Using computer vision to improve diagnosis

Facial recognition can do far more than unlock your smartphone.

In 2016, Associate Professor Tracy Dudding-Byth watched a TV segment on facial recognition and its use in counter-terrorism. The program explained that facial recognition could pinpoint and track individual faces in a crowd by using new technology. A task that could take tens of thousands of person-hours takes just milliseconds with the software.

Tracy has a daughter with a rare health condition and has experienced the lengthy diagnostic journey associated with rare diseases.

As a clinical geneticist, Tracy sees children with significant global developmental delay or intellectual disability (ID). ‘All these children are beautifully unique in their appearance, but children with genetic conditions often share similar features.’

Genetic disorders affect millions of people worldwide but, for a parent trying to find a diagnosis for their child, Tracy explains that the odds aren’t in their favour.

Unfortunately, despite testing around 1500 developmental disorder genes in a single analysis, 50–70 per cent of children with moderate to severe ID remain undiagnosed. An estimated 2000 developmental disorder genes are still to be characterised.

Matching facial features is a commonly used approach to discovering new ID genes, as around half of these children and adults have facial features providing a clue to the diagnosis.

When an individual remains undiagnosed, doctors rely on manual face-matching techniques, which are slow, subjective and unreliable.

‘I wondered whether facial recognition technology could help parents find a diagnosis for their child by matching their faces with other diagnosed and undiagnosed children around the world,’ says Tracy.

Using artificial intelligence, FaceMatch can match facial images and is more accurate than the human eye.

FaceMatch is a platform that helps parents and doctors contribute to an international, secure image database of both diagnosed and undiagnosed children across the globe. A user can upload a front profile image of their child’s face to the library, and the software will determine whether the child shares a close facial match with other children within the database.

‘The technology has the potential to ease years of heartache and uncertainty for families and parents faced with the challenges of an intellectually disabled child with an unknown diagnosis,’ says Tracy.

The technology is not just available to doctors and clinicians – Tracy encourages parents to get involved.

Parents of children with both diagnosed and undiagnosed moderate to severe ID are encouraged to upload images of their child at different ages through the secure, ethically approved FaceMatch recruitment site.

Tracy is proud to be a finalist for a Research Australia Health and Medical Research Data Innovation Award.

Please visit FaceMatch.org.au or contact [email protected] if you have any questions.