A combined ex vivo fundus imaging–histology protocol for clinicopathological validation of human donor eyes with limited medical history Authors: Sarrvesa H Vijay Singh, Waiwai Yeap, Lay Khoon Too, Alexandra Allende, Geoffrey Hall, Alex Hunyor, Michele C Madigan, Rohan Merani, Juan Ortiz, Michael Rodriguez, James Wong, Alessandro Invernizzi, Murray Killingsworth, Victoria Pye, Hemal Mehta, Constantinos Petsoglou, Svetlana Cherepanoff Published: 2026 ABSTRACT AbstractThis study evaluated a combined ex vivo fundus imaging–histology protocol to improve accuracy of diagnosis in human donor eyes with limited medical history. Fifty-one formalin-fixed eyes from 32 donors underwent a standardised workflow comprising ex vivo fundus photography, selective spectral-domain optical coherence tomography (SD-OCT), and comprehensive histological examination. Medical/surgical retinal ophthalmologists independently reviewed imaging, while ophthalmic pathologists performed masked histological assessment. Diagnostic metrics were calculated using histology as the reference standard. Ex vivo imaging identified definitive pathology in 10 of 51 eyes (19.6%) and possible abnormalities in 4 eyes (7.8%). Histological analysis, however, revealed definitive pathology in 20 eyes (39.2%), detecting a broader range of diseases, including early age-related macular degeneration, hypertensive vasculopathy, and a rare choroidal tumour. Concordance was observed in 27 eyes (52.9%), mainly where no pathology was present. Significant discordance occurred in 24 eyes (47.1%), comprising 13 false negatives and 11 false positives (including misclassified pathology). Consequently, the sensitivity and specificity of ex vivo imaging for detecting pathology were 18.8% and 68.6%, respectively. While ex vivo imaging is a practical screening tool, imaging alone may miss or misclassify pathology due to postmortem artefacts and a lack of validated interpretive criteria. A combined imaging–histology approach validates the limited tissue available from eye-bank programs and maximises research value.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-58078-w.Keywords: Ex vivo imaging, Postmortem eyes, Ophthalmic pathology, Ex vivo validation, Clinicopathological correlationSubject terms: Diseases, Medical research, Oncology Introduction Human donor eyes are an essential resource for translational research in vision science and ophthalmology1. However, in practice, corneal donation programs represent the only reliable source of this tissue2. As these programs are resourced to prioritise corneal transplantation, retrieval often involves only the corneal button, making intact posterior eye cups rare3. Furthermore, donor anonymity legislation precludes access to in vivo clinical history or imaging4. Consequently, researchers are frequently compelled to include tissue in experimental protocols without prior clinicopathological validation, often relying on eyes with extended postmortem delays and no documentation to guide selection. To mitigate these limitations, some researchers have utilised ex vivo imaging to screen eyes for specific disease models, particularly age-related macular degeneration (AMD). However, this approach often relies on restricting selection to eyes with short postmortem intervals which is logistically impractical for large-scale research. While a small number of research‑committed eye banks have developed dedicated procurement protocols that substantially reduce post‑mortem interval, these represent specialised programmes dependent on research‑specific funding and investigator partnerships rather than standard operational models, and they supply only a minority of research tissue globally1–3. Moreover, short post‑mortem interval alone does not account for the wide range of pre‑mortem factors that influence intraocular tissue quality4. Validation studies have been limited to small case series, leaving much of the tissue recovered through corneal donor programs uncharacterised and unvalidated for research use5. Ex vivo fundus photography and, less commonly, optical coherence tomography (OCT) can be used to screen donor eyes for macroscopic abnormalities5,6. This approach provides a pragmatic solution to the lack of clinical data and allows researchers to exclude tissue with obvious advanced disease or structural disruption. Standardised interpretation criteria for ex vivo postmortem imaging remain limited and vary between research laboratories. Fixation related artefacts such as retinal detachment, corneal oedema, pigment displacement and RPE tears are common in this setting and further complicate interpretation. Importantly, many research programs do not incorporate imaging at all, and it is rarer still for eyes to undergo histological validation before selection for experimental workflows. The limited application of clinicopathological validation in research eyes is largely due to the absence of established protocols that combine ex vivo imaging with histology. To address this gap, we implemented a systematic ex vivo validation protocol combining macroscopic fundus imaging, selective OCT, and comprehensive histological analysis. We hypothesised that: (i)ex vivo fundus imaging alone may be insufficient to detect the full range of pathological findings; ex vivo fundus imaging alone may be insufficient to detect the full range of pathological findings; (ii)histological validation can either confirm or expand upon abnormalities identified by ex vivo imaging, revealing additional or subtler disease processes; and. histological validation can either confirm or expand upon abnormalities identified by ex vivo imaging, revealing additional or subtler disease processes; and. (iii)a combined ex vivo imaging–histology protocol could provide a reproducible framework for accurate and reliable classification of research eye tissue. a combined ex vivo imaging–histology protocol could provide a reproducible framework for accurate and reliable classification of research eye tissue. To our knowledge, a systematic, cohort-wide validation of eye bank–derived research tissue integrating ex vivo imaging and histology has not previously been undertaken. Published protocols have focused mainly on ex vivo imaging to identify or enrich for specific diseases—most often age-related macular degeneration5 and none have incorporated comprehensive histological validation across all available donor tissue. This study is the first to implement a comprehensive imaging–histology protocol for an unselected series of corneal eye bank–derived eyes. By establishing a logistically feasible workflow that leverages the complementary expertise of retinal specialists and pathologists, this study provides a realistic framework for evaluating the true diagnostic yield and research value of donor tissue obtained under real-world conditions. Methods Ethics statement The research protocol was approved by the South Eastern Sydney Local Health District Human Research Ethics Committee (HREC Reference: 14/215) and adhered to the tenets of the Declaration of Helsinki. Cohort selection Human donor eyes were sourced from the Lions New South Wales (NSW) Eye Bank and Tissue Bank, Australia. Eligible donors were aged 18 years or older and only donors with explicit informed consent for research use were included. In keeping with real-world eye bank practice, retrieval and fixation of donor eyes were determined by corneal transplant priorities rather than research protocols. As a result, postmortem and fixation delays were variable—sometimes prolonged (Supplementary Table S1). This variation in postmortem delay reflects the realities of tissue availability for research from corneal donors, which is already limited as full enucleation is uncommon across eye banks and cultural factors worldwide already limit the pool of available transplant donors. All eyes were received after removal of the corneas and fixation in 10% neutral buffered formalin. To reflect real-world eye bank donor tissue, no exclusion criteria were applied, and no limit was imposed on the postmortem interval prior to fixation. Fifty-one eyes from 32 donors were included. Both eyes were not always available for this study because eye bank allocation prioritised corneal transplantation and, where tissue was not used for transplant, globes were sometimes allocated to other teams on the eye bank research waiting list. Each eye was identified by donor ID and accompanied by a donor information sheet containing basic demographic and clinical details: age, sex, time and date of death, time of enucleation and preservation, cause of death, and limited medical history retrieved from hospital notes or provided by the senior next of kin. Comprehensive ophthalmic history was unavailable. Ex vivo imaging A unique study ID was assigned to each eye on receipt. The iris and lens were removed if present. Each eye cup was rinsed with tap water, transferred to a fused quartz crucible filled with 70% ethanol, and photographed for macroscopic documentation. All eyes underwent ex vivo fundus photography as the primary imaging modality. Ex vivo OCT was performed selectively, limited to eyes where macroscopic findings were equivocal or where two independent ophthalmologist reviewers held differing interpretations. Fundus photography was performed using an SZ61TR stereomicroscope and DP26 digital camera (Olympus Corporation, Tokyo, Japan), with image capture managed using cellSens Standard 1.1 software (Olympus Corporation). The eye remained submerged in ethanol during imaging. For select eyes, SD-OCT imaging was performed. The eye cup was placed in a bespoke holder and submerged in 70% ethanol. Imaging was conducted using the Spectralis OCT2 85kHZ module on the Spectralis imaging platform (Heidelberg Engineering GmbH, Heidelberg, Germany). Horizontal A scans including Enhanced Depth Imaging (EDI) scans in a dense raster at high resolution (1024 × 49) were taken through the macula area and through areas of potential pathology identified on ex vivo imaging. Ex vivo fundus and OCT assessment Ex vivo fundus photographs for each donor eye was independently reviewed by at least two ophthalmologists with subspecialty training in medical (JW, RM, and AI) or surgical (AH) retina. Where imaging findings were discordant or indeterminate, cases were discussed at a consensus meeting. Ex vivo OCT was performed selectively as a problem‑solving modality rather than a screening tool. It was used in eyes where fundus findings were equivocal or where disagreement occurred between clinical reviewers. Final imaging classifications were recorded prior to histological processing and assessment. Imaging assessment and histological evaluation were conducted in separate stages to minimise observer bias. Imaging classification criteria Ex vivo imaging findings were classified into three predefined categories: definitive pathology, possible abnormality, and postmortem change/normal. Definitive pathology was defined as a structural abnormality on ex vivo fundus photography or OCT consistent with recognised retinal or choroidal disease. Examples included chorioretinal scars, choroidal masses, geographic atrophy–like lesions, and clearly demarcated areas of retinal atrophy. Possible abnormality referred to subtle or equivocal findings that could represent pathology but might also reflect postmortem artefact or fixation-related changes. Examples included focal pigment abnormalities, suspected gliosis or Bergmeister papilla remnants, and poorly defined areas of chorioretinal alteration. Postmortem change/normal included findings attributable to fixation or postmortem processes rather than disease. These included retinal folds, vascular congestion, retinal detachment, RPE separation, and other artefacts commonly encountered in postmortem ocular tissue. Initial classifications were made independently by retinal specialists and subsequently finalised through consensus discussion where discrepancies occurred. Histological processing and assessment After consensus on ex vivo imaging findings, each eye was bisected in the horizontal plane through the macula and optic nerve and embedded in paraffin in its entirety. Serial sections were cut from both blocks, ensuring systematic assessment of the macula while also allowing evaluation of all other intraocular tissues. Histological evaluation was therefore not limited to targeted macular sampling but encompassed the full bisected globe, ensuring that histological assessment remained independent of imaging classification. Formalin-fixed, paraffin embedded (FFPE) serial sections of both blocks were stained using haematoxylin and eosin (H&E), astral blue/periodic acid–Schiff (AB/PAS), and Masson’s trichrome. All histopathological assessments for this study were conducted by pathologists using physical glass slides. Whole-slide scanning was performed for digital archiving of case slides using a SLIDEVIEW VS200 high-throughput slide scanner (Olympus Corporation) with 60x oil immersion objective lens, using brightfield one-plane scan mode. Scanned slides were archived and viewed using the OlyVIA software platform (Olympus Corporation). For each eye, a minimum of four stained FFPE sections per block were independently reviewed by at least two ophthalmic pathologists (AA, GH, JO, and SC), all blinded to imaging findings. Additional levels or targeted sections were examined when abnormalities were suspected on ex vivo imaging. In cases of discordant histological interpretation, a third pathologist conducted an independent review, and consensus was subsequently reached through discussion at pathology review meetings. Histological classification and mapping to imaging categories Histological diagnoses were established by consensus among ophthalmic pathologists using standard histopathological criteria. For age-related macular degeneration (AMD), classification followed clinicopathological criteria established in prior studies of macular degeneration5,6. Early AMD was defined by the presence of continuous basal laminar deposit (BLamD) at the macula, consistent with established histopathological definitions of early AMD. For the purposes of diagnostic performance analysis, histological findings were grouped into two categories: Pathology present Retinal, RPE, choroidal, or scleral disease processes identified on histology, including early AMD, retinal layer atrophy, inflammatory or fibrotic lesions, tumour-related pathology, vascular lesions, and other disease processes. No pathology (postmortem change only) Eyes showing only non-specific postmortem artefacts without identifiable disease. Clinicopathological consensus Cases with discrepancies between ex vivo imaging and histological findings were reviewed in multidisciplinary clinicopathological meetings. Participants included the original ophthalmologists and pathologist assessors and a neuropathologist (MR). Research scientists (MM and MK) attended these meetings as part of the research team. A flowchart of the full clinicopathological workflow is presented in Fig. 1. Clinicopathological validation workflow for donor eyes. 51 eyes were accessioned and prepared for ex vivo fundus imaging. Fundus photos were independently assessed by two retinal subspecialists. Discordant cases proceeded to a clinical consensus meeting, and ex vivo OCT was performed on a subset of eight eyes requiring further clarification. All eyes then underwent histological processing, with at least two anatomical pathologists independently assessing a minimum of eight stained sections per eye. Histological classifications were reviewed for concordance. Cases with discrepant ex vivo imaging and histology classifications underwent multidisciplinary clinicopathological review. This multi-stage process resulted in a final integrated clinicopathological consensus diagnosis for each eye. Data management and analysis All donor data, including donor sheet information, ex vivo fundus and OCT images and microscopic photographs of histology slides, were stored in a FileMaker Pro database (Claris International Inc., Santa Clara, CA, USA) for assessment. Ex vivo fundus and OCT assessments, histology assessments, and final classifications were similarly recorded in FileMaker Pro (Claris International Inc.) and exported to Microsoft Excel (Microsoft Corp., Redmond, WA, USA) for analysis. Descriptive statistics were used to summarise frequencies and proportions. Concordance between ex vivo imaging (clinical) and histology (pathological) consensus classifications was assessed by cross-tabulation. Diagnostic performance of ex vivo imaging was evaluated using histology as the reference standard. For the assessment of concordance, findings recorded as ‘possible’ were pooled with ‘definite’ findings to represent a positive result. Additionally, imaging findings that were discordant with the histological diagnosis (e.g., misclassified pathology) were categorised as false positives. Performance metrics included sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), diagnostic odds ratio (DOR), and overall accuracy. Sensitivity, specificity, PPV, NPV, accuracy, and their 95% confidence intervals were calculated using the Wilson score method for binomial proportions. Diagnostic odds ratios were calculated with 95% confidence intervals on the log scale. Importantly, the final clinicopathological consensus diagnoses were not used in this analysis to avoid circularity. Formal interobserver agreement statistics were not calculated because discrepancies between reviewers were resolved through structured consensus review. The aim of the study was to assess concordance between ex vivo imaging and histological findings rather than to quantify observer variability. The datasets generated and analysed during the study are available from the corresponding author on reasonable request. Results Cohort characteristics A total of 51 eyes from 32 donors were included in the cohort (Table 1). The mean donor age was 67 years (95% CI: 64–80), with a predominance of males (M: F = 23:9). Cancer was the most common medical history (81.3%) and the leading cause of death (71.9%). Most donors passed away in a hospital or health facility. Table 1Summary of demographics, medical history and ophthalmic history. Cohort Overview Total Number of Donors (n =)32Total Number of Eyes (n =)51 Demographics Mean Donor Age (Years)67 (32–81), (95% CI: 64–80)Donor Sex (M: F)23:9 Medical history Donor Cause of Death*CancerCardiovascular diseaseCerebrovascular diseasen (% of donors)23 (71.9%)8 (25.0%)1 (3.1%)Donor Medical HistoryCancerCardiovascular diseaseCerebrovascular diseaseRespiratory diseaseRenal diseaseObesityDiabetesOther^n (% of donors)26 (81.3%)24 (75.0%)1 (3.1%)15 (46.9%)8 (25.0%)3 (9.4%)8 (25.0%)21 (65.6%) Donor eye tissue Mean Time Between Donor Death & Eye Enucleation (hours)12 (95% CI: 11–164)Mean Time Between Donor Death & Eye Tissue Preservation (hours)14 (95% CI: 13–16)Ophthalmic historyNilBrief/ Non-SpecificCataractViral Conjunctivitisn (% of eyes)23 (45.1%)20 (39.2%)7 (13.7%)1 (2.0%)The cohort comprised 51 eyes from 32 donors, representing a random subset of eye bank donors who consented to research use of residual tissue following cornea removal. The mean donor age was 67 years (95% CI: 64–80), with a male-to-female ratio of 23:9. Cancer was the most common cause of death (71.9%). No ophthalmic history was documented for 23 eyes (45.1%). Among those with available information, 20 eyes (39.2%) had only brief or non-specific entries, while specific diagnoses included cataract (13.7%) and viral conjunctivitis (2.0%).*Two donors had multiple causes of death.^Other diseases include gastrointestinal, musculoskeletal, endocrine, infectious, haematological, neurological, psychiatric conditions, and trauma. Summary of demographics, medical history and ophthalmic history. Donor Cause of Death* Cancer Cardiovascular disease Cerebrovascular disease n (% of donors) 23 (71.9%) 8 (25.0%) 1 (3.1%) Donor Medical History Cancer Cardiovascular disease Cerebrovascular disease Respiratory disease Renal disease Obesity Diabetes Other^ n (% of donors) 26 (81.3%) 24 (75.0%) 1 (3.1%) 15 (46.9%) 8 (25.0%) 3 (9.4%) 8 (25.0%) 21 (65.6%) Ophthalmic history Nil Brief/ Non-Specific Cataract Viral Conjunctivitis n (% of eyes) 23 (45.1%) 20 (39.2%) 7 (13.7%) 1 (2.0%) The cohort comprised 51 eyes from 32 donors, representing a random subset of eye bank donors who consented to research use of residual tissue following cornea removal. The mean donor age was 67 years (95% CI: 64–80), with a male-to-female ratio of 23:9. Cancer was the most common cause of death (71.9%). No ophthalmic history was documented for 23 eyes (45.1%). Among those with available information, 20 eyes (39.2%) had only brief or non-specific entries, while specific diagnoses included cataract (13.7%) and viral conjunctivitis (2.0%). *Two donors had multiple causes of death. ^Other diseases include gastrointestinal, musculoskeletal, endocrine, infectious, haematological, neurological, psychiatric conditions, and trauma. Medical history retrieval relied on eye bank technicians reviewing hospital records, which often lacked ophthalmic details. No ophthalmic history was documented for 23 eyes (45.1%), while 20 eyes (39.2%) had brief or non-specific entries (Table 1). Specific diagnoses included cataract (13.7%) and viral conjunctivitis (2.0%). These limitations reflect the difficulty of obtaining reliable and detailed ophthalmic history in anonymised cornea bank donor cohorts. Ex vivo imaging findings Ex vivo fundus examination of 51 eyes from 32 donors revealed definitive pathology in 10 eyes (19.6%) from 7 donors (21.9%). Ex vivo OCT was performed in eight eyes where fundus findings were equivocal or where reviewer disagreement occurred. Of these, 5 eyes had pathology confirmed on OCT. The majority of eyes (37, 72.6%) exhibited only postmortem changes, such as artefactual retinal detachment and RPE tears, none of which were routinely assessed with OCT. Possible fundus abnormalities were identified in 4 eyes (7.8%) from 3 donors (9.4%), with gliosis/Bergmeister papilla confirmed on OCT in 2 eyes. Paired eyes from some donors displayed inter-eye variability, including cases where one eye showed definitive pathology and the other exhibited only postmortem changes (Table 2). Table 2Ex Vivo Imaging Findings.Ex vivo Imaging FindingsDiagnostic ConfidenceNumber of Eyes (n)% Total Cohort(n = 51)OCT Confirmed (n)% Total Cohort(n = 51) Retina / RPE-Related Geographic AtrophyDefinitive23.9%23.9%RPE Pigment AbnormalitiesDefinitive12.0%00.0%RPE TearDefinitive12.0%00.0%RPE Pigment AbnormalitiesPossible12.0%00.0% Chorioretinal Pathology Chorioretinal Atrophy / ScarDefinitive23.9%23.9% Choroidal Pathology Choroidal PigmentationDefinitive23.9%00.0%Choroidal MassDefinitive23.9%12.0%Choroidal MetastasisPossible12.0%00.00% Other Vitreoretinal Findings Gliosis / Bergmeister PapillaPossible23.9%23.9% Post-Mortem Changes Only (Normal) 3772.5%12.0% Totals Definitive Abnormal Findings 10 19.6% 5 9.8% Possible Abnormal Findings 4 7.8% 2 3.9% Post-Mortem Changes Only 37 72.5% 1 2.0% Overall Total 51 100.0% 8 15.7% Ex vivo fundus examination identified definitive pathology in 10 eyes (19.6%) from 7 donors, with 5 cases confirmed definitive pathology on OCT. Most eyes (37; 72.6%) showed only postmortem changes, which were not routinely assessed with OCT. Possible fundus abnormalities were noted in 4 eyes (7.84%) from 3 donors, with gliosis/Bergmeister papilla being the only OCT confirmed finding . Ex Vivo Imaging Findings. Ex vivo fundus examination identified definitive pathology in 10 eyes (19.6%) from 7 donors, with 5 cases confirmed definitive pathology on OCT. Most eyes (37; 72.6%) showed only postmortem changes, which were not routinely assessed with OCT. Possible fundus abnormalities were noted in 4 eyes (7.84%) from 3 donors, with gliosis/Bergmeister papilla being the only OCT confirmed finding . Histological findings Histological analysis (Table 3) identified definitive pathology in 20 of 51 eyes (39.2%), providing a diagnostic baseline for the cohort. Two eyes (3.9%) were classified as having possible findings. The remaining 29 eyes (56.9%) were confirmed as histologically normal, exhibiting only non-specific postmortem changes such as RPE detachment, RPE tears and retinal folding. Importantly, histological processing reduced diagnostic ambiguity compared to macroscopic examination, limiting the number of possible cases to just these two instances. Table 3Histological findings (n = 51 eyes from 32 donors).Histological FindingsDiagnostic ConfidenceNumber of Eyesn (% of cohort)Number of Donorsn (% of cohort) Retinal Disease Early age-related macular degenerationDefinitive2 (3.9%)2 (6.3%)Outer plexiform layer atrophy, epiretinal membrane, subretinal membraneDefinitive1 (2.0%)1 (3.1%)Chorioretinal scarDefinitive1 (2.0%)1 (3.1%)Small hard drusen, numerousDefinitive2 (3.9%)1 (3.1%)Outer retinal atrophy with minimal RPE abnormalitiesDefinitive2 (3.9%)1 (3.1%)Photoreceptor atrophyDefinitive2 (3.9%)2 (6.3%)Subretinal exudateDefinitive1 (2.0%)1 (3.1%) Ischaemic Changes Cytoid bodiesDefinitive2 (3.9%)2 (6.3%) Tumour-related changes Choroidal ganglioneuromaDefinitive2 (3.9%)1 (3.1%)Metastatic adenocarcinomaDefinitive2 (3.9%)2 (6.3%) Age-related changes Scleral calcificationDefinitive3 (5.9%)2 (6.3%)Hypertensive vasculopathyPossible2 (3.9%)2 (6.3%)Postmortem Changes Only (Normal)29 (56.9%)21* Totals Definitive Pathology 20 (43.1%)16* Possible Pathology 2 (3.9%)2* Postmortem Changes Only (Normal) 29 (56.9%)21* Overall Total 51 (100.0%)39*Histology identified definitive pathology in 20 eyes (39.2%) from 16 donors (50.0%), nearly doubling the diagnostic yield compared to ex vivo imaging alone. Retinal disease was the most frequent category, including early AMD, layer-specific retinal atrophy, and inflammatory or fibrotic sequelae such as epiretinal or subretinal membranes. Postmortem changes without identifiable microscopic pathology were present in 29 eyes (56.9%), and bilateral concordance was variable, with inter-eye differences observed in 14 eyes from 7 donors.*The donor total (n = 39) in this table exceeds the cohort total (n = 32) because 7 donors contributed eyes with differing histological findings. In these 14 eyes, one eye showed only postmortem changes while the fellow eye exhibited pathology, or different pathology was found in each eye. Histological findings (n = 51 eyes from 32 donors). Histology identified definitive pathology in 20 eyes (39.2%) from 16 donors (50.0%), nearly doubling the diagnostic yield compared to ex vivo imaging alone. Retinal disease was the most frequent category, including early AMD, layer-specific retinal atrophy, and inflammatory or fibrotic sequelae such as epiretinal or subretinal membranes. Postmortem changes without identifiable microscopic pathology were present in 29 eyes (56.9%), and bilateral concordance was variable, with inter-eye differences observed in 14 eyes from 7 donors. *The donor total (n = 39) in this table exceeds the cohort total (n = 32) because 7 donors contributed eyes with differing histological findings. In these 14 eyes, one eye showed only postmortem changes while the fellow eye exhibited pathology, or different pathology was found in each eye. Among the 20 eyes with definitive pathology, retinal pathology was the most frequent finding, observed in 11 eyes (21.6%). Other definitive findings included tumour-related changes (7.8%), age-related changes (5.9%), and ischaemic changes (3.9%). The two equivocal cases mentioned above displayed features suggestive of possible hypertensive vasculopathy. Microscopic examination identified specific pathological features known to be occult to macroscopic observation. These included continuous basal laminar deposit (BLamD) characteristic of early AMD, focal cytoid bodies without associated clinical cotton wool spots, and selective retinal layer atrophy. These features were present even in eyes that appeared macroscopically unremarkable, underscoring the necessity of microscopic validation for accurate tissue characterisation. Concordance: ex vivo imaging vs. histology Overall, diagnostic concordance between ex vivo imaging and histology was observed in 27 of 51 eyes (52.9%). To rigorously evaluate diagnostic utility, imaging grades recorded as “possible” were pooled with definitive findings; these were classified as discordant if they were not confirmed by the opposing modality. A Stuart-Maxwell test for marginal homogeneity confirmed a significant divergence in the distribution of findings between the two methods (χ2 = 33.9, p < 0.0001), indicating that ex vivo imaging and histology do not yield equivalent diagnostic classifications. As the distribution of errors was balanced between false positives (n = 11) and false negatives (n = 13), no statistically significant bias was observed between the two modalities. The majority of concordant cases (24 of 51 eyes; 47.1%) consisted of normal eyes exhibiting only postmortem changes (Table 4). However, for the detection of definitive pathology, ex vivo imaging demonstrated low sensitivity, agreeing with histological findings in only 3 cases (5.9%). These concordant pathology cases involved gross structural abnormalities: a chorioretinal scar (Figure 3) and choroidal masses (Figure 2) confirmed as metastatic adenocarcinoma. This suggests that while ex vivo fundus examination can reliably detect large or structurally obvious lesions, it frequently fails to capture subtle disease. Table 4Concordance and discordance between ex vivo fundus imaging and histological findings (n = 51 donor eyes, 32 donors).Ex vivo Imaging FindingsHistological FindingsNumber of Eyesn (% of cohort)Number of Donorsn (% of cohort) Concordant cases Normal on ex vivo imaging and histology (total) 24 (47.1%)19 (59.4%)Postmortem Changes OnlyPostmortem changes only24 (47.1%)19 (59.4%) Pathology on ex vivo imaging (total) 3 (6.0%)3 (9.4%)Chorioretinal Atrophy / ScarChorioretinal scar1 (2.0%)1 (3.1%)Choroidal MassMetastatic adenocarcinoma1 (2.0%)1 (3.1%)Choroidal Mass *Metastatic adenocarcinoma1 (2.0%)1 (3.1%) Total Concordant Cases 27 (52.9%) 22 (68.8%) ^ Discordant cases Pathology missed on ex vivo imaging, total13 (25.5%)10 (31.3%)Postmortem Changes OnlyCytoid bodies2 (4.0%)1 (3.1%)Postmortem Changes OnlyEarly age-related macular degeneration2 (3.9%)2 (6.3%)Postmortem Changes OnlyAtrophy of the plexiform layers, epiretinal membrane and subretinal fibrocellular membrane1 (2.0%)1 (3.1%)Postmortem Changes OnlySubretinal exudate1 (2.0%)1 (3.1%)Postmortem Changes OnlyPhotoreceptor atrophy2 (3.9%)2 (6.3%)Postmortem Changes OnlyScleral calcification3 (5.9%)2 (6.3%)Postmortem Changes OnlyHypertensive vasculopathy2 (3.9%)1 (3.1%)Overcalled/misclassified on ex vivo imaging, total11 (21.6%)8 (25.0%)Geographic atrophyOuter retinal atrophy without RPE atrophy2 (3.9%)1 (3.1%)Chorioretinal atrophyPostmortem changes only (normal)1 (2.0%)1 (3.1%)Pigment abnormalities *Choroidal ganglioneuroma1 (2.0%)1 (3.1%)RPE tear *Choroidal ganglioneuroma1 (2.0%)1 (3.1%)Gliosis/Bergmeister papillaSmall hard drusen2 (3.9%)1 (3.1%)Choroidal pigmentation *Postmortem changes only (normal)2 (3.9%)1 (3.1%)Pigment abnormalities*Postmortem changes only (normal)1 (2.0%)1 (3.1%)Choroidal metastasis*Postmortem changes only (normal)1 (2.0%)1 (3.1%) Total Discordant Cases 24 (47.1%) 18 (56.3%) ^ Ex vivo imaging findings were concordant with histology in 27 eyes (52.9%) obtained from 22 donors (68.8%). The majority of concordant cases were eyes without pathology (24 eyes; 47.1%), while only 3 cases (5.9%) had concordant pathological diagnoses on both imaging and histology. Discordance was observed in nearly half the cohort: 13 eyes (25.5%) from 10 donors (31.3%) were classified as normal on imaging despite showing definitive pathology on histology (False Negatives). Conversely, 11 eyes (21.6%) from 8 donors (25.0%) were classified as False Positives; these were incorrectly identified as abnormal on imaging due to either overestimation of postmortem changes or diagnostic misclassification of pathology.^* OCT not performed. ^ The total number of donors reported in this table exceeds the unique study cohort (n = 32) due to inter-eye variability; in some cases, paired eyes from a single donor yielded different outcomes (e.g., concordant in one eye and discordant in the fellow eye), causing the donor to be counted in multiple categories. Concordance and discordance between ex vivo fundus imaging and histological findings (n = 51 donor eyes, 32 donors). Ex vivo imaging findings were concordant with histology in 27 eyes (52.9%) obtained from 22 donors (68.8%). The majority of concordant cases were eyes without pathology (24 eyes; 47.1%), while only 3 cases (5.9%) had concordant pathological diagnoses on both imaging and histology. Discordance was observed in nearly half the cohort: 13 eyes (25.5%) from 10 donors (31.3%) were classified as normal on imaging despite showing definitive pathology on histology (False Negatives). Conversely, 11 eyes (21.6%) from 8 donors (25.0%) were classified as False Positives; these were incorrectly identified as abnormal on imaging due to either overestimation of postmortem changes or diagnostic misclassification of pathology. ^* OCT not performed. ^ The total number of donors reported in this table exceeds the unique study cohort (n = 32) due to inter-eye variability; in some cases, paired eyes from a single donor yielded different outcomes (e.g., concordant in one eye and discordant in the fellow eye), causing the donor to be counted in multiple categories. Concordant ex vivo imaging and histological findings of a chorioretinal scar. A. Epi- and trans-illuminated fundus image of the left eye from a 68-year-old male (DID: 1954) donor, showing a well-demarcated chorioretinal scar nasal to the optic disc (arrow). B. Corresponding ex vivo OCT reveals increased reflectivity in an area of marked retinal thinning, with loss of normal retinal lamination in the affected region (arrow). C–D. Low (C) and intermediate (D) magnification H&E-stained paraffin sections confirm retinal atrophy (arrow), with loss of normal retinal layers. E. High-magnification view confirms retinal gliosis and atrophy (arrow), with complete loss of the RPE monolayer, consistent with a longstanding chorioretinal scar. Concordant ex vivo imaging and histological findings of choroidal metastases. A. Epi- and trans-illuminated fundus photographs of the right eye from a 57-year-old male donor (DID:1658), showing a pale, raised choroidal mass in the superotemporal quadrant (arrow). C. Epi-illuminated fundus photograph of the left eye from a 75-year-old male donor (DID:1804), demonstrating a large, multinodular, pale choroidal mass (arrow) extending from the juxtapapillary region toward the superior macula. D. Corresponding ex vivo OCT from the donor in C shows elevation of the retinal pigment epithelium (RPE) consistent with an underlying choroidal mass (blue arrows). E. Low-magnification H&E paraffin section from the donor in (A) reveals a cellular choroidal tumour composed on higher magnification of epithelial sheets (F) and glandular structures (G), consistent with metastatic adenocarcinoma. Both donors had a documented history of metastatic adenocarcinoma (oesophageal and colorectal, respectively). Despite the lack of systematic bias, clinically significant discrepancies were noted in both directions. In 13 eyes (25.5%), imaging yielded a False Negative result, classifying the tissue as having only postmortem changes when histology revealed underlying pathology (Table 4). These missed diagnoses included subtle findings such as early AMD, cytoid bodies, and retinal layer-specific atrophy. Conversely, 11 eyes (21.6%) were classified as False Positives. This category encompassed two types of diagnostic error: overestimation and misclassification. In five cases, lesions suspected on imaging—such as chorioretinal atrophy, choroidal pigmentation, or suspected metastasis—were not confirmed histologically and were identified as postmortem artefacts or normal tissue. In the remaining six cases (misclassified pathology), imaging correctly identified a lesion but assigned an incorrect diagnosis. For example, areas suspected to be geographic atrophy or RPE tears were reclassified on histology as outer retinal atrophy or choroidal ganglioneuroma, respectively. Subtle lesions not identifiable on ex vivo fundus exam Histological analysis revealed a range of clinically significant pathologies in eyes initially classified as having “postmortem changes only” on ex vivo imaging. These included retinal ischaemic changes (cytoid bodies), photoreceptor atrophy, and more complex lesions involving subretinal exudates and combined epiretinal and subretinal fibrous membranes with atrophy of the plexiform layers. Notably, early AMD (Fig. 4), characterised histologically by thin, continuous basal laminar deposit (BLamD)5, often manifests as a clinically normal fundus6 and can therefore be subclinical on in vivo imaging. Discordant ex vivo imaging and histology- early AMD. A–B. Epi- (A) and trans-illuminated (B) fundus photographs of the left eye from a 79-year-old male donor (DID: 1665) show no significant abnormalities. The retinal fold near the macula and congested vasculature are normal postmortem artefacts. C–D. H&E-stained macular sections show no pathology at low magnification (C). At intermediate magnification (D), a granular eosinophilic deposit is visible between the RPE BrM (arrow), consistent with basal laminar deposit (BLamD). E–F. Masson trichrome–stained sections confirm the presence of sub-RPE material, faintly visible at low magnification (E) and clearly defined at intermediate magnification (F). G–H. High-power magnification confirms BLamD as a granular, eosinophilic sub-RPE deposit on H&E (G) and as a granular, green-staining sub-RPE deposit on Masson trichrome (H). These findings indicate thin, continuous macular BLamD, the defining histological lesion of early AMD. Abbreviations: RPE – retinal pigment epithelium; BrM – Bruch’s membrane; BLamD – basal laminar deposit; AMD – age-related macular degeneration. In one donor, histology revealed bilateral choroidal ganglioneuroma, while ex vivo fundus imaging showed only non-specific findings—namely pigment abnormalities and an artefactual RPE tear (Fig. 5). This rare, often asymptomatic lesion may be subtle even on in vivo imaging and may require histological confirmation via biopsy for diagnosis. The bilateral nature of the finding is strongly suggestive of neurofibromatosis type 17–9, illustrating how postmortem histological validation can uncover rare and otherwise inaccessible pathological entities, offering valuable material for research. Discordant ex vivo imaging and histology- bilateral choroidal ganglioneuroma. A-B. Epi- and transilluminated ex vivo fundus photos of the right (A) and left (B) eyes from a 56-year-old male donor (DID 1963 & 1964) show no significant abnormalities. A small RPE tear in the left eye is a postmortem artefact. Small retinal haemorrhages in both eyes are non-specific. C. Low-power H&E-stained section of the choroid shows no increase in thickness but reveals increased cellularity. D–E. Intermediate magnification demonstrates a spindle cell proliferation arranged parallel to the sclera, admixed with ganglion cells (arrowheads) and scattered mast cells (arrow), shown again in (E). F. Ganglion cells (arrowheads) are conspicuous within the choroidal spindle cell proliferation. These findings, diffusely involving the choroid in both eyes, are consistent with choroidal ganglioneuroma. Bilaterality suggests an underlying genetic syndrome such as neurofibromatosis type 1. Additionally, scleral calcification, is not detectable on fundus examination even in vivo, and would require B-scan ultrasonography or orbital imaging for diagnosis. These findings highlight the limitations of ex vivo fundus and OCT assessment alone in detecting subtle, early, or deep structural lesions, and emphasise the importance of histological validation to avoid misclassification of donor eyes in research cohorts. Misclassification of advanced pathology on ex vivo Imaging Even lesions that appeared obvious on ex vivo imaging were subject to misclassification. Two eyes (3.9%) from the same donor demonstrated clearly abnormal fundus findings (Fig. 6A–B) and were initially diagnosed as bilateral geographic atrophy (complete outer retinal and RPE atrophy; advanced dry AMD). However, histological analysis revealed near complete photoreceptor atrophy in the absence of RPE atrophy. The absence of any BLamD definitively excludes AMD (Fig. 6E–G). A-D. Discordant ex vivo imaging and histology- cORA. A-B. Epi- and transilluminated ex vivo fundus photos of the right (A) and left (B) eye from a 70-year-old male donor (DID:1668 & 1669), classified as bilateral geographic atrophy (cRORA). C-D. Ex vivo OCT confirmed the area of atrophy (arrows) in the right (C) and (D) eye (boxed). E. At intermediate magnification, H&E-stained paraffin section of the macula shows an intact layer of RPE (arrowheads). Loss of the outer nuclear (photoreceptor) layer, is abrupt (arrow). F. Higher magnification confirms well demarcated loss of the outer nuclear layer (bounded by arrows). A few pigment laden epithelioid cells (likely melanophagocytic microglia) are seen in the subretinal space. G. The RPE appear as an intact monolayer (arrowhead) in the area of outer nuclear and outer plexiform layer atrophy (boxed). No basal laminar deposit (AMD specific) is seen. There is some RPE hypertrophy to the right of the image. The histological findings are consistent with outer retinal atrophy without RPE atrophy (cORA), and compatible with photoreceptor atrophy of hereditary or non-AMD origin rather than advanced atrophic AMD (cRORA). These histological findings prompted retrospective review of the OCT from both eyes during a clinicopathological consensus meeting, where the RPE was confirmed to be intact (C, D; arrows). These findings are more consistent with complete outer retinal atrophy (cORA) rather than complete RPE and outer retinal atrophy (cRORA), the defining lesion of geographic atrophy10. This distinction raises alternative diagnostic considerations, including inherited retinal diseases and vitamin A deficiency—all of which can produce bilateral, symmetrical fundus changes that mimic geographic atrophy11,12. A closer re-examination of the ex vivo OCT images, prompted by the histological findings, confirmed an intact RPE layer in both eyes (Fig. 6C–D). This underscores the challenges of interpreting ex vivo OCT, where tissue fixation and postmortem changes may obscure critical landmarks, and reflects the clinical ambiguity surrounding whether cORA falls within the AMD spectrum, a distinction that remains debated in clinical practice. Sensitivity and specificity: ex vivo fundus imaging vs. histology Using histological assessment as the reference standard, the diagnostic performance of ex vivo imaging was calculated. Ex vivo imaging demonstrated a sensitivity of 18.8% (95% CI: 6.6–43.0%) and a specificity of 68.6% (95% CI: 52.0–81.5%). While the modality showed moderate specificity, the low sensitivity resulted in 13 False Negatives, indicating that a significant proportion of pathology was missed. Conversely, the specificity was impacted by 11 False Positive results, which included both overestimation of artefacts and misclassification of pathology. Predictive values were also calculated to assess clinical utility. The positive predictive value (PPV) was 21.4% (95% CI: 7.6–47.6%), indicating that only a small proportion of eyes flagged as abnormal on imaging were confirmed to have the specific suspected pathology on histology. The negative predictive value (NPV) was 64.9% (95% CI: 48.8–78.2%), suggesting moderate reliability in ruling out disease when the imaging appeared normal. The diagnostic odds ratio (DOR), which combines sensitivity and specificity into a single global measure, was 0.50 (95% CI: 0.12–2.12). A DOR below 1.0 reflects the impact of the high false positive rate (driven by misclassifications) and false negative rate relative to correct calls. The overall diagnostic accuracy was 52.9% (95% CI: 39.5–66.1%). Collectively, these metrics highlight the limitations of using ex vivo fundus imaging as a standalone screening tool and underscore the necessity of histological confirmation for accurate classification. Discussion This study tested a systematic clinicopathological validation protocol for donor eyes, integrating ex vivo imaging with histological assessment. The approach was designed to address limitations in current donor eye classification methods by evaluating the performance of ex vivo imaging relative to histological validation. The study highlights a fundamental challenge in translational ophthalmic research: corneal eye banks are currently the only consistent source of donor human ocular tissue. As their primary mandate is corneal transplantation, many retrieve corneas by in situ trephination rather than enucleation of the whole globe, leaving the remaining ocular tissues unavailable for research. Even in eye banks that perform enucleation, donor anonymity usually prevents linkage to clinical records or in vivo imaging. As a result, most research tissue lacks both clinical context and histological validation. Our findings emphasise the need for systematic validation frameworks to extract maximum diagnostic and scientific value from this limited resource. Importantly, the present study does not suggest that longer post‑mortem intervals are preferable for research tissue. Instead, it demonstrates that meaningful clinicopathological classification can still be achieved in donor eyes obtained through routine corneal‑donor programs, where retrieval and preservation times are determined primarily by transplant workflows. Accordingly, the workflow presented here is intended to complement rather than replace specialised research programs that prioritise rapid retrieval. Ex vivo imaging alone is insufficient to detect the full range of pathology Ex vivo imaging, though useful for identifying macroscopically conspicuous lesions, missed a substantial proportion of clinically relevant findings. Over half of the eyes were classified on imaging as having only postmortem changes. However, histology revealed underlying pathology in nearly a third of these cases, including early AMD, photoreceptor degeneration, and retinal vascular lesions. Ex vivo OCT, while superior to fundus photography for cross-sectional structural microanatomy, failed to detect basal laminar deposit or retinal-layer–specific atrophy. Two eyes initially diagnosed with geographic atrophy based on ex vivo imaging were found on histology to have complete outer retinal atrophy (cORA) with intact RPE and no BLamD—findings that exclude AMD and suggest alternative diagnoses. Although ex vivo imaging protocols have been published, most are designed to enrich donor-eye cohorts for AMD research13. This limits the availability of well-characterised, phenotypically diverse ocular tissue that could advance understanding across a broader range of ophthalmic diseases. Ex vivo OCT was performed in eight eyes and was used selectively as a problem-solving modality in equivocal or disputed cases rather than as a uniform screening test. Separate performance analysis for OCT versus non-OCT cases would therefore be underpowered and subject to substantial selection bias. Future studies should establish disease-specific interpretive criteria for ex vivo OCT and evaluate its diagnostic performance against histology in larger cohorts undergoing systematic multimodal imaging. Histological validation refines or expands imaging-based assessment Histology provided definitive diagnostic clarification in nearly half the cohort, including early AMD, hypertensive vasculopathy, and other subtle pathology. In one donor, bilateral choroidal ganglioneuroma was identified despite normal imaging findings—a rare diagnosis suggestive of neurofibromatosis type 17–9. This study extends prior ex vivo imaging protocols13 that selected for AMD eyes with very short postmortem delays suitable for resin embedding and ultrastructural analysis. While those methods preserve morphology, they preclude downstream applications such as immunohistochemistry, spatial transcriptomics, proteomics, and metabolomics. Our workflow, using routinely processed FFPE tissue, allows both robust histological validation and compatibility with next-generation molecular platforms, providing a flexible foundation for integrated morphological and multi-omic research on donor eyes. By systematically validating all available donor tissue, this protocol also addresses a key limitation of real-world practice—the inadvertent inclusion of eyes with undetected pathology in control cohorts. Such misclassification reduces study power, confounds biomarker discovery, and undermines reproducibility, particularly in imaging, transcriptomic, and machine-learning experiments. Histology remains the diagnostic reference standard and is indispensable for ensuring cohort integrity; studies relying solely on ex vivo imaging should be interpreted with these limitations in mind. A combined imaging–histology protocol enables accurate tissue classification The integrated clinicopathological workflow enabled consensus diagnosis for all 51 eyes. It proved effective not only for identifying pathology but also for distinguishing artefactual from disease-related changes—an essential step in preventing misclassification. Statistical analysis using the Stuart-Maxwell test confirmed a significant divergence between ex vivo fundus and histological classifications (χ2 = 33.9, p < 0.0001), underscoring the limitations of imaging alone and the necessity of histological validation. Because the protocol relies on standard imaging and histopathology infrastructure, it can be readily implemented within existing eye-bank workflows without disrupting corneal transplantation priorities. Standardising this process across centres could generate large, well-phenotyped tissue repositories suitable for downstream molecular, imaging, and AI-based analyses. This current study demonstrates the feasibility and performance of this approach and its potential to serve as a reference framework for multicentre ocular biobanking. Limitations and future directions Several limitations merit consideration. Interpretation of ex vivo imaging posed a learning curve for retinal specialists accustomed to in vivo conditions. The absence of vascular perfusion, fixation effects, and postmortem artefacts introduced visual ambiguities that likely contributed to under-recognition of pathology14. There are currently no established interpretive criteria for ex vivo OCT, and cross‑sectional imaging often reveals additional postmortem artefacts not readily apparent on fundus imaging. For these reasons, ex vivo OCT was used selectively in this study rather than applied to all eyes. Establishing reference standards for postmortem fundus and OCT imaging will be critical to broader adoption. The donor cohort, derived from a corneal transplant eye bank, was skewed towards older individuals with systemic disease and variable postmortem intervals. Artefacts such as retinal detachment, RPE separation, or pigment displacement further complicate interpretation. While the final validated cohort included a broad range of retinal and choroidal pathologies, the small number of cases within each diagnostic category limits disease-specific performance analysis and generalisability. The overall cohort size was modest, reflecting the exploratory nature of this protocol study and the limited availability of intact donor globes from corneal‑donor programs. As such, the diagnostic performance estimates should be interpreted cautiously and are intended primarily to illustrate the feasibility of the combined imaging–histology workflow rather than to provide definitive performance metrics. Furthermore, postmortem changes including retinal opacification, retinal detachment, RPE or retinal tears, and vitreous debris are major impediments to macroscopic assessment on ex vivo fundus photography and OCT. These artefacts, together with the limited experience of retinal specialists in interpreting ex vivo imaging, likely contributed to instances of discordance between imaging and histological diagnoses. Despite these constraints, the workflow demonstrates that meaningful clinicopathological validation can be achieved using resources already available in most eye-bank and diagnostic pathology settings. Future adaptation using AI-assisted image analysis and automated histology review could reduce observer bias and improve efficiency. Prospective donation programs that integrate in vivo imaging, consent for research, and postmortem validation would represent the next step towards sustainable, well-characterised ocular biorepositories15. The post-fundus imaging, grossing stage represents a critical decision point for tissue triage. Refining protocols at this step will enable allocation of optimal samples for histological validation while preserving adjacent material for whole-mount, organotypic, or in vitro culture and a range of molecular applications—including spatial transcriptomics, proteomics, metabolomics, lipidomic, and multiplex imaging. Our findings support a standard methodological approach for histopathological studies using eye‑bank eyes that includes systematic ex vivo imaging prior to tissue processing, predefined imaging and histological classification criteria, independent masked review with consensus adjudication, and explicit mapping between imaging and histological categories. Importantly, this workflow remains feasible and reproducible across heterogeneous donor tissue, including eyes retrieved under routine corneal‑donor conditions and regardless of post‑mortem interval barriers. Conclusion This study shows that systematic histological validation markedly improves diagnostic accuracy in donor eyes and reveals important retinal, RPE, and choroidal diseases that ex vivo imaging alone may miss. It demonstrates that real-world corneal eye-bank tissue—even with variable post-mortem delays—can be reliably classified and provide valuable research material. The workflow uses widely available, low-cost ex vivo fundus imaging and routine FFPE histology, generating tissue compatible with a broad range of downstream research platforms. By maximising the diagnostic and scientific yield of scarce donor eyes through a feasible, scalable approach, this protocol offers a practical model for integrating imaging–histology validation into eye-bank practice and establishing sustainable, high-quality ocular biorepositories. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements We gratefully acknowledge Drs Shirley and John Sarks for their foundational contributions to clinicopathological correlation in eye research, and for their participation in the clinicopathological consensus meetings held in 2017 and 2018. We also thank the eye donors and their families for consenting to donation and enabling the further use of tissue for research. Author contributions M.C.M., S.C. and V.P. contributed to the conception and design of the work, with V.P. also responsible for project coordination. S.H.V.S., W.Y., and L.K.T. performed the analysis and interpretation of data. S.H.V.S., W.Y., and S.C. drafted the original manuscript. J.W., R.M., A.H., and A.I. served as the retinal specialists for ex vivo imaging assessment. A.A., G.H., J.O., M.R., and S.C. served as the pathologists for histological assessment. M.K was involved in the microscopy work and consensus meetings of the project.H.M. and C.P. contributed to the interpretation of data and made substantial revisions to the work. S.C. supervised the project and acquired funding. All authors reviewed and approved the submitted version. Funding Shirley & John Sarks Macular Degeneration Research Foundation (SC), St Vincent’s Curran Foundation Shirley & John Sarks Research Grant (VP), UNDA Shirley & John Sarks Higher Degree by Research Scholarship (SVS), University of Sydney Save Sight Institute Shirley & John Sarks Research Grant (LKT). Data availability The datasets generated and analysed during the current study are available from the corresponding author on reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Sarrvesa Hari Vijay Singh and Waiwai Yeap contributed equally to this work. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The datasets generated and analysed during the current study are available from the corresponding author on reasonable request.