Jan 20, 2022 12:00:00

AI that scans 'human eyes' and predicts whether to die prematurely

An international research team from China, Australia and Germany has announced that it has developed a 'deep learning model that scans a person's eyes and predicts whether that person will die prematurely.' The newly developed algorithm can predict the risk of premature death of a person with high accuracy based on the gap between the biological age predicted from the

retina and the actual age.

Retinal age gap as a predictive biomarker for mortality risk | British Journal of Ophthalmology
https://bjo.bmj.com/content/early/2021/11/17/bjophthalmol-2021-319807

Something in Your Eyes May Reveal if You're at Risk of Early Death, Study Shows
https://www.sciencealert.com/looking-deep-into-your-eyes-could-reveal-whether-you-re-a-fast-ager

The research team is based on a total of 80,169 fundus images taken by 46,6969 subjects enrolled in UK Biobank, of which 19,200 fundus images of 11,052 with no prior medical history. We trained and verified a deep learning model using. The deep learning model developed by the research team was trained to predict a person's biological age simply by scanning the retina, which converts visual information into neural signals.

After that, using the developed deep learning model, we scanned the fundus image data including the remaining 35,917 people and compared the predicted age with the actual age of the subjects. The research team then reported that the deep learning model was able to predict the actual age of the subject with an error of less than 3.55 years on average.

In addition, 1871 subjects had died by about 11 years after the fundus image was taken, and an analysis comparing the gap between biological age and actual age by retinal scan with this mortality data. I also did. As a result, it was also found that the older the biological age predicted by the retinal scan than the actual age, the higher the risk of premature death. For example, if the age predicted by the deep learning model was one year older than the actual age, the risk of death in the following 11 years increased by 2%. On the other hand, death from cardiovascular disease and cancer was not associated with the biological-age-real-age gap from retinal scans.

The results of this study are based on observation only, and the biological factors that link the increase in age and mortality due to retinal scan are unknown, but 'the retina is a very sensitive tissue to aging.' It can be said to support past research. The retina contains both blood vessels and nerves, and may convey important information about the health of blood vessels and the brain.

Studies have been conducted to detect signs of cardiovascular disease , kidney disease, and aging from retinal scans, but the 'gap between biological age and actual age by retinal scan' can be used to predict mortality. This is the first time that he has shown the possibility. 'The important association between the retina-real-age gap and mortality from factors other than cardiovascular disease and cancer is neurological in the retina, along with increased evidence of a link between the brain and the eye. It may support the notion of being the 'window' of the disease. '

Past studies have also shown that fundus images can predict the risk of cardiovascular disease, but recent advances in medicine have reduced cardiovascular disease-related mortality, leading to the biological age and heart of the retina. It is possible that the association of mortality from vascular disease has decreased. In addition, since only 20 subjects died from dementia in total, the research team could not link the biological age expected from the retina with the risk of developing dementia.

Google researchers announce technology that can predict heart disease just by scanning the eyes-GIGAZINE

Other methods of determining the biological age of the body include DNA methylation and transcriptome measurements, but these methods are less accurate than retinal scans and have cost and physical burden issues. Retinal scans, on the other hand, can be easily completed in less than 5 minutes, so further research into the biological association between the retina and the body could lead to better tools for clinicians to measure a patient's mortality risk. It is said that there is.