Artificial intelligence for early diabetes diagnosis

Diabetes affects over ten percent of Americans, but many of those with the disease are undiagnosed. Leaving diabetes untreated harms essential organs and damages blood vessels, causing them to swell, leak, or grow abnormally. When this occurs in the eyes, a condition known as diabetic retinopathy, it causes blindness.

Deepa and Sivasamy examined the benefits and ethical implications of utilizing artificial intelligence (AI) to detect warning signs of diabetic retinopathy for early diabetes intervention.

Retinopathy screens detect signs of damage using a special camera that captures detailed images of the back of the eye. Typically, doctors carefully examine these images, looking for diabetes warning signs in the retina’s blood vessels caused by high blood sugar. If found, the patient should be tested for diabetes to prevent further damage.

Algorithms trained to identify blood vessel damage can enhance this process by identifying warning signs quickly, remotely, and at earlier stages.

“AI and deep learning models have shown remarkable promise in the early detection of diabetic retinopathy,” said author Rangasamy Deepa. “These systems can analyze retinal images to identify signs of retinopathy and provide automated assessments. They can be integrated into primary care settings and telemedicine platforms, improving accessibility.”

The authors warn that ethical concerns about AI’s inherent bias and privacy must be addressed. They recommend additional work to improve accuracy, reduce costs, and integrate with healthcare systems. Still, this work highlights the potential for AI in medicine.

“We emphasize how AI techniques are revolutionizing medical diagnostics by accurately predicting diabetes risk and automating the analysis of retinal images for retinopathy detection,” said Deepa. “This technology can significantly enhance healthcare efficiency overall.”

Source: “Advancements in early detection of diabetes and diabetic retinopathy screening using artificial intelligence,” by Rangasamy Deepa and A. Sivasamy, AIP Advances (2023). The article can be accessed at https://doi.org/10.1063/5.0172226 .