Retinoblastoma is one of the most common childhood cancers worldwide, with an incidence rate of 1 in 16,000-28,000 births. In developed countries, early detection and proper treatment have resulted in therapy success rates of up to 99%. However, in Indonesia, delayed diagnosis remains a serious issue, leading to 40-70% of patients who have vision loss, permanent disability, or even death. This situation underscores the urgent need for a rapid, accurate, and cost-effective screening system to enhance recovery rates and reduce childhood blindness.
Responding to this challenge, a team of students from Universitas Gadjah Mada (UGM), through the Student Creativity Program for Intellectual Creation (PKM-KC), has developed an innovative solution called Rb-AID (Retinoblastoma Artificial Intelligence Detection).
This innovation utilizes a two-step convolutional neural network (TSCNN) system integrated into a mobile application to perform real-time eye fundus examinations in toddlers.
The team is led by Jonathan Setiawan (Nuclear Engineering, 2023) with members Ammar Ali Yasir (Information Technology, 2023), Ammar (Mechanical Engineering, 2023), Muhammad Hafidz Al Farisi (Information Technology, 2023), and Emeliana Putri Ayu Ningsih (Medicine 2024), under the supervision of Professor Herianto.
Jonathan Setiawan explained that Rb-AID consists of two main components: the Rb-AID tool and the Rb-AID mobile decision system.
The first component maintains the stability and quality of fundus image acquisition, while the second employs a two-stage TSCNN system to analyze examination results quickly and objectively.
“The screening decision does not depend on the operator’s subjectivity and can be applied in primary care services,” Setiawan told reporters on Monday, Oct. 13, 2025.
Technically, the Rb-AID device was designed using Autodesk Inventor, printed with eco-friendly PLA materials, and equipped with a 20D lens to ensure sharp visual results.
“The tool’s mechanism allows a smartphone to be mounted securely on a special holder at an ideal working distance of about 50 mm to achieve optimal lighting and focus on the retina,” he explained.
On the software side, the AI system was trained using thousands of fundus images comprising both retinoblastoma and normal cases for comparison.
Through the TSCNN architecture, the application can identify clinical signs such as white retinal masses, abnormal blood vessel patterns, and uneven light reflections caused by lesions.
The application also features two operating modes: offline, which allows manual referrals by health workers, and online, where it automatically recommends the nearest ophthalmology facilities and sends examination results to doctors for verification.
In terms of novelty, Rb-AID is the first system to combine device-based image stabilization with two-stage deep learning analysis.
“This approach addresses the limitations of previous methods that relied on non-fundus photos or handheld tools prone to bias,” Setiawan added.
This innovation is expected to accelerate retinoblastoma detection in regions with limited access to conventional fundus equipment and medical specialists.
“Through Rb-AID, we aim to provide faster, more consistent, and affordable early screening so that every child in Indonesia has the same opportunity to see the world,” he said.
In the long term, the team aims to develop a fully functional prototype that integrates seamlessly with the mobile application, accompanied by scientific reports and educational publications.
This initiative aligns with the 2025 PKM theme in the field of Health and Community Nutrition, and supports the achievement of Sustainable Development Goal (SDG) 3: Good Health and Well-being.
Author: Kezia Dwina Nathania
Editor: Gusti Grehenson
Post-editor: Lintang Andwyna
Photographs: PKM Team on Rb-AID and Freepik