Gangbuk Samsung Hospital Develops Personalized Algorithm to Predict Progression of Macular Degeneration

[Asia Economy Reporter Kim Young-won] Professor Song Su-jeong of the Department of Ophthalmology at Kangbuk Samsung Hospital and Professor Shin Ji-tae of the Department of Electronic and Electrical Engineering at Sungkyunkwan University, along with their research team, announced on the 2nd that they have developed an algorithm capable of predicting changes in patients with macular degeneration using a type of artificial intelligence method called 'Generative Adversarial Network' (GAN).

Generative Adversarial Networks are models in which a generator and a discriminator compete to produce information values, creating fake images using real images. Several recently popular applications that predict future facial appearances also utilize GANs.

This study was conducted on early to intermediate macular degeneration patients who could be followed up for more than five years at the Health Screening Center and Ophthalmology Department of Kangbuk Samsung Hospital. Based on their fundus photographs, the research team developed an algorithm using GANs that generates expected fundus images 1, 3, and 5 years later when inputting the current fundus images of macular degeneration patients.

In particular, the research team explained that the newly developed algorithm is significant in that it predicts deterioration according to each patient's individual fundus condition, representing personalized disease prediction using artificial intelligence. Until now, developed algorithms for predicting macular degeneration have been limited to simply indicating the risk level of deterioration.

Professor Song Su-jeong of Kangbuk Samsung Hospital's Department of Ophthalmology said, "Macular degeneration ranks first or second as a leading cause of blindness worldwide, so predicting the worsening of early or intermediate macular degeneration holds very important meaning."

Professor Song added, "Although improving the predictive performance and additional external validation are necessary for clinical use of the algorithm, this study is highly meaningful as it realizes future prediction, considered the ultimate stage of artificial intelligence research, in the medical field."

This study was published in the April issue of the journal Computer Methods and Programs in Biomedicine.

This content was produced with the assistance of AI translation services.