
My research on Automated MGD Assessment Got Featured on South Korea's Largest Online News Portal
I am delighted to share that my research work during my MS at Gwangju Institute of Science and Technology (GIST) was featured in Naver News, South Korea’s largest online news portal. The article, published on August 11, 2022, highlighted our groundbreaking development of an AI-based automated system for diagnosing dry eye syndrome through meibomian gland analysis.
Here is the link: https://n.news.naver.com/article/030/0003036257
The english translation is given below:
AI Technology Developed for Faster and More Accurate Dry Eye Diagnosis Than Doctors
Professors Euiheon Chung (GIST) and Ho Sik Hwang (Catholic University) collaborate on research that proves higher accuracy than clinical diagnosis The Gwangju Institute of Science and Technology (GIST) announced on August 11 that Professor Euiheon Chung’s biomedical engineering team, in collaboration with Professor Ho Sik Hwang’s ophthalmology team from Catholic University (Yeouido St. Mary’s Hospital), has developed AI-based quantitative analysis technology for quick and accurate diagnosis of dry eye syndrome. Dry eye syndrome is broadly categorized into aqueous deficient dry eye, where tear production is insufficient, and evaporative dry eye caused by insufficient oil layer production. Among these, evaporative dry eye is most common in modern people who use their eyes extensively. The meibomian glands, which produce oil and form the tear film layer inside the eyelids, are typically responsible for evaporative dry eye syndrome due to their loss or dysfunction. Therefore, imaging meibomian glands to assess their loss is crucial for accurate diagnosis and appropriate treatment of dry eye syndrome. However, the current method relies on clinicians subjectively scoring meibomian gland loss on a scale of 0-3, which has limitations in accuracy and reproducibility. The research team achieved faster and more accurate diagnostic results than doctors using patient meibomian gland images from actual hospitals and their newly developed deep learning model. Using 1,000 meibomian gland images from Yeouido St. Mary’s Hospital, they marked eyelid and meibomian gland regions and had two dry eye specialists score the gland loss. The deep learning model was trained on 800 randomly selected images and showed more consistent and accurate results than clinical readings when tested on 200 images.
When applied to 600 additional images from Korea University Ansan Hospital, the AI continued to demonstrate faster and more accurate readings than clinical experts. The team has made all research materials and analysis data publicly available online for other researchers to verify and develop new AI technologies.
Professor Chung stated, “We can now quickly, accurately, and objectively diagnose meibomian gland dysfunction, a major cause of dry eye syndrome, using our AI-based deep learning model.” Professor Hwang emphasized that “this research can be easily applied in medical settings since the deep learning model was developed using data from commercial equipment that photographs meibomian glands.” The research was published online in the prestigious ophthalmology journal “The Ocular Surface.”