Statistical comparison of MLE for Class P images (n = 49).
| Comparison | Test | Statistic | RawP-value | BonferroniP-value | Mean difference (A–B) |
|---|---|---|---|---|---|
| Friedman | χ²(2) | 71.306 | 3 × 10–16 | – | – |
| RetinaRegNet vs EyeLiner | Wilcoxon V | 238 | 2 × 10–4 | 6 × 10–4 *** | −1.513 |
| RetinaRegNet vs GeoFormer | Wilcoxon V | 1 | 1 × 10–9 | 4 × 10–9 *** | −6.957 |
| EyeLiner vs GeoFormer | Wilcoxon V | 41 | 1 × 10–8 | 4 × 10–8 *** | −5.444 |
Asterisks denote Bonferroni-adjusted significance: * P < 0.05, ** P < 0.01, *** P < 0.001.
The supplementary materials for this article are available at: https://www.explorationpub.com/uploads/Article/file/101194_sup_1.pdf.
TD: Conceptualization, Methodology, Data curation, Formal analysis, Project administration, Writing—original draft, Writing—review & editing, Visualization, Writing—review & editing, Funding acquisition. NS: Conceptualization, Investigation, Methodology, Formal analysis, Data curation, Visualization, Writing—original draft, Writing—review & editing. YWC: Supervision, Conceptualization, Methodology, Validation, Writing—review & editing, Funding acquisition. SA: Supervision, Writing—review & editing, Writing—original draft. KD: Investigation, Writing—original draft, Writing—review & editing. PG: Writing—review & editing. YTC: Writing–review & editing. AJ: Validation. NP: Supervision, Project administration, Validation, Writing—review & editing, Funding acquisition. All authors read and approved the submitted version.
NP is a patent holder of PCT/EP2023/076614 filed by UCL Business. The authors declare that they have no other relevant conflicts of interest.
This study utilised the FIRE dataset, which is publicly available, so it involves no new data collection or direct interaction with human participants. All data are fully anonymised. Therefore, ethical approval was not required.
This study involved secondary analysis of publicly available, anonymised retinal image data and did not involve direct participation of human subjects. Accordingly, consent to participate was not required.
No identifiable individual data are presented in this study. Therefore, consent to publication was not required.
The FIRE dataset analysed during this study is publicly available at: https://projects.ics.forth.gr/cvrl/fire/ and was used in accordance with the terms of use. No additional datasets were generated or analysed. Code used for analysis is available from the corresponding author upon reasonable request.
The research was supported by a grant from the National Institute for Health Research (NIHR) Biomedical Research Centre (BRC) at MEH NHS Foundation Trust and UCL Institute of Ophthalmology (grant no. NIHR203322). N.P. is funded by an Artificial Intelligence in Health and Care Award (NIHR AI Award grant no. AI_AWARD02488). The Artificial Intelligence in Health and Care Award is part of the NHS AI Laboratory, which has made funding available to accelerate the testing and evaluation of artificial intelligence technologies that meet the aims set out in the NHS Long Term Plan. The NHS AI Laboratory is a joint unit of teams from the Department of Health and Social Care and NHS England, driving forward the digital transformation of health and social care (https://transform.england.nhs.uk/ai-lab/). N.P. is also funded by Sight Research UK (grant no. TRN004). N.P. and Y.W.C are also funded by Medical Research Foundation and Moorfields Eye Charity (grant no. MRF-JF-EH-23-122) and Fight for Sight (grant no. RESSGA250). T.D. and N.P. are also funded by NIHR i4i THRIVE (grant no. NIHR505133). N.P. was also previously funded by Retina UK as part of the UK IRD Consortium, Moorfields Eye Charity Career Development Award (grant no. R190031A), HDRUK (grant no. MC_PC_18036) and by a Translational Innovation grant awarded by the UCL Translational Research Office, which has seed funded this work. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
© The Author(s) 2026.
Open Exploration maintains a neutral stance on jurisdictional claims in published institutional affiliations and maps. All opinions expressed in this article are the personal views of the author(s) and do not represent the stance of the editorial team or the publisher.