Researchers at Yeouido St. Mary’s Hospital have developed an artificial intelligence (AI)-based real-time glaucoma vision test result extraction method and proved its efficacy.
Glaucoma is a disease that causes abnormalities in the optic nerve function from compressed optic nerve or impaired blood supply due to causes such as increased intraocular pressure. Damage to the optic nerve can lead to vision loss and eventually blindness.
According to the hospital, it is essential to analyze the visual field changes for glaucoma diagnosis. However, the visual field test results were in the form of images in the hospital medical information system, which, in turn, took a lot of time to analyze.
Therefore, the team, led by Professor Jang Dong-jin of the Department of Ophthalmology, developed an artificial intelligence model that can digitize and analyze accumulated visual field test image big data in real-time and conducted a study to analyze its accuracy.
In developing the system, the team analyzed 325,310 vision test papers and extracted 5,530,270 pieces of information from 2010 to 2019.
Afterward, the team applied two independently developed AI models -- semi-AI digitizer and full-AI digitizer -- to analyze the inspection processing results, time, and accuracy.
The semi-AI digitizer extracts major clinical information through text recognition after processing the classification and location of key information according to the rules set by the medical staff when analyzing the visual field examination image, and the full-AI digitizer acquires clinical information through text recognition after deep-learning the visual field test image.
The study results confirmed that the semi-AI digitizer and full-AI digitizer extraction accuracy was 99.3 and 98.3 percent, respectively.
“The team has integrated the research results to the hospital’s anonymized medical information analysis system, enabling analysis by linking data and visual field information from various clinical departments as well as ophthalmology,” Professor Jang said. “It is meaningful that the study laid the foundation for AI-based glaucoma research, and we expect that the system will help analyze test results in various clinical tests with images.”
Computer Methods and Programs in Biomedicine published the results of the study.