3D facial analysis to detect disease

A Perth doctor is creating the first 3D photo library of Australian Aboriginal children’s faces, an innovative medical resource that could improve the diagnosis of rare diseases. 

Gareth Baynam hopes new research into human faces, or ‘biological billboards’ as he terms them, using 3D technology can assist doctors in the early diagnosis of disease, while also improving health outcomes for those living in remote communities.

It is estimated up to 8,000 different rare diseases affect one in 12 Australians, and are commonly life threatening or chronically debilitating.

These include conditions such as muscular dystrophy, cystic fibrosis, Huntington’s disease, early-onset Alzheimer’s and Parkinson’s disease, as well as foetal alcohol syndrome.

About half of rare disease cases occur in children and account for 35 per cent of deaths before a child’s first birthday.

Initially trained in paediatrics, Dr Baynam is a clinical geneticist at Genetic Services Western Australia. He has been trialling the 3D imaging technology for the past six years as method of examining rare diseases.

“Our face is a biological billboard that advertises our physical and mental wellness, our ageing and disease, and approximately one third of genetic and rare diseases are thought to have subtle facial cues,” Dr Baynam told Business News.

“But these facial variations are individually rare so a general practitioner, paediatrician or specialist might see none or a few of these individual cues with that particular syndrome in their working life.”

Dr Baynam said using the objectivity and precision of 3D facial analysis to compare and collate facial pattern information could improve detection by picking up subtle variations in facial contours often invisible to the naked eye.

In 2017, Dr Baynam will lead ‘Pilbara Faces’, a three-year health initiative that aims to create the first 3D-photo resource of Aboriginal children’s faces, which will assist in rare disease diagnosis, treatment monitoring and clinical research.

The Princess Margaret Hospital Foundation’s Emerging Leaders in Philanthropy group recently secured $298,000 in funding for the program from the Roy Hill Community Foundation, which is a registered charity that focuses on initiatives in the Pilbara.

In time, Pilbara Faces will reduce the need for families to travel from their remote communities to a regional or even Perth hospital for diagnosis.

“We’re not at the point where we take a photo and get all the answers, the technology is still being built and implemented,” Dr Baynam said.

“The key outcomes to my mode so far have been developing the means to engage communities and create bridges across multiple sectors to build a new analytic capacity in WA.”

Dr Baynam has already formed numerous local and international collaborations in order to pool a sufficient number of facial images to make comparisons and investigate new detection approaches.

His team at King Edward Memorial Hospital and Princess Margaret Hospital have been at the forefront of a 3D surface imaging technology known as stereophotogrammetry.

“In Perth we’ve developed a beta version called ‘3D Fast’ that can show a face as a 3D image; we can look at and manipulate that image to put some standard landmarks on a face,” Dr Baynam said.

“Our analysis can take out about 10,000 data points, then you can compare those data points and how they relate in that individual’s face and across the population.”

The technology disrupts the traditional clinical practice of comparing a panel of 2D photo configuration standards, where a tape measure is used to record a child’s physical proportions such as head circumference, eye space, and upper-lip-to-nose ratio.

The non-invasive 3D photo test and analysis only takes about 20 minutes.

“A child sits in a seat in front of the camera, you press click on the mouse, flash goes off and in less than two milliseconds the image is taken,” Dr Baynam said.

“The computer software then stitches that image together from the different cameras and creates a 3D photo.”

He said although there was no definitive number of faces needed, a sample of hundreds of faces from different regions would provide the relevant conditions to make comparisons, as accuracy depended on determining the normal range of facial variation across the Aboriginal community.

“There are different guides for Caucasian and African American children, so that may equally be the case for Australian Aboriginal children,” Dr Baynam said.

Future developments of the technology involve exploring the potential of turning pixels into text descriptions of the facial variation detected, generated once a 3D image is taken.

Dr Baynam said while the technology had so far proved useful in a scientific sense, it also held value in its scope for community engagement.

“One study that looked at facial variation between Caucasians and African Americans found that about 9 per cent of total variation was due to ethnic background,” he said.

“That’s one of the beauties of this approach – while it investigates diversity, it’s also a message about harmony.

“We’re 91 per cent similar, at least according to that study, and that to me is a lovely thing.”