This new computer program can use a digital picture to help diagnose rare disorders.
According to New Scientist, 6 percent of all people in the world are thought to have rare genetic disorders. In first world countries genetic testing is available for some of the more common ones, like Down Syndrome; but the rarer disorders require a visit with a geneticist that is trained to make diagnoses based on pronounced facial features. Regrettably, there are very few people who are trained to do this, and for many that means they might never get an accurate diagnosis. Due to the hard work of a group of scientists at the University of Oxford, these life-changing analyses will soon be more available.
Christoffer Nellåker and Andrew Zisserman oversaw a team at the University of Oxford and together have created a new software that is capable of changing the future of healthcare. Using this technology, computers are able to analyze digital photographs and compare them to publicly available photos of people with genetic disorders. This helped the computer to learn to identify conditions based off factors like the shape and size of eyes, eyebrows, lips, and noses. This testing revealed that 93 percent of all predictions were accurate.
Initially, computers were fed 1363 images of people with eight different genetic disorders, each of which was represented by 100 to 283 images, but those numbers have increased. Currently, the software is able to identify 90 different rare genetic disorders, and the database contains 2754 images of people with these disorders. While the system can not definitely make a diagnosis, studies have shown that the program makes it 30 times more likely that a correct diagnosis will be determined compared to a clinician alone.
As beneficial as this software could be for people in first world countries, its impact will be much greater in developing countries without access to doctors and medical supplies. A photo could be taken on a smartphone and sent to a computer halfway across the world, resulting in a prompt diagnosis.
Nellåker is hoping to train the algorithm to analyze frontal as well as profile pictures. In addition, he hopes to pair the system with DNA analyzing programs so that facial and genetic features of disorders can be evaluated together. Regardless of whether or not those ideas ever come to fruition, it’s clear that this software marks the beginning of more accessible genetic diagnoses, which in turn should lead to better treatment.