Living with diabetes can put a person's vision at risk. Doctors often look for diabetic retinopathy, a condition where high blood sugar damages the blood vessels in the eyes. Detecting these changes early is vital for protecting a patient's sight.
Researchers looked at 17 different studies to see how artificial intelligence performs when checking wide-view eye images. The AI showed an 85 percent sensitivity rate, meaning it was good at catching cases of the disease. However, its specificity was lower at 72.5 percent, which means it sometimes struggled to correctly identify those without the condition.
While the results are promising, the evidence is still early. The researchers noted that the small number of available studies means we need much larger and more diverse data sets. More testing across different types of imaging platforms is also needed before this technology can be fully relied upon in every clinic.
Common questions
How accurate is AI at finding diabetic retinopathy?
The artificial intelligence showed an 85 percent sensitivity rate when checking wide-field eye images. This means it is quite capable of identifying the disease. However, its specificity was 72.5 percent, which means it is less consistent at correctly identifying patients who do not have the condition.
Is this AI technology ready to replace regular eye exams?
Not yet. While the results are promising, the study notes that the findings should be interpreted with caution. Because there are a limited number of studies available, more large and diverse data sets are needed to confirm how well the technology works across different platforms.
What specific eye issues can this AI detect?
The AI was tested on its ability to identify diabetic retinopathy, distinguish it from other retinal disorders, and classify the stage of the disease. It was also used to look for referable retinal problems and predict how the disease might progress over time.
