Computers can see genetic disorders in your face

Scientists have developed an artificial intelligence system that can identify rare genetic conditions from a patient’s photograph.

Software has been trained to recognise more than 200 genetic syndromes that result in particular facial features. An artificial-intelligence concept, above. Image via iStock

Scientists have developed an artificial-intelligence system that can identify rare genetic conditions from a photograph of a patient’s face.

The DeepGestalt system offers new hope for parents of children who suffer developmental delays because of genetic variations.

“We hope to shorten the length of the diagnostic odyssey these families often go through,” said Karen Gripp, co-author of a paper published in the journal Nature Medicine.

However, there are fears that the technology could be abused.

Peter Bannister, of the Institution of Engineering and Technology, said: “If your insurance company can tell from a passport photo that you have a genetic disorder, could they discriminate against you?”

The new system, developed by FDNA, an American technology company, has been trained to recognise more than 200 genetic syndromes that result in particular facial features.

People with Angelman syndrome, for instance, which affects the nervous system, tend to have small heads and wide mouths, as well as severe intellectual and physical disabilities.

The system can recognise foetal alcohol syndrome and conditions such as Noonan syndrome, which can involve learning disabilities. It relies on a learning algorithm which was “trained” on about 17,000 labelled images of patients with distinct syndromes.

It is designed to be used by clinical geneticists, Dr Gripp said.

When they enter a photograph of a patient the system gives them a list of possible syndromes.

This information can then be relayed to a laboratory that will analyse the patient’s DNA. Having a shortlist of possible diagnoses narrows the number of genes the lab will focus on.

This should streamline the diagnostic process, and the system has already been used by hundreds of clinical geneticists, Dr Gripp said.

“There are thousands of genetic syndromes and it is not possible for [a human] to know each one. But the AI is never exhausted, in terms of learning.”

The Times