Primary diabetes care and diabetic retinopathy (DR) screening persist as major public health challenges due to a shortage of trained primary care physicians (PCPs), particularly in low-resource settings. Here, to bridge the gaps, we developed an integrated image–language system (DeepDR-LLM), combining a large language model (LLM module) and image-based deep learning (DeepDR-Transformer), to provide individualized diabetes management recommendations to PCPs. In a retrospective evaluation, the LLM module demonstrated comparable performance to PCPs and endocrinology residents when tested in English and outperformed PCPs and had comparable performance to endocrinology residents in Chinese. For identifying referable DR, the average PCP’s accuracy was 81.0% unassisted and 92.3% assisted by DeepDR-Transformer. Furthermore, we performed a single-center real-world prospective study, deploying DeepDR-LLM. We compared diabetes management adherence of patients under the unassisted PCP arm (n = 397) with those under the PCP+DeepDR-LLM arm (n = 372). Patients with newly diagnosed diabetes in the PCP+DeepDR-LLM arm showed better self-management behaviors throughout follow-up (P < 0.05). For patients with referral DR, those in the PCP+DeepDR-LLM arm were more likely to adhere to DR referrals (P < 0.01). Additionally, DeepDR-LLM deployment improved the quality and empathy level of management recommendations. Given its multifaceted performance, DeepDR-LLM holds promise as a digital solution for enhancing primary diabetes care and DR screening.

Aims To compare and rank the myopia control effects of different light wavelengths in children using a systematic review and Bayesian network meta-analysis (Bayesian NMA). Methods The review protocol was registered with PROSPERO. We searched PubMed, EMBASE and MEDLINE for relevant clinical and animal studies published as of 2 February 2023. We included studies comparing red, violet or full-spectrum light with controls. Data extracted included descriptive statistics and study outcomes (axial length (AL) elongation and progression of spherical equivalent (SE) refraction). After quality assessment, estimates of treatment effect outcomes (mean differences (MDs) and 95% CIs) were first pooled for the animal and clinical studies in a traditional meta-analysis. To compare and rank the different light wavelengths, the Bayesian NMA was then conducted for all the included clinical studies (12 studies) and separately for only randomised controlled trials (8 studies). MDs, 95% credible intervals (CrIs) and ranks of the various light wavelengths were estimated in the Bayesian NMA. Results When all clinical studies were included in the Bayesian NMA (12 studies), only red-light significantly slowed AL elongation, MD (95% CrI), −0.38 mm (−0.59 mm to −0.16 mm)/year and SE refraction progression, 0.72D (0.35D to 1.10D)/year compared with controls. It remained the only significant intervention when effect sizes from only RCTs (eight studies) were separately combined, (−0.28 mm (−0.40 mm to −0.15 mm)/year and 0.57D (0.22D to 0.92D)/year, for AL and SE refraction, respectively). Conclusion Myopia control efficacy varied among different wavelengths of light, with red light ranked as the most effective. PROSPERO registration number Clinical studies: CRD42022368998; animal studies: CRD42022368671.

Aims To develop and externally test deep learning (DL) models for assessing the image quality of three-dimensional (3D) macular scans from Cirrus and Spectralis optical coherence tomography devices. Methods We retrospectively collected two data sets including 2277 Cirrus 3D scans and 1557 Spectralis 3D scans, respectively, for training (70%), fine-tuning (10%) and internal validation (20%) from electronic medical and research records at The Chinese University of Hong Kong Eye Centre and the Hong Kong Eye Hospital. Scans with various eye diseases (eg, diabetic macular oedema, age-related macular degeneration, polypoidal choroidal vasculopathy and pathological myopia), and scans of normal eyes from adults and children were included. Two graders labelled each 3D scan as gradable or ungradable, according to standardised criteria. We used a 3D version of the residual network (ResNet)-18 for Cirrus 3D scans and a multiple-instance learning pipline with ResNet-18 for Spectralis 3D scans. Two deep learning (DL) models were further tested via three unseen Cirrus data sets from Singapore and five unseen Spectralis data sets from India, Australia and Hong Kong, respectively. Results In the internal validation, the models achieved the area under curves (AUCs) of 0.930 (0.885–0.976) and 0.906 (0.863–0.948) for assessing the Cirrus 3D scans and Spectralis 3D scans, respectively. In the external testing, the models showed robust performance with AUCs ranging from 0.832 (0.730–0.934) to 0.930 (0.906–0.953) and 0.891 (0.836–0.945) to 0.962 (0.918–1.000), respectively. Conclusions Our models could be used for filtering out ungradable 3D scans and further incorporated with a disease-detection DL model, allowing a fully automated eye disease detection workflow.

A specific form of drusen, known as pachydrusen, has been demonstrated to be associated with pachychoroid eye diseases, such as central serous chorioretinopathy (CSC) and polypoidal choroidal vasculopathy (PCV). These pachydrusen have been found in up to 50% of eyes with CSC and PCV and may affect the disease progression and treatment response. This study aims to investigate the association between pachydrusen and changes in fundus autofluorescence (FAF) in eyes with CSC and PCV. A total of 65 CSC patients and 32 PCV patients were evaluated. Pachydrusen were detected using both color fundus photography and spectral-domain optical coherence tomography. The relationships between pachydrusen and FAF changes were then investigated. The prevalence of pachydrusen in CSC and PCV eyes was 16.7% and 61.8%, respectively. The mean age of patients with pachydrusen was significantly older than those without pachydrusen (CSC: 56.3 vs. 45.0 years, p < 0.001; PCV: 68.8 vs. 59.5 years, p < 0.001). No significant difference was found in the mean subfoveal choroidal thickness between eyes with or without pachydrusen. Eyes with pachydrusen were significantly associated with more extensive FAF changes in both CSC and PCV (p < 0.001 and p = 0.037, respectively). The study demonstrated that pachydrusen are more prevalent in PCV than CSC. Increasing age and more extensive abnormalities in FAF are associated with the presence of pachydrusen, suggesting that dysfunction of retinal pigment epithelial cells is associated with pachydrusen.