Cup-to-Disc Ratio Grading is Accurately Measurable Using Artificial Intelligence
Cup-to-disc (C/D) ratio grading appears to be accurately measurable using artificial intelligence (AI), as the process was found to moderately correlate with traditional grading, according to a prospective study involving 45-degree color images. The results were presented in a poster presentation by Subhashini Chandrasekaran, BS, of the Institute of Ophthalmology and Visual Science at Rutgers University School of Medicine in Newark, NJ, during the American Academy of Ophthalmology’s 2018 annual meeting in Chicago.
The AI method was compared with glaucoma specialist grading. The images were captured using a nonmydriatic retinal camera and evaluated on a monitor using American Telemedicine Association standards. Images were then uploaded to an AI program to automatically measure C/D and image quality. Among the findings:
- Results were available within 35 seconds.
- A moderate positive Pearson correlation was observed in C/D ratio in both right and left eyes (0.72, 0.71).
- Each method was able to capture C/D ratio in 95% of the images.
- AI was unable to provide a C/D ratio in 4% of the images that the ophthalmologist was able to.
- There was a moderate positive correlation (0.55) for image quality.
Chandrasekaran S. Traditional vs. cloud-based artificial intelligence grading for glaucoma images. Talk presented at: AAO 2018 annual meeting; October, 26-30, 2018; Chicago.