In areas where conventional fundus cameras are logistically unaffordable, the smartphone-based approach is considered as a promising future of glaucoma screening due to their affordability and ease-of-use. Optic disc localization is a critical stage during glaucoma screenings. For fundus images captured with standard fundus cameras, the majority of the models available out there can locate the optic disc with satisfactory performance. However, for images that are captured with smartphone-based fundus cameras, the inherent noise and lower quality of the fundus image makes it difficult for models to detect the optic disc region. In this study, we have proposed to utilize YOLOv8n for optic disc localization due to the model’s cutting-edge performance in diverse tasks and its lightweight nature. We have used a public dataset which has 2000 low quality fundus images that are captured using a smartphone-based fundoscopy device. From these, 60% of the data was used for fine-tuning the model, and 25% for testing. By using a confidence of 50% set as threshold, the model was able to detect the optic disc successfully on over 97% of test images with intersection over union of above 0.85. These results highlight the potential of the lightweight YOLOv8n model for deployment on resource-constrained environments, offering a promising performance on accurately localizing the OD and enhancing the feasibility of affordable glaucoma screening on smartphones.
Fundus imaging is widely used for diagnosing and monitoring ocular diseases such as diabetic retinopathy, glaucoma, and age-related macular degeneration. However, the high cost and limited availability of conventional fundus cameras pose significant barriers, particularly in resource-constrained settings. This study introduces the Glaucoma Screening on Phone (GSoP), an affordable and portable smartphone-based fundus imaging adapter designed to address these challenges. The adapter was developed using accessible and cost-effective components. We recorded retinal videos from dilated pupil with a focus on the optic disc region, which provides critical information on the degeneration of the optic nerve. While clinical trials revealed artifacts such as glare and noise that reduced overall image quality, the GSoP demonstrated its ability to capture diagnostically relevant images of the optic disc region. A subjective qualitative comparison with the commercially available ophthalmoscope called oDocs-nun showed that although the GSoP’s field of view is smaller, it effectively highlights the optic nerve head, a critical area for glaucoma screening. Our approach is well-suited for mydriatic video-based screening due to its limited field of view. With a production cost of under C10, the GSoP offers a practical and accessible solution for primary healthcare and educational purposes. Future improvements, including glare reduction mechanisms, AI-driven automation, and modular design options, have the potential to enhance its diagnostic capabilities and broaden its impact.
Glaucoma is one of the leading causes of irreversible blindness worldwide, affecting more than 76 million people, with prevalence expected to increase subs
This study investigated the feasibility of deep learning–based glaucoma screening using fundus videos acquired with an ultra-low-cost, smartphone-based imaging system designed for resource-limited settings. Fundus videos were collected during routine clinical examinations using the glaucoma screening-on-phone (GSoP) adapter. A red-channel intensity-based strategy was employed to select diagnostically relevant frames from videos. Optic disc (OD) localizationwas performed using a fine-tunedYOLOv8n object detector, followed by OD-centered cropping and preprocessing. Glaucoma classification was conducted using MobileNetV3-Small model, fine-tuned via five-fold stratified group cross-validation. Performancewas evaluated using accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC), with additional testing on independent dataset. Comparative and ablation analyses assessed architectural choices and OD localization impact. Despite GSoP’s limited optical quality, the model achieved strong performance. On the independent test set, the single-frame MobileNetV3-Small reached an AUC of 0.91, with 93% sensitivity and 88% specificity. This semi-automated approach outperformed a fully automated multi-frame strategy (0.90 vs. 0.81 accuracy). Ablation analysis demonstrated that OD localization improved sensitivity and specificity by 13.3% and 9.5%, respectively, on MobileNetV3-Small model. The lightweight model performed comparably to deeper networks with significantly lower computational complexity. These findings indicate that clinically meaningful glaucoma-related features can be extracted from low-quality fundus images acquired with ultra-low-cost smartphone hardware, with semi-manual single-frame selection currently proving more effective than fully automated multi-frame-based prediction. This work serves as an encouraging proof-of-concept for combining affordable imaging systems with efficient deep learning pipelines to achieve scalable glaucoma screening in underserved settings.