This project addresses a growing public health challenge in Malaysia: the early detection of diabetic retinopathy (DR), a leading cause of preventable vision loss. Traditional DR diagnosis depends on manual ophthalmological examination, which is increasingly unsustainable due to rising patient numbers and limited specialist capacity, often resulting in delayed screening and referral.

To respond to this need, the project has developed a cost-effective, locally driven AI-based screening platform to support rapid DR diagnosis using retinal fundus images. The system is trained using a combination of public datasets and locally relevant data, ensuring both robustness and applicability to Malaysian healthcare settings. A key innovation lies in its clustered AI architecture, where multiple models are customised to different fundus cameras used across hospitals and clinics, rather than relying on a one-size-fits-all solution.

At its current stage, the platform adopts a clinician in the loop approach. Clinicians perform an initial visual check of image quality and guide the selection of the appropriate AI model, ensuring safe and reliable deployment while maintaining clinical oversight. As training improves, the system is expected to progressively handle images from lower-quality cameras, including those commonly used in community health clinics. Ultimately, the platform aims to enable machine-initiated preliminary screening, reducing waiting times at the initial diagnostic stage.

The project is designed for integration into routine clinical workflows and for scalable adoption through the Ministry of Health (KKM). It also places strong emphasis on community-level screening, supporting earlier detection closer to patients. In parallel, the initiative contributes to capacity building, including the training of a biomedical engineering student embedded in a clinical setting, and is supported by Ogawa Malaysia, reinforcing its translational and industry relevance.