Revolutionising Diabetic Eye Care
In a landmark development for rural healthcare, Universiti Malaysia Sarawak (UNIMAS) has created a Smartphone-Based Diabetic Retinopathy (DR) Screening System Using Deep Learning CNN Classification, a groundbreaking tool that promises to revolutionise eye care accessibility. This innovative project, which recently won the prestigious Research & Development Award under the Sarawak Digital Economy Awards (DEA) 2024, was developed from the collaborative efforts of the Faculty of Engineering, Faculty Medicine and Health Sciences and Faculty of Computer Science and Information Technology, aiming to address the critical shortage of eye care resources in Malaysia’s rural regions.
Diabetic Retinopathy: The Silent Threat
Diabetic Retinopathy (DR) is one of the leading causes of blindness worldwide. In Malaysia, access to DR screening has been limited, especially in rural areas where advanced ophthalmic diagnostic tools are scarce and costly. According to the Health Assessment Unit of the Ministry of Health, comprehensive DR screening involves vision assessment and retinal examination, typically using expensive equipment. For instance, a conventional fundus camera costs around RM100,000 per unit and a digital camera that captures a clear view of the eye could cost upwards of RM120,000.
Moreover, these diagnostic machines are often large, immobile, and requires frequent calibration to produce the results necessary for a proper diagnosis. Within Sarawak, such equipments are accessible only at select hospitals or eye clinics. Furthermore, it is impractical to bring such equipment to rural and remote areas whose topography (e.g., lack of road access, across a river) make the transportation difficult, if not outright impossible. This limited access makes it so that rural DR patients could not easily get the treatment they require before their condition deteriorates into blindness.
EYE-DR: Leveraging Deep Learning to Democratise Eye Care
The heart of this breakthrough system lies in its portable smartphone-based application system, EYE-DR, that integrates Deep Learning (DL) Convolutional Neural Network (CNN) classification model with a portable ophthalmic examination setup. Here is how it works in practice:
- Portable Hardware for Remote Eye Screening: The project’s hardware component includes a mobile phone (currently a Google Pixel and iPhone models are used as the application is compatible with both Android and iOS devices), a 3D-printed adapter, and a portable ophthalmoscope. The adapter, made with biodegradable materials, securely aligns the smartphone’s camera with the ophthalmoscope to capture high-resolution images of the retina, essential for detecting early signs of DR.
- Image Capture and Processing: Medical staff use the app “EYE-DR” to capture detailed retinal images via the portable ophthalmoscope and the smartphone’s camera. For best results, images are taken in well-lit conditions to ensure clear visibility of the retina.
- Deep Learning-Powered Screening: Once the image is captured, it’s immediately processed by the “EYE-DR” app, which incorporates a DL CNN model to classify the retina’s condition. The CNN is trained on a large dataset of retinal images, with varying grades of DR (from normal at grade 0 to severe at grade 4), to detect abnormalities accurately. At the time of writing, the system has been fine-tuned for high sensitivity and specificity, allowing it to distinguish between normal (grade 0) and extreme (grade 4) conditions with considerable accuracy.
- Immediate Screening Results and Cloud Storage: After processing, the app provides screening results in seconds. If the system detects signs of diabetic retinopathy, it notifies the user, who can then recommend further follow-up procedure with the patient for treatment. For mild case of diabetic retinopathy, advice will be given to observe and follow up the eye condition whereas patients with severe condition will be referred to eye specialist for further management in a timely manner.
As the app stores patient data and retinal images on a cloud database, it enables medical staff to easily share the images with an eye specialist if needed. This feature is crucial in remote areas where access to specialists is limited; the images and diagnosis can be quickly transferred for further examination and prompt intervention. The stored data also helps health providers track patients’ screening history, enhancing long-term care management for diabetes-related eye complications.
In contrast to standard DR diagnostic machines, this new project presents a low-cost, portable solution for DR’s early preventive measures. A single unit of this innovative system—which includes a smartphone (access to the EYE-DR application included), an adapter, and a portable ophthalmoscope—costs about RM10,000; a fraction of the traditional costs.
By bringing DR screening directly to the patient in rural areas instead of necessitating they make long journeys to selected eye specialist centres in the urban areas, this project enables greater access to routine checkups for eye care, making a substantial impact on both early intervention and ongoing diabetic care in underserved communities.
Overcoming Challenges and Building Partnerships
Being revolutionary in its concept and design, the journey to create a precise DR screening system was not without its challenges. Training the algorithm to accurately distinguish between normal and pathological fundus images requires extensive collaboration with hospitals and eye clinics. Acquiring enough data, particularly images of DR at varying stages, remains a challenge, with more data needed to refine the app’s accuracy in intermediate cases.
The team also faced issues with data security and patient consent, particularly when working with sensitive medical data. Despite these challenges, Associate Professor Dr Kuryati remains optimistic, highlighting ongoing collaborations to further enhance the app’s functionality.
The potential of this smartphone-based system to transform healthcare in Sarawak is enormous. Once fully developed, the project could drastically reduce the need for patients in rural areas to travel to urban centres for DR screening. As the system provides immediate results, it allows patients to seek prompt intervention if needed, which can significantly prevent disease progression before their vision deteriorates to blindness.
The project is currently at technology readiness level 5 as of May 2024; the team is working to increase the app’s accuracy by adding a broader range of DR images within the grading system and enhancing image-processing capabilities.
“Of course, the EYE-DR is a working prototype currently. But we intend to refine the AI’s ability of image processing and configuration even further. We are already in contact with local eye specialist centres and hospitals like KPJ Kuching Specialist Hospital and the Tun Hussein Onn National Eye Hospital for collaborations apart from our existing collaborator Eye Specialist Centre (ESC), Kuching.”
Looking ahead, the project aims to reach technological readiness level 7 and secure patents within the next three years, while plans for commercialisation and expansion into a national diabetic eye-screening network, incorporating telemedicine support are set for the next five to ten years. In addition to these goals, Associate Professor Dr Kuryati and the team also ambitions the possibility of creating their own portable ophthalmoscope device akin to a VR headset for greater portability and usability in self-retinal imaging.
“Perhaps one day, it could be expanded to screen other eye diseases beyond diabetic retinopathy. In any case, we are keen on developing this project to its full potential.”
Advice for Future Researchers
Winning the Research & Development Award affirms the team’s commitment to improving rural healthcare through technological innovation. The team sees this recognition as a testament to the project’s potential and encourages young researchers to enter the field of digital health.
For Associate Professor Dr Kuryati, digital healthcare is at the frontier of medical advancement, offering significant room for innovation across multiple specialties. She stresses that young researchers in this field have opportunities to create far-reaching change but that success relies on creative, lateral thinking and collaborative, cross-disciplinary work. In her view, the most impactful digital health solutions are crafted when medical specialists, engineers, and IT experts come together to tackle challenges with fresh perspectives and a common purpose.
One of her primary pieces of advice to new innovators in digital healthcare is to have a thorough understanding of ethical standards, as these are foundational to responsible and effective medical research.
“Informed consent and patient privacy is not to be taken likely. So be sure to understand the ethical pillars behind data security, and responsible practices when developing medical solutions, especially in the field of AI and digital health. This will help you navigate challenges and strengthen the integrity and value of your contributions,” she explains.
In addition to ethics, Associate Professor Dr Kuryati emphasises the importance of creating technology that is both user-friendly and sustainable. For devices intended for direct patient interaction, she highlights that accessibility and ease of use can greatly increase the impact of a project. For example, digital solutions for diagnostics should be crafted with practicality and ease of access in mind to benefit patients and healthcare providers alike.
She also warns innovators about the risks of over-claiming and encourages careful use of terminology.
“It’s essential to use the right words,” she says, “so the function and value of the tool are clear and well-understood. For instance, we chose the term ‘screening’ rather than ‘diagnosis’ to describe our project, as only eye specialists using advanced equipment can provide a formal diagnosis.”
Finally, she encourages new researchers to search for their personal motivation, or their “why,” behind their work.
“Finding your purpose isn’t just motivating—it shapes your journey, helping you persevere in making meaningful contributions to healthcare.”
With the ultimate vision of creating a “Community-Driven University for a Sustainable World,” the team’s smartphone-based DR screening system exemplifies a sustainable healthcare solution. By addressing a critical healthcare need through an accessible, AI-driven approach, this project represents a significant step toward equitable healthcare in Sarawak and beyond.
