AI Beats Doctors in Speed, Identifies Retinal Diseases with High Accuracy

New Delhi: Artificial Intelligence (AI) is making remarkable strides in the healthcare sector, and a new study from AIIMS Raebareli has shown that AI can sometimes detect eye diseases even before doctors. While ophthalmologists remain the most reliable diagnosticians, AI tools such as ChatGPT and Gemini are rapidly proving their capabilities. Early detection of retinal diseases is critical, as delayed diagnosis can lead to permanent vision loss.

The study, conducted by the Department of Ophthalmology at AIIMS Raebareli, analyzed 30 patients. Each patient’s fundus photograph (image of the retina) along with relevant medical information was provided for assessment.

Doctors vs AI: Accuracy and Speed

The cases were evaluated in three ways—by an experienced ophthalmologist, by ChatGPT, and by Gemini—and the results were compared with the confirmed diagnoses. The findings revealed:

• Doctors: Correctly diagnosed 29 out of 30 cases, achieving 97% accuracy.
• ChatGPT: Correctly diagnosed 90% of cases.
• Gemini: Achieved 87% accuracy.

However, when it came to speed, AI tools significantly outperformed doctors. ChatGPT took an average of 22 seconds per case, while Gemini took about 26 seconds. In contrast, doctors required approximately 2.5 minutes (150 seconds) per patient to reach a final diagnosis. The study also found that doctors excelled in explaining the disease to patients, providing detailed insights beyond the diagnosis.

Diseases Detected with Precision

The research team, led by Assistant Professor Dr. Absil Shrikhal at AIIMS Raebareli, reported that both ChatGPT and Gemini were able to identify diabetic retinopathy with 100% accuracy. Other conditions accurately detected included optic atrophy, retinal detachment, and myopic fundus, highlighting AI’s growing potential as a diagnostic support tool.

This study underscores the transformative potential of AI in healthcare—augmenting doctors’ expertise, speeding up diagnosis, and potentially preventing vision loss through early detection.