Eye diseases found early with highly accurate AI-based system in Japan

According to The Asahi Shimbun, people who are unaware that their eyes are succumbing to disease can now receive an early diagnosis, thanks to an artificial intelligence (AI)-based system with over 80 percent accuracy.
The system uses AI to analyse images of the eye taken using optical coherence tomography (OCT), which show the centre of the retina.
It was created by Nagoya City University and Tokyo-based information system developer Cresco Ltd. They are hoping the new system will be adopted at hospitals to give patients thorough physical check-ups, and for other uses.
“Combining OCT imaging and AI diagnosis for thorough check-ups will lead to the early detection and treatment of diseases,” said Tsutomu Yasukawa, an associate professor of vision science at the university’s Graduate School of Medical Sciences, who is a member of the team behind the system.
The equipment records patients’ eyes at ultrahigh resolution of several micrometres in an imaging process that requires just a few seconds.
Even when patients do not complain of any eye conditions, the system can detect early signs of disease, such as accumulated deposits and abnormal blood vessels, according to the researchers.
As the first step toward developing the diagnostic system, 1,200 CT images of healthy and unhealthy eyes of 300 people were prepared.
Physicians with at least 20 years of clinical experience examined 1,100 of such images and made their own diagnoses for AI to learn their skills.
The other 100 CT images were analysed by AI so that it would identify five possible diseases for each image.
The AI diagnostic system's first most likely diagnoses were consistent with doctors’ diagnoses in 83 percent of the images.
The accuracy rate rose to 90 percent when AI's second most likely diagnoses were included. The remaining 10 percent showed signs of rare diseases, according to the researchers.
OCT images are commonly used for diagnosis at medical institutions. But even a single disease causes many abnormalities and patients often contract multiple diseases, so imaging diagnosis requires technical knowledge and experience.

Ref: http://www.asahi.com/ajw/articles/AJ201612020002.html
 
If you want to read this article in Japanese, please see the following link:
http://www.j-abc.com/jp-blog/-ai-8 
​