The AI diagnostic device DrNoon captures and analyzes retinal images to detect various eye diseases as well as cardiovascular diseases. MediWhale Inc. a domestic eye image-based health diagnosis start-up, developed it in collaboration with Yonsei University Severance Hospital. DrNoon is drawing attention with the world’s highest level of accuracy, 95 percent of eye diseases and 83 percent of cardiovascular diseases.
So how can we diagnose eye disease and predict heart disease with a fundus photo of the retina inside the eye? Tae-geun Choi, CEO of Mediwhale, attended an open seminar hosted by YEHS, a next-generation engineering leader at Ewha-Womans University, and introduced AI technology that recognizes diseases through retinal images.
Under the theme of “How to predict cardiovascular disease with retinal images, simple diagnosis of diabetic retinopathy, and safe video as a new biomarker for heart disease,” we had time to share Mediwhale’s technology with next-generation engineering leaders through this open seminar.
How do you predict cardiovascular disease through the eye images?
He said, “It is the only organ that shows human blood vessels directly, non-invasive, with no the eyeball penetrating the skin or any hole in the body. So it is possible to predict vascular diseases through the eyes,” he explained.
In 2004, Dr. Tien Wong, a top ophthalmologist in the field of ophthalmology, published in the New England Journal of Medicine that “the ratio of the blood vessel thickness of arteries and veins in the retina can predict cardiovascular disease,” and in 2018 Lily Peng wrote in Nature BioMed, “You can accurately predict the age and sex of a person through the eyes, and that is the key factor in the cardiovascular system.”
Using this technology, Google analyzed 120,000 fundus photography of patients with cardiovascular diseases collected in India, and developed AI that predicts people with cardiovascular diseases with a 70 percent probability by deep learning 100,000 photos of retinal coherence tomography collected in the U.K. “AI, ‘DrNoon’, has increased the accuracy to 83% by learning heart CT photographs along with 120,000 fundus photography that have been precisely classified by Severance ophthalmologists,” CEO Choi said.
In addition, DrNoon can even predict the ”Coronary Artery Calcification Index (CACS)”, an official indicator of cardiovascular risk factors. CEO Choi said, “The number one cause of death worldwide is heart disease. The related global medical expenses amount to 800 trillion won. In particular, in the United States, medical expenses of 300 trillion won are incurred annually due to cardiovascular disease. To reduce this, it is important to predict the occurrence of cardiovascular disease, and the fundus photography is excellent in its effectiveness,” he stressed.
It’s easy to diagnose diabetic retinopathy, DrNoon
The big advantage of “DrNoon” is that it is cheap and simple than heart CT and that there is no risk of radiation exposure. DrNoon is currently undergoing clinical trials to obtain medical device approval from the Ministry of Food and Drug Safety. “Last year, the U.S. Food and Drug Administration (FDA) gave its first approval for IDx, an AI-based medical device that diagnoses diabetic retinopathy,” and “This will greatly help early diagnosis of diabetic retinopathy, which appears to be a complication of diabetes.” CEO Choi said.
Diabetic retinopathy is one of the major complications of diabetes patients. Since it occurs in about 90 percent of patients with diabetes more than 30 years old, it can be said that the longer the diabetes prevalence period, the more likely it is to occur in almost all patients. Even if there are no symptoms in the early stages and there is considerable progress in retinopathy, it is difficult to detect the disease with a simple eye test because the vision is good if the macula is not impaired.
“It is essential for diabetic patients to have regular eye examinations, but less than half of them can have eye examinations in internal medicine hospitals,” CEO Choi said. “Because missing the treatment period can lead to blindness, if ‘DrNoon’ is distributed to nearby hospital or health examination centers, it will be a great help for early diagnosis,” he stressed.
He is aware of the importance of early diagnosis more than anyone else because he missed the timing of glaucoma treatment and lost about 40 percent of his vision in his right eye, adding, “If the DrNoon is commercialized, people can easily check the health of their eyes at a nearby hospital or health examination center, and if any abnormalities are found, they can go to an ophthalmology hospital to help maintain healthy vision for more people.”
As a new biomarker for heart disease, Fundus photography
Recently, Professor Lim Hyung-taek, Singapore National Eye Center, co-founder of MediWhale and chief marketing officer in the medical sector, published a paper saying, “The veins and arteries of the retina are related to the risk of stroke, acute myocardial infarction, heart failure, and atrial fibrillation”. Therefore, MediWhale’s goal is to make fundus images a new biomarker for heart disease.
“Our goal is to analyze the fundus image through AI and to present the fastest way to detect the disease through it,” CEO Choi said. “We will expand the scope of diagnosis through the eyes to include eye examinations for ophthalmological and cardiovascular diseases such as cataracts, glaucoma and retinal diseases as well as chronic diseases in the systemic.”
Meanwhile, the Young Engineers Honor Society (YESHS), which hosted the seminar, is a group of next-generation engineering leaders under the National Academy of Engineering of Korea and was established in 2005 to establish a nationwide network of engineering students. Through this open seminar, next-generation engineering leaders had time to learn about technology trends and share knowledge in the AI-based bio industry.