Search

BackgroundDeep learning (DL) is promising to detect glaucoma. However, patients’ privacy and data security are major concerns when pooling all data for model development. We developed a privacy-preserving DL model using the federated learning (FL) paradigm to detect glaucoma from optical coherence tomography (OCT) images.MethodsThis is a multicentre study. The FL paradigm consisted of a ‘central server’ and seven eye centres in Hong Kong, the USA and Singapore. Each centre first trained a model locally with its own OCT optic disc volumetric dataset and then uploaded its model parameters to the central server. The central server used FedProx algorithm to aggregate all centres’ model parameters. Subsequently, the aggregated parameters are redistributed to each centre for its local model optimisation. We experimented with three three-dimensional (3D) networks to evaluate the stabilities of the FL paradigm. Lastly, we tested the FL model on two prospectively collected unseen datasets.ResultsWe used 9326 volumetric OCT scans from 2785 subjects. The FL model performed consistently well with different networks in 7 centres (accuracies 78.3%–98.5%, 75.9%–97.0%, and 78.3%–97.5%, respectively) and stably in the 2 unseen datasets (accuracies 84.8%-87.7%, 81.3%-84.8%, and 86.0%–87.8%, respectively). The FL model achieved non-inferior performance in classifying glaucoma compared with the traditional model and significantly outperformed the individual models.ConclusionThe 3D FL model could leverage all the datasets and achieve generalisable performance, without data exchange across centres. This study demonstrated an OCT-based FL paradigm for glaucoma identification with ensured patient privacy and data security, charting another course toward the real-world transition of artificial intelligence in ophthalmology.

BackgroundOral corticosteroid remains the first-line treatment of IgG4-related ophthalmic disease, but steroid-dependence is common and serious. Factors associated with steroid dependence and relapse have to be further explored.Study populationA city-wide, biopsy-proven, Chinese cohort.MethodsRetrospective, masked review of medical records, orbital images and histopathology reports.ResultsThere were 101 patients with at least 24-month follow-up. Up to 82% (82/101) received oral corticosteroid as first-line treatments, and 7 of them received also concomitant steroid-sparing agents (SSA)/biological agents as primary treatment. There was 61% (50/82) of patients required long-term corticosteroid (alone=23, with SSA=27) after 1.9±0.7 (range 1–5) relapses. When compared with the 21% (17/82) of patients who tapered corticosteroid successfully for 24 months, steroid dependence was associated with elevated baseline serum IgG4 level (94% vs 65%, p<0.01) and Mikulicz syndrome (46% vs 18%, p<0.05). Up to 13% (11/82) of patients tolerated residual disease after tapering off corticosteroid. There was 17% (17/101) of patients did not require any medications after biopsies. They were more likely to have debulking surgeries (71% vs 40%, p<0.05), discrete orbital lesions (65% vs 26%, p<0.05), normal baseline serum IgG4 level (24% vs 6%, p<0.05) and no Mikulicz syndrome (94% vs 61%, p<0.05).ConclusionIn this cohort, 60% of patients required long-term maintenance oral corticosteroid. Elevated pretreatment serum IgG4 level and Mikulicz syndrome were associated with steroid dependence. Debulking surgery is an alternative for a subgroup of patients with discrete orbital lesions, normal baseline IgG4 level and no Mikulicz syndrome.