Multimodal structure–function mapping of retinal sensitivity loss in pachychoroid spectrum diseases

Abstract

Purpose

Pachychoroid spectrum disease (PSD) is characterized by a thickened choroid and alterations of the retinal pigment epithelium (RPE). Although structural changes in PSD have been well described, the relationship between RPE integrity and retinal function has not been fully clarified. This study investigated the characteristics of retinal sensitivity in eyes with inactive PSD using near-infrared autofluorescence (NIRAF) and microperimetry.

Methods

Twenty-six eyes without recent exudative activity were retrospectively analyzed. Hypoautofluorescent areas on NIRAF images were manually outlined to identify regions of RPE alteration, and mean retinal sensitivity (mRS) was compared between the inside and outside of these lesions.

Results

Thirteen eyes showed macular neovascularization (MNV-positive), while the remaining thirteen had no MNV. In MNV-positive eyes, mRS was significantly reduced within hypoautofluorescent lesions compared with surrounding areas (26.98 ± 7.14 vs. 29.11 ± 4.53 dB, p = 0.021). In MNV-negative eyes, a similar but nonsignificant trend was observed. In an exploratory adjusted linear regression model, MNV status was associated with greater lesion-associated retinal sensitivity reduction (B coefficient = 1.55, p = 0.032). Sublesional choroidal thickness and symptom duration were not significantly associated with ΔmRS.

Conclusion

These results suggest that NIRAF-defined RPE alterations are topographically associated with localized retinal sensitivity reduction, particularly when MNV is present. The combined use of NIRAF, microperimetry, and OCT angiography may support topographic functional assessment in PSD, although prospective validation is required.

Introduction

Pachychoroid spectrum diseases (PSD) represent a continuum of macular disorders—including central serous chorioretinopathy (CSC), pachychoroid pigment epitheliopathy (PPE), pachychoroid neovasculopathy (PNV), and related entities—that share a characteristic choroidal phenotype of focal or diffuse thickening, dilated Haller’s layer vessels (pachyvessels), and attenuation of the inner choroid [1, 2]. Structural alterations of the retinal pigment epithelium (RPE) are common across this spectrum and may coexist with pigment epithelial detachment or type 1 macular neovascularization (MNV) [1–3]. Although best-corrected visual acuity (BCVA) is often preserved in the early stages, patients frequently report subtle visual disturbances that are not captured by standard acuity testing, highlighting the need for more sensitive functional biomarkers [3–5]. While numerous structural assessments using optical coherence tomography (OCT) and OCT angiography (OCTA) have been performed in PSD, functional evaluations—such as retinal sensitivity measurements—have remained relatively limited.

Near-infrared autofluorescence (NIRAF) imaging provides complementary information regarding RPE integrity by detecting melanin-related signals and is less affected by macular pigment than short-wavelength autofluorescence (SWAF), allowing clearer evaluation of RPE alterations in the macular region [6–8]. Microperimetry enables spatially resolved assessment of retinal sensitivity with fundus tracking, allowing localized functional measurements to be compared with multimodal structural findings [9–13]. The combination of NIRAF and microperimetry may therefore be useful for evaluating topographic structure-function relationships in PSD.

Although several studies have investigated the characteristics of fundus autofluorescence (FAF) and its structure–function correlation in central serous chorioretinopathy (CSC), research encompassing the pachychoroid spectrum remains limited, largely due to its relatively recent recognition as a distinct clinical entity [5–7, 12, 13]. In this study, we evaluated whether NIRAF-defined RPE alterations were topographically associated with localized retinal sensitivity differences in inactive PSD and explored whether this association differed according to MNV status.

Methods

This retrospective, observational study included patients with pachychoroid spectrum disease (PSD) who visited Asan Medical Center in Seoul between March 2022 and March 2024. The study protocol was approved by the Institutional Review Board of Asan Medical Center, and all procedures were conducted in accordance with the tenets of the Declaration of Helsinki. Eyes were included if they met all of the specified inclusion criteria and none of the exclusion criteria outlined below.

Inclusion criteria

Central serous chorioretinopathy (CSC), pachychoroid pigment epitheliopathy (PPE), and pachychoroid neovasculopathy (PNV) patients based on multimodal imaging were included. Clinical diagnoses were established by two retina specialists (S.K and J.L) on the basis of multimodal imaging, using the following operational definitions. CSC included both active and resolved forms but only eyes with inactive CSC were enrolled. Inactive CSC was defined as eyes that fulfilled pachychoroid criteria (subfoveal choroidal thickness ≥ 300 µm, dilated Haller’s layer vessels or choroidal hyper-permeability on indocyanine-green angiography) and showed no subretinal fluid on OCT at baseline, with characteristic stigmata of previous serous detachment, such as focal or track-like RPE mottling, window defects on fluorescein angiography. PPE was diagnosed when multifocal RPE alterations (mottling, pachydrusen or shallow irregular pigment epithelial detachment) were present in pachychoroid eyes. PNV required demonstrable type 1 macular neovascularization on OCT angiography or indocyanine-green angiography in the absence of soft drusen with the aforementioned pachychoroid criteria.

Enhanced-depth optical coherence tomography (EDI-OCT) had to demonstrate a pachychoroid configuration, defined as a subfoveal choroidal thickness of ≥ 300 µm. This quantitative threshold, widely adopted in pachychoroid research, distinguishes eyes with pathologically thickened choroid and dilated Haller’s layer vessels from age-matched normative values. If both eyes met the inclusion criteria, the eye with the longer duration of symptom was selected. For eyes classified as CSC or PPE, the absence of MNV was confirmed on OCTA; when OCTA findings were inconclusive, indocyanine-green angiography was additionally reviewed. Eyes with subclinical MNV were reclassified as PNV and excluded from the CSC and PPE subgroups.

Exclusion criteria

Any eye that had exhibited exudative activity – including subretinal or intraretinal fluid, hemorrhage or lipid deposition – within the preceding 6 months was excluded. This wash-out interval was selected to eliminate the confounding, short-term functional fluctuations in PSD. Eyes were excluded when either structural or functional testing failed to meet predetermined quality standards. Specifically, cases in which near-infrared autofluorescence (NIRAF) failed to provide a clearly delineated hypoautofluorescent lesion due to media opacity or low signal that compromised image quality were omitted. Microperimetry recordings obtained with the Compass device (CenterVue, Padova, Italy) were disregarded if fixation losses reached ≥ 20% or false-positives reached ≥ 15%; thus, only eyes with stable fixation were included to ensure the topographic accuracy of the structure-function mapping. Suboptimal EDI-OCT images in which the choroidoscleral interface could not be visualized with confidence were excluded as such.

Eyes with prior intravitreal anti-VEGF (vascular endothelial growth factor) treatment, focal laser or photodynamic therapy were ruled out. Eyes exhibiting complete RPE and outer-retinal atrophy on FAF and OCT, or eyes with soft drusen (> 125 µm), reticular pseudodrusen, or any other signs of intermediate or late non-neovascular age-related macular degeneration were excluded. Additional exclusion criteria were: (1) Prior pars plana vitrectomy or any other retinal surgery; (2) Vitreomacular interface disorders such as epiretinal membrane or vitreomacular traction; (3) Diabetic retinopathy; (4) Glaucoma; (5) Uveitis; (6) Spherical equivalent of ± 6.0 diopters or greater.

Ophthalmic examinations

We reviewed patients’ medical records for basic demographic data such as age, sex, symptom duration and medical history. Patients underwent a comprehensive ophthalmic examination at the initial visit, including best-corrected visual acuity (BCVA) measurement, intraocular pressure measurement, slit-lamp biomicroscopy, and fundus examination. The spectral-domain optical coherence tomography (OCT) images were obtained using a Spectralis HRA2 (Heidelberg Engineering, Heidelberg, Germany) and subfoveal choroidal thickness was measured on OCT based on enhanced depth images. For the OCT angiography (OCTA), AngioVue (OptoVue Inc., Fremont, CA) utilizing the split-spectrum amplitude-decorrelation system was used and outer retina and choriocapillaris slabs were inspected for subclinical macular neovascularization (MNV). Images with signal strength ≥ 6 were included, and the quality of each image and segmentation of the retinal layers were manually checked by a single retinal specialist (S.K.). Fluorescein angiography (FA) and indocyanine-green angiography (ICGA) were performed when OCTA was inconclusive, ensuring robust determination of MNV status. Eyes were categorized as MNV-positive or MNV-negative accordingly.

Fundus autofluorescence (FAF) was acquired with a confocal scanning laser ophthalmoscope (SLO) (Heidelberg Engineering, Heidelberg, Germany) using a 30-degree field. Images were captured at 768 × 768-pixel resolution. Short-wavelength autofluorescence (SWAF, using a blue excitation wavelength of 488 nm) and near-infrared autofluorescence (NIRAF, using an excitation wavelength of 787 nm) were recorded in the same session. Hypoautofluorescent regions were manually contoured on NIRAF images by two masked graders (S.K. and Y.E.L) and discrepancies were resolved by consensus. To ensure methodological rigor, inter-grader reliability was assessed for lesion area measurements, yielding a good intraclass correlation coefficient of 0.82 (95% CI, 0.39–0.97) for lesion area measurements. Sublesional choroidal thickness (SLCT) was measured beneath the geometric center of each NIRAF lesion as the perpendicular distance between Bruch’s membrane and the choroidoscleral interface using OCT calipers embedded in Heidelberg Eye Explorer software.

Retinal sensitivity was measured using Compass microperimetry, employing mesopic tests on a 10 − 2 grid with 68 loci. Manually delineated hypoautofluorescent lesions on NIRAF were superimposed onto the microperimetry fundus image for analysis. The representative images are shown in Fig. 1. Retinal sensitivity values were categorized into two groups – inside and outside the lesion boundary – and the arithmetic mean retinal sensitivity (mRS) and the difference between the two groups (ΔmRS) for each zone were calculated. Central foveal mRS was defined as arithmetic mean value of central four points.

Fig. 1.

Multimodal imaging and microperimetry of the left eye in a 58-year-old female with pachychoroid pigment epitheliopathy (PPE). Best-corrected visual acuity was 0.10 logMAR. (a) Infrared reflectance (IR) fundus image and enhanced-depth imaging optical coherence tomography (EDI-OCT) B-scan through the fovea showing a thickened choroid and pachyvessels. (b) En face OCT angiography (OCTA) slabs (superficial, deep, outer retina, and choriocapillaris) demonstrate no evidence of macular neovascularization (MNV). (c) Near-infrared autofluorescence (NIRAF) image reveals a well-delineated area of granular hypoautofluorescence (yellow dashed line). (d) Fundus-guided microperimetry (Compass) superimposed on the NIRAF-defined lesion. Despite the marked RPE alterations seen on NIRAF, the retinal sensitivity values (dB) within the lesion remained relatively preserved in this case

Statistical analysis

Continuous variables are presented as mean ± standard deviation, and categorical variables as counts (percentages). BCVA was converted to logarithm of the minimum angle of resolution (logMAR). Between-group comparisons were performed using the Mann-Whitney U test, and within-group comparisons between retinal sensitivity inside and outside the lesion were performed using the Wilcoxon signed-rank test. To explore factors associated with greater lesion-associated retinal sensitivity reduction (higher ΔmRS), univariable linear regression analyses were performed for age, sex, symptom duration, sublesional choroidal thickness, and MNV status. Given the limited sample size, the adjusted analysis was restricted to an exploratory model including symptom duration and MNV status, which were selected based on clinical relevance and the strongest univariable signals. Statistical analyses were performed using SPSS version 23.0 (IBM Corp., Armonk, NY, USA) and p-value < 0.05 was considered statistically significant.

Results

A total of 26 eyes from 26 patients were included. The mean age was 62.5 ± 9.6 years (range, 45–78 years) and 58% were male. Mean symptom duration at presentation was 8.2 ± 7.4 months and the mean logMAR BCVA was 0.17 ± 0.32. The mean sublesional choroidal thickness measured on EDI-OCT was 348.2 ± 88.3 µm and the central foveal mean retinal sensitivity (mRS) was 28.8 ± 2.1 dB. Seven eyes (27%) were classified as CSC, six eyes (23%) as PPE, and 13 eyes (50%) as PNV. (Table 1).

Table 1.

Demographics and clinical characteristics of the study eyes

Parameters	Value
Eyes/Patients (n)	26/26
Age (years)	62.5 ± 9.6 (range, 45–78)
Sex (male/female)	15/11
Spherical equivalent (D)	0.23 ± 1.3 (range, -5.0-+3.0)
logMAR BCVA	0.17 ± 0.32
Symptom duration (months)	8.2 ± 7.4
Central foveal mRS (dB)	28.8 ± 2.1
SLCT (µm) PSD subtypes (n)	348.2 ± 88.3
CSC	7 (26.9%)
PPE	6 (23.1%)
PNV	13 (50.0%)

SD, Standard deviation; SLCT, Sublesional choroidal thickness; mRS, mean retinal sensitivity; PSD, Pachychoroid spectrum disease; CSC, Central serous chorioretinopathy; PPE, Pachychoroid pigment epitheliopathy; PNV, Pachychoroid neovasculopathy

Table 2 summarizes the linear regression analyses for factors associated with lesion-associated retinal sensitivity loss. In univariable analysis, only MNV status was significantly associated with higher ΔmRS (unstandardized B = 1.42; 95% CI, 0.32 to 2.52; p = 0.034). Symptom duration showed a borderline association (p = 0.081), whereas age, sex and SLCT were not significantly associated with ΔmRS. In the exploratory multivariable model including symptom duration and MNV status, the presence of MNV remained significantly associated with greater lesion-associated retinal sensitivity loss (unstandardized B = 1.55; 95% CI, 0.61 to 2.50; p = 0.032).

Table 2.

Univariable and exploratory multivariable linear regression analyses for factors associated with ΔmRS

Predictors	Univariable			Multivariable
	unstandardized B (95% CI)		P value	unstandardized B (95% CI)	P value
Age (years)	0.31 (-0.11 to 0.73)		0.594	-
Sex	0.46 (-0.59 to 1.51)		0.617	-
Symptom duration (months)	0.08 (-0.01 to 0.17)		0.081	0.04 (-0.18 to 0.26)	0.184
SLCT (per 10 μm)	0.02 (-0.14 to 0.18)		0.356	-
Presence of MNV	1.42 (0.32 to 2.52)		0.034	1.55 (0.61 to 2.50)	0.032

ΔmRS was defined as mean retinal sensitivity outside the NIRAF-defined lesion minus mean retinal sensitivity inside the lesion; higher values indicate greater lesion-associated retinal sensitivity reduction. Sex was coded as male = 1 and female = 0. SLCT was modeled per 10-µm increase

Across the cohort, mRS within the RPE lesions was significantly lower than that of mRS outside the RPE damage (27.58 ± 8.19 dB vs. 28.83 ± 6.65 dB, p = 0.034). When stratified by MNV status, logMAR BCVA, SLCT, and mRS outside the RPE damage did not differ between groups. In contrast, MNV-positive group showed a marked functional decrement (26.98 ± 7.14 dB vs. 29.11 ± 4.53 dB, p = 0.021), whereas the MNV-negative group demonstrated a smaller, non-significant difference (28.42 ± 4.61 dB vs. 28.50 ± 2.95 dB, p = 0.132; Table 3). Figure 4 shows the individual distribution of ΔmRS stratified by MNV status. MNV-positive eyes showed higher ΔmRS values than MNV-negative eyes, but individual values varied across the eyes (p = 0.044).

Table 3.

The difference in mean retinal sensitivity between the inside and outside of RPE alterations (detected by NIRAF) in eyes without MNV (MNV (-), n = 13) and with MNV (MNV (+), n = 13)

Eyes (n)	Total	MNV (-)	MNV (+)	P value
	26	13	13
BCVA (logMAR)	0.17 ± 0.32	0.14 ± 0.21	0.19 ± 0.24	0.067
SLCT (µm)	348.2 ± 88.3	312.3 ± 62.4	381.7 ± 103.4	0.350
mRS outside the lesion (dB)	28.83 ± 6.65	28.50 ± 2.95	29.11 ± 4.53	0.210
mRS inside the lesion (dB)	27.58 ± 8.19	28.42 ± 4.61	26.98 ± 7.14	0.031
ΔmRS (dB)	1.09 ± 1.81	0.08 ± 0.12	2.04 ± 1.24	0.044
P value		0.132*	0.021*

The P values in the rightmost column indicate between-group comparisons (MNV (-) vs. MNV (+)) using the Mann-Whitney U test. The asterisked P values in the bottom row indicate within-group comparisons between retinal sensitivity inside and outside the lesion using the Wilcoxon signed-rank test

Fig. 2.

Individual distribution of lesion-associated retinal sensitivity difference (ΔmRS) according to macular neovascularization (MNV) status. ΔmRS was defined as the mean retinal sensitivity outside the near-infrared autofluorescence (NIRAF)-defined lesion minus the mean retinal sensitivity inside the lesion. Higher values indicate greater lesion-associated retinal sensitivity reduction. Each dot represents one eye. Box plots indicate the median and interquartile range, with whiskers indicating the minimum and maximum values

Representative cases illustrated the spectrum of structure-function relationships in PSD. In some eyes (Figs. 1 and 2), NIRAF demonstrated hypoautofluorescent RPE lesions with variable functional impact. In one case, retinal sensitivity within the lesion remained relatively preserved despite marked NIRAF abnormalities, whereas in another case, reduced retinal sensitivity was observed within a well-delineated hypoautofluorescent lesion despite relatively preserved outer retinal architecture on OCT. These findings demonstrate heterogeneous spatial correspondence between structural alterations and focal functional changes in PSD.

Fig. 3.

Multimodal imaging and topographic structure-function mapping in a 73-year-old male patient with pachychoroid neovasculopathy (PNV). Best-corrected visual acuity was 0.00 logMAR. (a) En face OCT angiography (OCTA) slabs (superficial, deep, outer retina, and choriocapillaris) and corresponding OCT B-scans with flow overlays. A type 1 macular neovascularization (MNV) is clearly visualized on the choriocapillaris slab. (b) Near-infrared autofluorescence (NIRAF) image demonstrates a well-delineated area of granular hypoautofluorescence (yellow dashed line). (c) Fundus-guided microperimetry (Compass) superimposed on the NIRAF-defined lesion. In contrast to the PPE case in Fig. 1, this PNV case demonstrates a spatial correspondence between the NIRAF-defined lesion and focal retinal sensitivity reduction

Figure 3 provides an additional example of this heterogeneity. The right eye of a 49-year-old male showed a serous pigment epithelial detachment (PED) on OCT without evidence of MNV. SWAF revealed a hyperautofluorescent area corresponding to the PED, whereas NIRAF displayed only subtle mottling without a distinct hypoautofluorescent lesion. Microperimetry demonstrated relatively preserved retinal sensitivity within the PED area. Collectively, these representative cases illustrate the diversity of structure–function profiles across the PSD spectrum and support the complementary use of NIRAF and microperimetry for topographic functional assessment.

Fig. 4.

Multimodal imaging and microperimetry in a 49-year-old male with a serous pigment epithelial detachment (PED) without evidence of macular neovascularization (MNV). Best-corrected visual acuity was 0.10 logMAR. (a) Infrared reflectance (IR) fundus image and enhanced-depth imaging optical coherence tomography (EDI-OCT) B-scan showing a serous PED and thickened choroid. (b) En face OCT angiography (OCTA) slabs (superficial, deep, outer retina, and choriocapillaris) and corresponding B-scans show no detectable flow signal suggestive of MNV. (c) Short-wavelength autofluorescence (SWAF) image shows a hyperautofluorescent region corresponding to the PED. (d) Near-infrared autofluorescence (NIRAF) image displays only subtle mottling in the corresponding region. (e) Fundus-guided microperimetry (Compass) demonstrates relatively preserved retinal sensitivity within the PED area, despite the structural abnormality

Discussion

This study demonstrated that RPE alterations visualized as hypoautofluorescence on NIRAF were associated with reduced retinal sensitivity in inactive PSD, particularly in eyes with MNV. Retinal sensitivity within NIRAF-defined lesions was modestly lower than that in surrounding retina, and MNV status showed the most consistent association with greater lesion-associated sensitivity reduction in the exploratory adjusted analysis whereas BCVA, SLCT, and symptom duration did not. The individual distribution of ΔmRS further demonstrated inter-eye variability among MNV-positive eyes, indicating that lesion-associated structure-function correspondence was heterogeneous rather than uniformly robust. These observations extend prior structure-function reports in chronic CSC [5, 7, 14–16], by suggesting that NIRAF-defined RPE alterations may help localize areas of functional vulnerability across the broader pachychoroid spectrum.

In a cohort of resolved CSC, Kim et al. demonstrated that NIRAF revealed additional hypoautofluorescent foci that SWAF failed to detect, and these NIRAF lesions showed high consistency with window defects detected by FA, underscoring NIRAF’s superior sensitivity for subtle RPE changes [7]. Nakagawa et al. recently showed that distinct NIRAF and SWAF patterns in PNV indicate disease chronicity and NIRAF outperformed in localizing macular neovascularization [12]. Other studies have also shown that abnormal gray scale values on SWAF correlated with visual function, while comparable gray values on NIRAF did not [17, 18]. Consistent with prior reports, we hypothesize that in PSD, melanin-driven RPE alterations detected by NIRAF might represent an earlier stage of injury than lipofuscin-related changes on SWAF. Although this temporal sequence remains speculative, NIRAF abnormalities may identify areas associated with localized sensitivity reduction, particularly in eyes with MNV. The greater lesion-associated sensitivity reduction observed in MNV-positive eyes may be consistent with more extensive alteration of the RPE–photoreceptor complex, although this interpretation remains exploratory. Compass microperimetry may be useful for characterizing localized functional changes in this setting.

Frequent structure-function discordance was another key feature. Structural imaging with OCT, OCTA, and FAF did not always match functional impairment [19–21]. Some structural abnormalities showed preserved sensitivity, whereas localized sensitivity loss sometimes occurred in areas with near-normal structure. Unlike conditions such as pre-perimetric glaucoma where structural damage consistently precedes functional loss, PSD therefore cannot be assumed to follow a fixed sequence of structural change followed by functional loss.

The lack of association between SLCT and mRS in the present study supports emerging evidence that choroidal thickness per se is an imprecise surrogate for PSD progression [22]. Instead, changes such as vessel dilation or choriocapillaris attenuation may be more closely associated with overlying photoreceptor dysfunction than choroidal thickness alone [23]. Future work incorporating quantitative parameters such as choroidal vascularity index or swept-source OCT flow metrics may better clarify the microvascular determinants of structure-function in PSD.

Several limitations should be acknowledged. First, the retrospective, cross-sectional design and modest sample size rendered the multivariable analysis exploratory. To minimize the risk of model overfitting and instability, the model was strictly limited to MNV status and symptom duration. Potential confounders such as best-corrected visual acuity, lesion size, and foveal involvement were excluded, which remains a limitation that warrants validation in larger prospective cohorts. Second, while the mean lesion-associated retinal sensitivity loss was greater in eyes with MNV, a validated minimal clinically important difference for Compass microperimetry in inactive PSD has not yet been established. The approximately 2-dB lesion-associated difference observed in eyes with relatively preserved BCVA (mean 0.17 logMAR) should be interpreted as a potentially relevant localized functional difference rather than as a validated clinical decision threshold. Third, symptom duration was based on patient-reported history, which is susceptible to recall bias. Furthermore, the lack of data on the duration of quiescence or baseline functional status limits our ability to infer longitudinal functional decline or causal relationships. Fourth, although manual delineation of NIRAF lesions was found to be highly reproducible (ICC = 0.82), potential observer bias remains a factor. Finally, the absence of axial length data and the focus on retinal sensitivity rather than other visual symptoms, such as metamorphopsia, should be considered when interpreting our findings. Prospective, longitudinal studies with larger cohorts are necessary to validate these preliminary structure-function correlations.

In conclusion, NIRAF-based lesion mapping with microperimetry may provide a useful approach for assessing structure-function characteristics in PSD. Identification of subclinical MNV by multimodal imaging may be important when interpreting localized retinal sensitivity differences. Prospective longitudinal studies are required to validate these associations and clarify the temporal relationship between structural RPE damage and functional decline.

Author contributions

Conception and design: Seungmo Kim, Ye Eun Han, and Junyeop Lee: Analysis and interpretation: Seungmo Kim, Yea Eun Lee, Ye Eun Han, and Junyeop Lee: Data collection: Seungmo Kim, Yea Eun Lee, Ye Eun Han, and Junyeop Lee: Writing – original draft: Seungmo Kim, and Yea Eun Lee: Writing – review & editing: Ye Eun Han, and Junyeop Lee: Overall responsibility: Seungmo Kim and Junyeop Lee:

Funding

This research was supported by a grant of the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (grant number: RS-2024-00438689), and by the University of Ulsan Glocal University COMPaaS joint research project grant funded by the Korea Ministry of Education (MOE) (2025 − 0626).

Data availability

The datasets used and/or analyzed during this study can be made available by the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study protocol was reviewed and approved by the Institutional Review Board of Asan Medical Center (approval no. 2024 − 0304) and all procedures were conducted in accordance with the tenets of the Declaration of Helsinki. The approval included a waiver of informed consent for the chart review. The study does not include any animal experiments conducted by any of the authors. The requirement for informed consent was waived by the Institutional Review Board of Asan Medical Center due to the retrospective nature of the study and the use of anonymized clinical data.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.