Artificial intelligence (AI) is revolutionizing many fields, and medicine is no exception. Its potential to improve screening for diabetic retinopathya serious complication of diabetes that can lead to blindnessis particularly promising. Indeed, AI offers the possibility of making eye exams faster, cheaper et more accessiblehelping to prevent vision loss in millions of people with diabetes.
AI for early detection
The early detection of diabetic retinopathy is essential to prevent vision loss. It is currently based on the analysis offundus imagesa process that can be long and expensiverequiring the intervention of qualified healthcare professionals. This is where AI comes in. AI algorithms can be trained to analyze these images and detect signs of illness with remarkable precision.
Imagine a system capable of identifying suspicious anomalies on fundus images and report them to ophthalmologists for further examination. This type of system, already developed by companies like Right markersallowsoptimize specialists’ time and of reduce costs related to screening.
Obstacles to overcome
Despite the undeniable advantages of AI in diabetes screening, several challenges remain. One of the main obstacles is the need to havehigh quality images for a reliable diagnosis. Varying lighting conditions, dirty lenses, or lack of operator training can affect image quality and result in false positivesthus generating anxiety and unnecessary expenses.
The experience of Google Health in Thailand illustrates this problem well. The algorithm developed by the company, which performed very well in the laboratory, proved less effective in the field due to the variability of shooting conditions. This experience highlights the importance of working with diversified data and consult a wide range of professionals to develop AI systems robust and reliable.
A hybrid model for more efficiency
Faced with these challenges, a hybrid model combining AI and human expertise seems to be the most promising solution. In this model, the AI performs an initial screening of images, identifying those that show potential signs of diabetic retinopathy. These images are then reviewed by medical professionals for a final diagnosis.
A study carried out at Singapore demonstrated the effectiveness of this hybrid model, which proved not only more precise than thefull automationbut also more economical than thehuman assessment alone. This model will also be integrated into the national IT platform of the Singapore health service in 2025.
Accessibility for all: a major issue
While AI offers considerable potential to improve diabetes screening, it is crucial to ensure that this technology is accessible to alland not just to rich countries. As pointed out Bilal Mateenhead of AI at the NGO PATH, it is essential to ask whether AI solutions are designed for everyone or only for a privileged minority. L’health equity must be at the heart of the development and deployment of these technologies.
AI has the power to transform how we screen for and treat diabetic retinopathy. By making tests faster, cheaper and more accessible, it can help prevent blindness in millions of people around the world. Now is the time to address the remaining challenges and ensure that this technology benefits everyone, without distinction.
