOP measurement via noncontact tonometry, and anterior/posterior segment assessment via slit-lamp biomicroscopy (BQ-900, Haag-Streit, Switzerland; Ocular 90D Slit Lamp Lens, Ocular Instruments, USA). Standard automated perimetry (SAP) was performed with the Humphrey Field Analyzer using the SITA 30-2 program (Carl Zeiss Meditec, Dublin, CA, USA).

Wide-Field SS-OCT and OCTA Scanning

Wide-field SS-OCT (VG200, SVision Imaging) was used to assess RNFL thickness and macular ganglion cell inner plexiform layer thickness (GCIPLT) using a 12-mm early treatment diabetic retinopathy study (ETDRS) grid (S3–S12, central 1-mm circle and concentric rings at 1–3 mm, 3–6 mm, 6–9 mm, and 9–12 mm, subdivided into temporal, nasal, superior, and inferior quadrants, comparison of superior vs. inferior quadrant data), and SS-OCTA (1050 nm laser, 200 kHz scan rate) was employed to quantify VD in optic nerve head (ONH) and macular regions (6 × 6 mm) as well as the choroidal vascular index (CVI) using Garway–Heath and ETDRS mapping protocols. Wide-angle angiograms were further subdivided into 3 × 3 mm grids (T1/T2/N2/N1), measure the inner retinal VD per grid, including VD of the radial peripapillary capillary plexus (RPCP), superficial vascular plexus (SVP), intermediate capillary plexus (ICP), and deep capillary plexus (DCP). VD was measured using the built-in algorithm (including large vessels and capillaries). For the peripapillary region (ppVD), analyses were performed both with and without large-vessel exclusion to assess capillary-specific changes, whereas all other regions retained large vessels in the analyses. Capillary-only results were validated through automated large-vessel removal (see Supplementary Table S2).

Confocal AOSLO

A confocal AOSLO system (Mona II; 840 nm light source) was used for ultra-high-resolution retinal imaging. The scanning range was 5.0° × 5.0° (1.5 × 1.5 mm) with horizontal and vertical scanning frequencies of 8 kHz and 14 Hz (14 frames per second), respectively. The system incorporated adaptive optics (Shack–Hartmann wavefront sensor) and real-time eye-tracking to maintain accurate focus and positioning, with an imaging power kept below 600 μW in compliance with ANSI safety standards. Automatic frame processing (deblurring, alignment, and averaging) improved signal-to-noise ratio, and an AI-based algorithm was applied for quantitative photoreceptor analysis, including density, spacing, and regularity. The diameters of third-order branches of the central retinal vein were measured in the comparison area using ImageJ.

Development of an In Vivo RNFL Injury Grading System

A three-tier RNFL grading system was established from AOSLO imaging to objectively assess NAION-related fiber damage: Grade 1 (normal), well-delineated RNFL bundles with individually visible fibers and intact vessel coverage; grade 2 (edematous), blurred fiber margins with partial vessel obscuration; grade 3 (atrophic)—markedly thinned or absent fibers with disrupted fiber–vessel patterns, partially visible below are cone and rod cells. Interobserver agreement was excellent (Cohen’s κ, see Quality Control, Supplementary Figs. S1 and S3), and representative images for each grade are shown in Fig. 1.

Fig. 1.

RNFL injury grading on AOSLO. a Grade 1, clear parallel fibers with normal vessel coverage. b Grade 2, swollen fibers with blurred margins and partial vessel obscuration. c Grade 3, disorganized or absent fibers with severe signal loss, partially visible below are cone and rod cells. Scale bar = 50 μm

Hyperreflective Dot-Like Foci (HDF) Identification

HDF were defined as discrete, round or oval hyperreflective structures with suspected pseudopodia located in the RNFL layer, clearly visible in at least two consecutive frames or on repeated scans to minimize false positives. Two independent graders (blinded to clinical stage) assessed HDF presence in each sector. Discrepancies were adjudicated by consensus. Inter-grader agreement was quantified using Cohen’s κ statistics (Supplementary Table S4 and Fig. S2).

Quality Control and Image Review

All OCTA and AOSLO images underwent standardized quality control; scans with low signal strength (< 7), motion artifacts, or segmentation errors were excluded (< 10% of scans). Two masked graders independently reviewed images, and disagreements were resolved by consensus. RNFL grading reproducibility was excellent (Cohen’s κ = 0.71, P < 0.0001). HDF were confirmed only if visible in ≥ 2 consecutive frames or repeat scans, minimizing false positives. Inter-grader agreement for vessel density and HDF identification was high (κ = 0.72, P < 0.001).

Structure–Function Mapping

SAP 30-2 covers an approximately 18-mm retinal diameter, VF 30-2 results were topographically aligned with 12 × 12 mm SS-OCTA scans by matching T1/T2/N1/N2 regions to hemifield defects. AOSLO images were co-localized using OCTA vascular landmarks to minimize positional errors.

Statistical Analysis

All analyses were performed using SPSS v20.0. Baseline demographics were compared using Kruskal–Wallis tests for continuous variables and χ²/Fisher’s exact tests for categorical variables. The assessment of visual function and ocular structural parameters included BCVA, mean deviation (MD), pattern standard deviation (PSD), and visual field index (VFI), optic disc/macular parameters (e.g., peripapillary retinal nerve fiber layer thickness [pRNFLT], GCIPLT), and retinal parameters (sectoral RNFL thickness, VD in T1/T2/N1/N2, RPCP, SVP, ICP, DCP). The average densities of RPCP, SVP, ICP, and DCP (calculated as [T1 + T2 + N2 + N1]/4) and the grid-based densities are detailed in Supplementary Tables S1 and S2, respectively. Parametric and non-parametric data are expressed as mean ± SD and median (IQR), and compared using t tests and Kruskal–Wallis tests with Dunn’s post hoc test, respectively. For all analyses, a Bonferroni-corrected P < 0.05 was considered significant. Structure–function relationships were assessed using Spearman's rank correlation, with P values further adjusted for the false discovery rate (FDR) using the Benjamini–Hochberg method (reported as q values). Before multivariable regression, all functional outcomes were assessed for normality and transformed when necessary. Among the four functional measures (BCVA, MD, PSD, VFI), only acute-stage MD and chronic-stage BCVA met model-assumption requirements and were therefore selected as dependent variables. Candidate structural and vascular predictors, together with age and sex, were screened for collinearity (VIF < 2.0). The acute model evaluated predictors of MD, and the chronic model evaluated predictors of BCVA. All models satisfied linear regression assumptions, including acceptable Durbin–Watson statistics.

Results

Baseline Characteristics

Table 1 summarizes baseline demographics. There were 22 patients (22 eyes) in the acute stage (6 women, 16 men; median age 47.00 [36.00, 69.00] years), 25 patients (27 eyes) in the chronic stage (6 women, 18 men; median age 53.00 [50.00, 60.00] years), and 10 patients (20 eyes) in the HC group (3 women, 7 men; median age 44.50 [39.75, 52.75] years, see Participants). No significant differences in age, gender, or laterality were found between groups (all P > 0.05). The acute stage showed significantly worse PSD and VFI compared to HC (all adjusted P < 0.0001), while the chronic stage exhibited significantly worse BCVA, MD, PSD, and VFI compared to HC (all adj-P < 0.0001). Additionally, the chronic stage showed worse BCVA and MD compared to the acute stage (all adj-P < 0.0001).

Table 1.

Demographic and clinical characteristics by eye group

Characteristic Acute stage	HC Matched region (n = 20)	Acute VF Defect regiona (n = 22)	Chronic VF Unaffected regionb (n = 27)	H/P χ2/P	Acute vs. HC adj-P	Chronic vs. HC adj-P	Acute vs. chronic adj-P
Age, years	44.50 [39.75, 52.75]	47.00 [36.00, 69.00]	53.00 [50.00, 60.00]	2.99 0.22	0.99	0.05	0.26
Women, n (%)	3 (30.00%)	6 (27.30%)	6 (25.00%)	0.09 0.95	0.99	0.99	0.99
Right eye, n (%)	10 (50.00%)	15 (68.20%)	13 (48.10%)	2.26 0.32	0.35	0.99	0.25
BCVA, logMAR	0.00 [0.00, 0.00]	0.22 [0.00, 0.70]	0.30 [0.22, 0.82]	19.38 < 0.0001****	0.99	0.0001***	0.0001***
MD, dB	0.00 [− 1.46, 0.00]	− 9.80 [− 10.79, − 4.71]	− 13.75 [− 18.73, − 9.58]	34.90 < 0.0001****	0.99	< 0.0001****	< 0.0001****
PSD, dB	1.46 [1.22, 1.69]	8.91 [4.82, 11.40]	11.89 [8.13, 14.92]	27.06 < 0.0001****	< 0.0001****	< 0.0001****	0.29
VFI, %	98.00 [97.00, 98.00]	78.00 [63.00, 88.00]	61.00 [52.00, 71.00]	41.50 < 0.0001****	< 0.0001****	< 0.0001****	0.99

Values are median [IQR]. Kruskal–Wallis with Dunn’s post hoc test (Bonferroni correction). Women and laterality: χ² or Fisher’s exact test

P values < 0.05: *adj-P < 0.05, **adj-P < 0.01, ***adj-P < 0.001, ****adj-P < 0.0001

^aVF defect = region with perimetric loss

^bVF unaffected = region without perimetric loss

Wide-Field SS-OCTA

Topographic RNFL Changes

In the acute stage, pRNFLT showed a graded increase across groups (VF defect > unaffected > HC; adj-P < 0.05 for all comparisons), with VF defect regions being the thickest (339.00 [252.00, 379.00] µm vs. 205.00 [146.30, 253.00] µm vs. 131.00 [117.00, 143.00] µm). Macular GCIPLT in unaffected regions was also elevated compared with HC (adj-P < 0.05). Sector-wise analysis revealed pronounced swelling in S6 to S12 versus HC, while unaffected regions in S6 and S9 also showed thickening (all adj-P < 0.05). Within-eye comparison confirmed no differences between defect and unaffected regions in S3, S6, or S9, except for a significant increase in S12 (adj-P < 0.01). In the chronic stage, pRNFLT in VF defect regions was markedly reduced compared with both unaffected regions and HC (all adj-P < 0.05). Macular GCIPLT showed a graded decrease (defect < unaffected < HC; all adj-P < 0.05), with defect regions showing the most pronounced thinning (51.21 [42.27, 65.43] µm vs. 69.08 [47.49, 75.83] µm vs. 76.82 [74.26, 78.93] µm). Sector-wise analysis confirmed thinning across S3–S12 compared with HC (all adj-P < 0.0001). Notably, unaffected regions (S3 and S12) also showed significant RNFL thickness loss versus HC, suggesting axonal degeneration extends beyond the VF defect area. Within-eye comparison further revealed thinning in S6–S12 of defect regions relative to unaffected regions (all adj-P < 0.05, Table 2, Fig. 2a, b).

Table 2.

Wide-field sectoral RNFLT (S3–S12) in VF defect and VF unaffected regions of NAION eyes

Characteristic Acute stage	HC Matched region (n = 20)	Acute VF Defect regiona (n = 22)	Acute VF Unaffected regionb (n = 22)	H/P	Defect vs. HC adj-P	Unaffected vs. HC adj-P	Defect vs. unaffected adj-P
pRNFLT	131.00 [117.00, 143.00]	339.00 [252.00, 379.00]	205.00 [146.30, 253.00]	38.81 < 0.0001****	0.02*	< 0.0001****	0.001**
GCIPLT	76.82 [74.26, 78.93]	74.34 [70.93, 80.55]	73.31 [71.31, 79.04]	8.58 0.014*	0.09	0.02*	0.99
S3vsI3	33.34 [32.33, 38.56]	33.90 [32.22, 39.25]	34.25 [31.92, 36.18]	0.05 0.98	0.99	0.85	0.84
S6vsI6	43.17 [41.52, 45.35]	49.14 [51.32, 54.60]	51.95 [50.17, 54.35]	31.61 < 0.0001****	< 0.0001****	< 0.0001****	0.99
S9vsI9	53.32 [52.39, 61.59]	71.13 [58.11, 90.54]	67.44 [61.45, 71.19]	25.54 < 0.0001****	< 0.0001****	0.0002***	0.99
S12vsI12	57.06 [51.86, 57.47]	62.76 [60.79, 66.29]	49.76 [42.06, 53.18]	32.79 < 0.0001****	0.002**	0.12	< 0.0001****

Characteristic Chronic stage	HC Matched region (n = 20)	Chronic VF Defect regiona (n = 27)	Chronic VF Unaffected regionb (n = 27)	H/P	Defect vs. HC adj-P	Unaffected vs. HC adj-P	Defect vs. unaffected adj-P
pRNFLT	131.00 [117.00, 143.00]	62.00 [55.00, 71.00]	110.0 [90.00, 126.00]	52.94 < 0.0001****	< 0.0001****	0.07	< 0.0001****
GCIPLT	76.82 [74.26, 78.93]	51.21 [42.27, 65.43]	69.08 [47.49, 75.83]	34.71 < 0.0001****	< 0.0001****	0.001**	0.03*
S3vsI3	33.34 [32.33, 38.56]	21.13 [17.98, 27.30]	25.95 [19.07, 36.36]	24.61 < 0.0001****	< 0.0001****	0.02*	0.05
S6vsI6	43.17 [41.52, 45.35]	23.92 [18.06, 27.94]	39.83 [23.02, 47.29]	29.40 < 0.0001****	< 0.0001****	0.29	0.0004***
S9vsI9	53.32 [52.39, 61.59]	20.02 [18.00, 46.00]	46.32 [37.40, 67.55]	26.01 < 0.0001****	< 0.0001****	0.75	0.0003***
S12vsI12	57.06 [51.86, 57.47]	25.11 [21.31, 33.56]	36.01 [29.45, 42.84]	45.12 < 0.0001****	< 0.0001****	0.0002***	0.02*

RNFLT sectors S3–S12 were analyzed [see Methods for definitions]

Values are median [IQR]. Kruskal–Wallis with Dunn’s post hoc test [Bonferroni correction]

P values < 0.05: *adj-P < 0.05, **adj-P < 0.01, ***adj-P < 0.001, ****adj-P < 0.0001

^aVF defect = region with perimetric loss

^bVF unaffected = region without perimetric loss

Fig. 2.

Wide-field SS-OCTA mapping of axonal and vascular remodeling in NAION. a Acute stage findings. A Representative mapping in HC and acute NAION, VD quantification included large vessels. VF plots (left) were vertically flipped and co-registered with RNFL thickness maps (middle) and wide-field OCTA scans (12 × 12 mm; right). B Quantitative comparisons of sectoral RNFL thickness (S3, S6, S9, S12) showing RNFL swelling and inner retinal VD across predefined grids (T1, T2, N2, N1) demonstrate early vascular dilation in acute VF defect regions (all adj-P < 0.05). b Chronic stage findings. A Representative mapping in HC and chronic NAION eyes. The pattern diagram is arranged according to the acute phase. B Quantitative comparisons of sectoral RNFL thickness (S3, S6, S9, S12) showing progressive RNFL thinning in chronic VF defect regions compared to HC and VF unaffected regions and inner retinal VD across predefined grids (T1, T2, N2, N1) demonstrate significant vascular attenuation in chronic NAION (all adj-P < 0.05)

Topographic Inner Retinal VD Findings

In the acute stage, both ppVD and macular vessel density (mVD) were significantly higher in VF defect and unaffected regions compared to HC (all adj-P < 0.0001). Inner retinal VD was also increased in VF defect regions at T1 and T2 compared with HC (62.51 [57.68, 64.70] vs. 57.83 [53.47, 60.73], adj-P = 0.02; 76.25[72.94, 78.63] vs. 71.81[70.21, 74.18], adj-P = 0.01). Additionally, ICP VD were significantly increased in VF defect regions compared to HC (adj-P = 0.003). In the chronic stage, ppVD was significantly reduced in VF defect regions compared to unaffected regions (adj-P < 0.01). mVD was markedly lower in VF defect regions compared to both HC and unaffected regions (56.74 [53.76, 64.79] vs. 69.49 [67.17, 73.94] and 67.08 [62.81, 72.62], adj-P < 0.0001). Inner retinal VD at T2, N2, and N1, along with RPCP, SVP, and ICP VD, were all significantly reduced in VF defect regions compared to HC (all adj-P < 0.01), notably, unaffected regions also showed VD loss in N2 and N1 (all adj-P < 0.05). Additionally, CVI was significantly elevated in VF defect regions during the chronic stage (adj-P < 0.05, Table 3, Fig. 2a, b, Supplementary Tables S1 and S2).

Table 3.

Wide-field SS-OCTA inner retina VD for VF defected and unaffected regions in NAION-affected eyes

Characteristicc Acute stage	HC Matched region (n = 20)	Acute VF Defect regiona (n = 22)	Acute VF Unaffected regionb (n = 22)	H/P	Defect vs. HC adj-P	Unaffected vs. HC adj-P	Defect vs. unaffected adj-P
ppVD	66.90 [58.91, 73.06]	81.53 [72.12, 81.54]	76.61 [73.49, 78.41]	20.44 < 0.0001****	< 0.0001****	0.002**	0.99
mVD	69.49 [67.17, 73.94]	70.25 [67.69, 76.42]	68.11 [63.09, 69.57]	23.20 < 0.0001****	< 0.0001****	0.001**	0.99
T1	57.83 [53.47, 60.73]	62.51 [57.68, 64.70]	60.83 [56.95, 65.27]	8.19 0.02*	0.02*	0.06	0.99
T2	71.81 [70.21, 74.18]	76.25 [72.94, 78.63]	73.88 [71.28, 76.49]	9.48 0.01*	0.01*	0.31	0.41
N2	81.07 [77.16, 83.35]	85.60 [77.56, 89.02]	81.12 [78.89, 82.63]	6.63 0.04*	0.12	0.99	0.07
N1	88.05 [84.73, 91.22]	87.91 [84.83, 91.25]	78.72 [74.07, 83.77]	24.10 < 0.0001****	0.99	< 0.0001****	< 0.0001****

Characteristicc Chronic stage	HC Matched region (n = 20)	Chronic VF Defect regiona (n = 27)	Chronic VF Unaffected regionb (n = 27)	H/P	Defect vs. HC adj-P	Unaffected vs. HC adj-P	Defect vs. unaffected adj-P
ppVD	66.90 [58.91, 73.06]	59.86 [49.43, 64.58]	72.33 [62.99, 77.49]	9.66 0.008**	0.25	0.79	0.006**
mVD	69.49 [67.17, 73.94]	56.74 [53.76, 64.79]	67.08 [62.81, 72.62]	25.79 < 0.0001****	< 0.0001****	0.53	0.001**
T1	57.83 [53.47, 60.73]	54.85 [44.88, 57.94]	59.46 [42.00, 63.02]	5.33 0.07	0.18	0.99	0.05
T2	71.81 [70.21, 74.18]	65.81 [60.82, 70.09]	70.71 [61.73, 74.10]	13.42 0.001**	0.001**	0.14	0.21
N2	81.07 [77.16, 83.35]	71.92 [69.95, 80.67]	77.11 [68.41, 81.29]	13.04 0.001**	0.001**	0.04*	0.62
N1	88.05 [84.73, 91.22]	75.11 [69.10, 90.54]	75.60 [69.54, 79.01]	22.09 < 0.0001****	0.0003***	< 0.0001****	0.99

Values are median [IQR]. Kruskal–Wallis with Dunn’s post hoc test [Bonferroni correction]

P values < 0.05: *adj-P < 0.05, **adj-P < 0.01, ***adj-P < 0.001, ****adj-P < 0.0001

^aVF defect = region with perimetric loss

^bVF unaffected = region without perimetric loss

^cppVD = peripapillary vessel density, 6 × 6 mm, mVD = macular vessel density, 6 × 6 mm; VD values at T1, T2, N1, and N2 grids are shown [see Methods for grid definitions]

Structure–Function Correlations

Correlation Analysis

In the acute stage, both structural parameters (pRNFLT, GCIPLT) and wide-field vascular parameters (e.g., T1, N1, ICP), the latter represented by the mean VD of the RPCP, SVP, ICP, and DCP, showed strong correlations with visual function indices (e.g., BCVA, MD, PSD; |r| up to 0.69, all q < 0.05). Regarding structural parameters, pRNFLT was significantly correlated with BCVA (r up to − 0.54, q = 0.04). However, notably, sectoral RNFL thickness (S3–S12) did not show significant correlations with VF indices. Interestingly, wide-field vascular parameters (T1, SVP, and ICP VD) were also highly correlated with optic disc and macular structure edema indices (r up to 0.98, q < 0.0001) (Fig. 3a). In the chronic stage, visual function (MD and VFI) remained significantly associated with wide-field vascular parameters, especially T1 and T2 VD (r up to − 0.68, q < 0.01). GCIPLT was significantly associated with BCVA (r up to − 0.46, q = 0.03). Conversely, correlations between broad sectoral structural parameters and global visual field indices were significantly weaker than correlations observed for wide-field vascular parameters and visual function (Fig. 3b).

Fig. 3.

Structure–vascular function relationships in acute and chronic NAION. Spearman correlation matrices show associations among visual function, optic disc and macular structural metrics, and retinal microvascular parameters in a acute and b chronic NAION. The color scale denotes Spearman’s correlation coefficient (r), ranging from − 1.0 (blue) to + 1.0 (red). Statistical significance was adjusted using the Benjamini–Hochberg FDR procedure (*q < 0.05, **q < 0.01, ***q < 0.001, ****q < 0.0001). c Forest plots illustrate standardized regression coefficients (β) and 95% confidence intervals from the final multivariable models after collinearity screening (VIF < 2.0). A In the acute stage, DCP VD was the only independent predictor of MD (β = − 0.67, P = 0.01). B In the chronic stage, both ppVD (β = − 0.14, P = 0.04) and GCIPLT (β = − 0.42, P = 0.04) independently predicted BCVA

Multivariable Analysis of Factors Associated with Visual Function

To ensure statistical robustness and prevent model overfitting, candidate structural and vascular parameters were screened using variance inflation factors (VIF < 2.0), and only variables meeting collinearity and model stability criteria were retained. Separate models were constructed for the acute and chronic stages. Model assumptions were satisfied, with normally distributed residuals, no significant heteroscedasticity, and Durbin–Watson statistics of 1.6–2.2.

In the model for acute-stage MD, sex, age, DCP VD, and pRNFLT were included. DCP VD was the only significant predictor (β = − 0.67; 95% CI − 1.19 to − 0.16; P = 0.01), whereas sex, age, and pRNFLT were not significant. The model showed moderate explanatory ability (adjusted R² = 0.30, P = 0.04) (Table 4 and Fig. 3c).

Table 4.

Multivariable linear regression analysis of factors associated with VF indicators (combined acute and chronic stages)

Acute stage (n = 22)
Predictor	B	SE	β	95% CI	t	P
Women	3.02	1.88	0.36	− 0.11 to 0.83	1.61	0.13
Age	0.10	0.05	0.37	− 0.044 to 0.79	1.89	0.08
DCP VD	− 0.53	0.19	− 0.67	− 1.19 to − 0.16	− 2.76	0.01*
pRNFLT	0.01	0.01	0.24	− 0.26 to 0.73	0.99	0.33

Chronic stage (n = 27)
Predictor	B	SE	β	95% CI	t	P
Women	0.10	0.13	0.13	− 0.20 to 0.47	0.84	0.41
Age	− 0.001	0.01	− 0.20	− 0.59 to 0.19	− 1.07	0.30
ppVD	− 0.01	0.01	− 0.42	− 0.81 to − 0.03	− 2.24	0.04*
pRNFLT	− 0.003	0.004	− 0.14	− 0.49 to 0.21	− 0.85	0.41
GCIPLT	− 0.01	0.01	− 0.42	− 0.81 to − 0.02	− 2.20	0.04*

Model summary for acute mean deviation (MD)

R² = 0.44; Adjusted R² = 0.30; F = 3.29; Model P = 0.04*

All variables demonstrated acceptable VIF (< 2.0); *P < 0.05

Model summary for chronic best-corrected visual acuity (BCVA), presented as BCVA logMAR

R² = 0.47; Adjusted R² = 0.34; F = 3.67; Model P = 0.02*

All variables demonstrated acceptable VIF (< 2.0); *P < 0.05

For chronic-stage BCVA (logMAR), sex, age, ppVD, pRNFLT, and GCIPL thickness were included. Significant predictors were ppVD (β = − 0.14; P = 0.04) and GCIPLT (β = − 0.42; P = 0.04). Age, sex, and pRNFLT were not significant. The model showed moderate explanatory ability (adjusted R² = 0.34, P = 0.02) (Table 4 and Fig. 3c).

Confocal AOSLO

Cellular-Level Axonal Changes

In the acute stage, AOSLO-based RNFL injury gradings were significantly higher in VF defect regions compared to HC (2.00 [2.00, 2.00] vs. 1.00 [1.00, 1.00]; adj-P < 0.0001) and VF unaffected region (2.00 [2.00, 2.00] vs. 1.00 [1.00, 1.00]; adj-P < 0.0001), indicating axonal swelling and boundary blurring. In the chronic stage, RNFL injury grading indicated severe atrophy in VF defect regions compared to HC (3.00 [3.00, 3.00] vs. 1.00 [1.00, 1.00]; adj-P < 0.0001) and unaffected regions (3.00 [3.00, 3.00] vs. 1.00 [1.00, 1.50]; adj-P < 0.0001). Retinal vessel diameter was significantly reduced in VF defect regions to HC and unaffected region (all adj-P < 0.01; Table 5, Fig. 4a, b).

Table 5.

AOSLO for VF defected and unaffected regions in NAION-affected eyes

Characteristic Acute stage	HC Matched region (n = 20)	Acute VF Defect regiona (n = 22)	Acute VF Unaffected regionb (n = 22)	H/P	Defect vs. HC adj-P	Unaffected vs. HC adj-P	Defect vs. unaffected adj-P
RNFL injury grading	1.00 [1.00, 1.00]	2.00 [2.00, 2.00]	1.00 [1.00, 1.00]	57.37 < 0.0001****	< 0.0001****	0.76	< 0.0001****
Density/mm2	12,051.00 [11,115.00, 12,278.00]	12,390.00 [11,197.00, 13,026.00]	11,915.00 [10,927.00, 12,598.00]	1.45 0.48	0.28	0.93	0.32
Spacing, μm	7.31 [6.77, 7.78]	7.50 [7.27, 7.63]	7.58 [7.39, 7.82]	4.36 0.11	0.27	0.11	0.32
Discrepancy, %	15.30 [14.68, 18.29]	15.93 [0.14.60, 18.05]	15.39 [0.1512, 16.55]	0.11 0.95	0.91	0.84	0.75
Regularity, %	94.17 [92.78, 95.77]	94.09 [93.25, 96.58]	94.81 [93.51, 95.56]	1.05 0.59	0.65	0.31	0.56
Retinal vessels diameter, μm	64.59 [61.85, 72.02]	63.82 [63.06, 70.78]	64.60 [61.61, 72.33]	0.004 0.99	0.95	0.97	0.99

Characteristic Chronic stage	HC Matched region (n = 20)	Chronic VF Defect regiona (n = 27)	Chronic VF Unaffected regionb (n = 27)	H/P	Defect vs. HC adj-P	Unaffected vs. HC adj-P	Defect vs. unaffected adj-P
RNFL injury grading	1.00 [1.00, 1.00]	3.00 [3.00, 3.00]	1.00 [1.00, 1.50]	60.44 < 0.0001****	< 0.0001****	0.52	< 0.0001****
Density/mm2	12,051.00 [11,115.00, 12,278.00]	11,218.00 [10,329.00, 12,072.00]	11,926.00 [11,277.00, 12,222.00]	6.16 0.046*	0.16	0.07	0.99
Spacing, μm	7.31 [6.77, 7.78]	7.50 [7.27, 7.63]	7.58 [7.39, 7.82]	3.13 0.21	0.11	0.13	0.92
Discrepancy, %	15.30 [14.68, 18.29]	15.93 [14.60, 18.05]	15.39 [15.12, 16.55]	0.61 0.74	0.96	0.51	0.50
Regularity, %	94.17 [92.78, 95.77]	94.09 [93.25, 96.58]	94.81 [93.51, 95.56]	0.63 0.73	0.68	0.75	0.43
Retinal vessels diameter, μm	64.59 [61.85, 72.02]	63.82 [63.06, 70.78]	64.60 [61.61, 72.33]	25.58 < 0.0001****	0.001**	0.60	0.001**

Values are median [IQR]. Kruskal–Wallis with Dunn’s post hoc test [Bonferroni correction]

P values < 0.05 : *adj-P < 0.05, **adj-P < 0.01, ***adj-P < 0.001, ****adj-P < 0.0001

^aVF defect = region with perimetric loss

^bVF unaffected = region without perimetric loss

Fig. 4.

Confocal AOSLO detects cellular-level axonal changes in NAION. a Acute stage findings. A, B Representative images between HC and acute NAION eyes, VF defect localized RNFL swelling and disorganized axonal bundles were observed (green boxes). In acute eyes, RNFL injury grading were significantly elevated in VF defect regions compared to HC and VF-unaffected regions, while density, spacing, regularity, and retinal vessels diameter showed preserved structural integrity (all adj-P < 0.05). b Chronic stage findings. A, B Representative images between HC and chronic NAION eyes, RNFL atrophy and axonal “darkening” were evident in chronic VF defect region (red boxes), RNFL injury grading were significantly elevated and retinal vessels diameter showed significant narrowing in VF unaffected regions (all adj-P < 0.05) and while photoreceptors remained preserved. c HDF aggregates image. A–D representative AOSLO images from patients with NAION showing HDF aggregates (red circles) located adjacent to retinal vessels or within the peripapillary region. The imaging location and field of view are indicated on each panel. E Comparison of time from symptom onset to examination between patients with detectable HDF infiltration (HDF+) and those without visible HDFs (HDF−). Patients in the HDF+ group exhibited a significantly shorter disease onset time at the time of imaging (*P < 0.05). Scale bar = 50 μm

Photoreceptor Integrity

Across both stages, photoreceptor cell density, spacing, discrepancy, and regularity showed no significant differences between VF defect and unaffected regions (all P > 0.05; Table 5, Fig. 4a, b).

HDF

In 8 of 22 acute NAION eyes, AOSLO revealed HDF clustered adjacent to swollen RNFL bundles (Fig. 4c, Supplementary Table S4 and Fig. S2). These punctate aggregates appeared as bright, radially arranged foci, frequently localized near peripapillary vessels. Patients with these structures were imaged at an earlier stage after symptom onset (mean 4.44 ± 8.38 days vs. 13.43 ± 4.48 days, P = 0.02). No similar aggregates were observed in chronic cases or controls. There was no significant difference in visual function and structural parameters between the two groups (all P > 0.05, Supplementary Table S5).

Discussion

This study combined wide-field SS-OCTA and confocal AOSLO to delineate multiscale NVU alterations in NAION. By integrating pan-retinal RNFL and VD maps with VF data, we overcame the limited spatial coverage and poor reproducibility of conventional perimetry and pRNFLT scans. This approach enabled detection of pathological changes beyond the optic disc and macula, providing a more complete picture of retinal injury distribution.

In the acute stage, we observed widespread RNFL thickening (S6–S12 sectors) and temporal vascular dilation (T1/T2) across ICP. In the chronic stage, RNFL thinning (S3–S12) was accompanied by nasal-to-temporal capillary rarefaction (T2/N2/N1). In the acute stage, these spatial patterns corresponded well with hemifield VF defects and showed a near-linear association with structural edema indices (r ≈ 0.98), although some of this relationship may reflect shared edema effects. Notably, VD metrics, especially T1/T2 and ICP VD, correlated more strongly with VF indices than RNFL thickness, suggesting that vascular parameters may serve as robust functional surrogates when perimetry is unreliable [14, 21–24]. Interestingly, RNFL thickening was also detected in VF unaffected sectors (S6, S9), later manifesting as chronic sectoral thinning (S12).

Our data further support a temporal cascade: transient hyperperfusion and edema dominate the acute phase (approx. 2–3 weeks), followed by capillary dropout and axonal atrophy in chronic disease. This indicates a time-sensitive therapeutic window where vascular and inflammatory modulation may be most effective. Previous studies have predominantly reported a decrease in RPCP VD [23, 25–27]; however, this finding requires more detailed and stratified analysis. Our supplementary analyses provide mechanistic insight into the apparent discrepancy between our results and several prior reports (see Supplementary Table S2). When large vessels were retained, peripapillary and pan-retinal VD significantly increased in acute NAION, indicating widespread vascular expansion during the early stage. However, after the removal of large vessels, the VD within the RPCP showed a significant decrease, notably in the superior hemidisc, a result fully consistent with most previous SD-OCTA studies reporting capillary hypoperfusion [23, 25–27]. These findings suggest that true capillary level rarefaction is confined to the ischemic core, whereas the overall pan-retinal VD increase largely reflects compensatory vasodilation and collateral recruitment in surrounding regions. Notably, similar results have been reported by studies that applied manual large-vessel exclusion, which also demonstrated a transient increase in composite peripapillary vascular metrics [28–30]. This phenomenon is physiologically plausible: autoregulatory mechanisms are likely activated in response to retinal ganglion cell hypoxia, resulting in dilation of surviving capillaries and collateral networks to maintain perfusion. In the macular region, we observed a spatially heterogeneous pattern, with VD reduction being predominant in nasal sectors and VD elevation frequently occurring in temporal sectors (T2/N2), the latter of which strongly correlated with visual function indices. This heterogeneity in reported findings explains the current inconsistent literature on macular VD, which ranges from studies demonstrating significant reductions to those reporting no difference compared with healthy controls [22, 31]. This suggests that NAION-related vascular remodeling is region-specific, with sectoral capillary recruitment possibly contributing to functional compensation. Collectively, these results reconcile previous discrepancies, support the interpretation of acute capillary dilation as a true biological phenomenon rather than an imaging artifact, and highlight the value of capillary-specific analyses for staging NAION and guiding future therapeutic trials.

The structure–function analysis demonstrated a clear stage-dependent pattern in NAION. In the acute stage, global pRNFLT correlated with BCVA, whereas sectoral RNFLT (S3–S12) showed no meaningful association with global VF indices. In contrast, wide-field vascular parameters (ICP and DCP VD) displayed substantially stronger correlations with functional deficits. Multivariable modeling confirmed this pattern, identifying DCP VD as an independent predictor of acute-stage MD. Although macular deep-plexus alterations have been described in ischemic neuroretinal disorders [16, 32], stage-specific multivariable evidence for their functional relevance in NAION remains limited. Our results therefore suggest that DCP VD may serve as a sensitive in vivo marker of early visual functional impairment, warranting validation in longitudinal studies. In the chronic stage, functional status shifted toward correlating with irreversible neural loss. GCIPLT showed the strongest association with BCVA, and both GCIPLT and ppVD emerged as independent predictors. Wide-field vascular metrics (T1, T2 VD) also remained significantly associated with visual function, with consistently stronger correlations than broad sectoral RNFL measures. These findings collectively indicate that chronic visual function primarily reflects macular neural integrity, while persistent vascular alterations provide additional information regarding disease severity.

While SS-OCTA provides valuable structural and vascular biomarkers, it remains limited to indirect surrogate measures. In contrast, confocal AOSLO enables in vivo visualization at cellular resolution [20, 33, 34]. Although AOSLO has been primarily used for cone and rod analysis, its high-resolution capability also offers valuable insights for evaluating the RNFL. As a result of the lack of dedicated RNFL analysis tools for AOSLO, we developed a three-tier RNFL grading system (κ = 0.714). In this study, most grade 2 swollen axons were observed in acute-phase VF defect areas, consistent with RNFL thickening observed on OCT. Similarly, in chronic cases, imaging revealed axonal atrophy meeting grade 3 criteria within functionally impaired regions. AOSLO thus directly visualizes axonal damage, providing clear pathological evidence for determining the timeline of NAION progression. Furthermore, photoreceptor integrity remained preserved across all stages, confirming that functional impairment entirely arises from inner retinal axonal injury [35, 36].

Current evidence indicates that ischemic injury to retinal ganglion cell axons can elicit a range of secondary structural alterations. Whether such injury-associated processes contribute to additional visual dysfunction in human NAION remains unclear, but identifying in vivo imaging correlates of these changes may help guide future mechanistic studies. In our study, we observed clusters of HDF adjacent to swollen axon bundles on AOSLO, which may represent a potential imaging marker of early axonal-level alterations following ischemic insult. Crucially, structures meeting the following criteria were classified as HDF: discrete, bright dot-like signals visible in at least two consecutive AOSLO frames or across repeated scans, located within or adjacent to edematous RNFL bundles. These putative cellular structures, which were more frequently observed in early-onset cases, resemble inflammatory cells previously visualized by AOSLO in uveitis [33] and hyperreflective foci identified on OCTA in diabetic retinopathy and NAION [37, 38]. This timing aligns with animal studies consistently demonstrating macrophage infiltration within 3–5 days after ischemic insult [39, 40], corresponding to the detection window observed in our cohort. The coexistence of acute axonal swelling and HDF raises the possibility that these foci may represent a secondary tissue response to ischemic injury; however, their biological nature cannot be determined from imaging alone. These findings should therefore be interpreted as preliminary and hypothesis-generating. Further longitudinal clinical imaging and experimental studies, including direct histological correlation, are required to clarify the origin, significance, and potential diagnostic value of these structures.

Our multimodal analysis revealed a staged disruption of the NVU. In the acute stage, RNFL swelling co-localized with transient peripapillary and macular hyperperfusion, AOSLO imaging revealed swollen axons accompanied by HDF, which may represent microstructural alterations associated with axonal injury. Notably, transient RNFL swelling was also observed in VF unaffected regions (S6, S9). Our data indicate that RNFL thickening extends beyond the core ischemic territory. This suggests that axoplasmic stasis secondary to primary injury has a broader topographic distribution than the initial vascular insult itself [41, 42], a point warranting further mechanistic investigation. The strong correlation (reaching r = 0.98) between vascular expansion in the SVP/ICP and edema measured by pRNFLT/GCIPLT further supports the presence of regulated vascular remodeling near injured axons, consistent with animal studies linking excitotoxicity and glial activation to vascular dilation and barrier disruption. In the chronic stage, progressive thinning of the RNFL and GCIPL occurred alongside rarefaction of the RPCP and SVP, despite stable VF loss. This suggests that vascular dropout may plateau after axonal loss stabilizes, and that capillary rarefaction in the RNFL likely represents a secondary effect of retrograde axonal degeneration and the associated reduction in metabolic demand, rather than a primary pathogenic event. This finding requires further validation through subsequent mechanistic studies [43–45]. Our findings provide a detailed, in vivo map of this chronic atrophic process.

The key contributions of this study are threefold. First, by integrating wide-field SS-OCTA with cellular-resolution AOSLO, we established neurovascular phenotyping as a surrogate for functional testing. Acute NAION lesions displayed axonal swelling and peripapillary capillary dilation, while chronic lesions exhibited axonal atrophy and capillary rarefaction. These stage-specific patterns offer an objective alternative to perimetry, with SS-OCTA providing practical topographic mapping and AOSLO adding localization precision neurovascular living image via OCTA-derived vascular landmarks. Second, vascular metrics showed superior predictive performance in multivariable modeling. DCP VD was the only independent predictor of acute-stage MD, whereas RNFL parameters were not significant. This suggests that deep capillary plexus measures may serve as more sensitive and robust functional endpoints for future clinical studies. Third, AOSLO identified acute axonal swelling accompanied by clusters of HDF, reflecting secondary microstructural changes adjacent to ischemic injury. In the chronic stage, combined AOSLO-OCTA imaging demonstrated progressive retrograde axonal loss with accompanying capillary rarefaction consistent with reduced metabolic demand. Collectively, these findings advocate for combined vascular and axonal biomarkers in staging, patient stratification, and guiding targeted interventions.

Several limitations of this study should be acknowledged. First, this was a single-center investigation with a modest sample size, and hyperacute cases (within < 7 days of symptom onset) were underrepresented. Second, posterior ciliary artery perfusion was not directly assessed. Third, the primary results are based on VD measurements that included large vessels. A key direction for future research is to validate these findings in the optic disc, macular, and pan-retinal regions, particularly by employing capillary-specific or flux-based analyses combined with cross-platform validation. Fourth, the identity of the HDF requires confirmation through histopathologic correlation. Furthermore, existing ETDRS grid and ring partitions inadequately capture the anatomical distribution of retinal nerve fiber bundles; future work should develop more anatomically aligned zoning methods. Finally, more frequent longitudinal imaging at earlier timepoints (e.g., days 1, 3, 5, and 7) is needed to clarify the dynamic timeline of NVU remodeling.

Conclusion

By integrating wide-field SS-OCTA and AOSLO, this study successfully established an in vivo neurovascular staging framework for NAION, achieving the study’s primary aim. The objective mapping revealed a pathological continuum from acute axonal swelling with vascular dilation to chronic axonal loss and capillary rarefaction. Vascular density metrics emerged as robust, stage-specific biomarkers that correlated more strongly with visual function than RNFL thickness, supporting their potential use as practical surrogates when standard perimetry is unreliable.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

Conceptualization: Shu Yang and Dazhuang Ren. Methodology: Shu Yang and Chang Li. Data acquisition: Shu Yang and Zhangrong Yang. Formal analysis: Shu Yang and Xiaoyun Hou. Investigation: Shu Yang and Xiaoyun Hou. Visualization: Shu Yang. Writing-original draft: Shu Yang. Writing-review & editing: Shu Yang, Dazhuang Ren, Chang Li, Zhangrong Yang, Xiaoyun Hou, Cece Zhao, Lejia Niu and Zhiqing Li. Supervision: Zhiqing Li. Funding acquisition: Zhiqing Li. All authors reviewed and approved the final manuscript for submission.

Funding

This study was funded by Tianjin Key Medical Discipline Construction Project (TJYXZDXK-3-004A-2). The journal’s Rapid Service fee was funded by the authors.

Data Availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Conflict of Interest

Shu Yang, Dazhuang Ren, Chang Li, Zhangrong Yang, Xiaoyun Hou, Cece Zhao, Lejia Niu, and Zhiqing Li confirm that they have no competing interests to declare.

Ethical Approval

The observational pilot study was reviewed and approved by the Ethics Committee of Tianjin Medical University Eye Hospital (Approval No. 2023KY(L)-46) and conducted in accordance with the tenets of the Declaration of Helsinki. Informed consent was obtained from all participants.