A new technique developed by Indian and US researchers utilizes artificial intelligence to detect diabetes without the need for traditional blood tests. By capturing a high-resolution photo of the retina, the AI system can identify individuals with high blood sugar levels. This method relies on identifying subtle signs in the blood vessels of the eye that are not visible to the naked eye, distinguishing between those with and without diabetes without the use of a finger-prick blood test.
Published in the Diabetes Technology and Therapeutics journal, the study demonstrated the AI’s ability to detect diabetes through retinal images, offering a potential breakthrough in early diagnosis. Dr. V. Mohan, a renowned diabetologist based in Chennai and a Padma Shri awardee, highlighted the significance of this innovation in a country with over 100 million diabetes cases, many of which go undiagnosed. The use of AI tools with retinal photos could revolutionize real-time diabetes screening in the future.
Researchers, including Dr. Sudeshna Sil Kar from Emory University in the US, trained the AI to analyze specific shapes and patterns in retinal images to differentiate individuals with diabetes from those without. By examining 273 retinal images from 139 participants, the team extracted quantitative vessel tortuosity features for arteries and veins using machine vision-based techniques. The AI system exhibited 95% sensitivity in correctly identifying diabetes in the test group and could even detect ‘prediabetes,’ a stage where lifestyle modifications can prevent the onset of diabetes.
This non-invasive AI technique offers a cost-effective and efficient means of early diabetes detection, eliminating the need for complex laboratory equipment, fasting, or blood tests. The simplicity of capturing a quick photo of the eye’s retina makes it a convenient screening tool for patients. While the results are promising, further validation in a larger population is essential to establish the technique’s reliability.
