Background/Aims To objectively classify eyes as either healthy or glaucoma based exclusively on data provided by peripapillary retinal nerve fiber layer (pRNFL) and ganglion cell-inner plexiform (GCIPL) measurements derived from spectral-domain optical coherence tomography (SD-OCT) using machine learning algorithms. Methods Three clustering methods (k-means, hierarchical cluster analysis -HCA- and model-based clustering-MBC-) were used separately to classify a training sample of 109 eyes as either healthy or glaucomatous using solely 13 SD-OCT parameters: pRNFL average and sector thicknesses and GCIPL average and minimum values together with the six macular wedge-shaped regions. Then, the best-performing algorithm was applied to an independent test sample of 102 eyes to derive close estimates of its actual performance (external validation). Results In the training sample, accuracy was 91.7% for MBC, 81.7% for k-means and 78.9% for HCA (p value = 0.02). The best MBC model was that in which subgroups were allowed to have variable volume and shape and equal orientation. The MBC algorithm in the independent test sample correctly classified 98 out of 102 cases for an overall accuracy of 96.1% (95% CI, 92.3–99.8%), with a sensitivity of 94.3 and 100% specificity. The accuracy for pRNFL was 92.2% (95% CI, 86.9–97.4%) and for GCIPL 98.0% (95% CI, 95.3–100%). Conclusions Clustering algorithms in general (and MBC in particular) seem promising methods to help discriminate between healthy and glaucomatous eyes using exclusively SD-OCT-derived parameters. Understanding the relative merits of one method over others may also provide insights into the nature of the disease.
16 January 2026 in British Journal of Ophthalmology
Aim: To explore associations between artificial intelligence (AI)-based baseline optical coherence tomography (OCT) fluid compartment quantifications and 12-month visual outcomes in diabetic macular oedema (DME) eyes treated with the intravitreal dexamethasone implant. Methods: This was a multicentre, real-world, national DME database and associated OCT dataset study. Demographics, visual acuity (VA), treatments and visit data were collected using a validated web-based tool (Fight Retinal Blindness). Fluid compartment quantifications, including intraretinal fluid (IRF) and subretinal fluid (SRF), were measured in nanolitres (nL) using a validated AI tool (Discovery). Univariate and multivariate regression mixed models evaluated associations between anatomical variables and VA outcomes. Results: A total of 101 treatment-naïve DME eyes were grouped into quartiles according to their fluid volume for each fluid compartment (Q1: lowest volume, Q4: highest volume). Baseline IRF was associated with greater VA gains at month 12 (+6.34 letters, p=0.07) but poorer final VA (−8.95, p=0.07), while SRF was associated with worse final VA at 12 months (−12.5, p=0.01). At month 3, IRF was associated with a VA decrease at 12 months (−13.7, p=0.02) and lower final VA (−29.8, p<0.001). At month 12, IRF was associated with lower final VA (−11.6, p=0.03). Quantitatively, a reduction of 100 nL of IRF at 3 months was associated with a +1.54 letters gain (p=0.03) in the multivariate analysis. Conclusion: This real-world, multicentre study describes objective baseline fluid volumes that predict visual outcomes at 12 months in routine clinical care. Accurate quantification of baseline fluid volumes may play a predictive role for final visual outcomes.
21 November 2025 in Japanese Journal of Ophthalmology