Telegraf – Every natural disaster – fire, flood, avalanche, volcano, tsunami, cyclone – brings its own extraordinary challenges in rescue, but also in trying to name the dead. This is called disaster victim identification, or DVI, a topic the head of Monash’s Department of Forensic Medicine, Professor Richard Bassed, specialises in. Professor Bassed is also deputy director of the Victorian Institute of Forensic Medicine (VIFM). He says within five years, funding permitting, a fledgling project with the Department of Defence using AI facial recognition to better-identify the dead could be up and running.
“Very early stages,” he says, “but we are thinking that we might be able to identify quite large numbers of people, and thus reduce the cost and time required to identify the remainder.”
We’ll return to that. First, Türkiye-Syria, and the sheer scale of it, in a highly earthquake-prone part of the world.
“Remember that in an earthquake like this, you lose your infrastructure,” Professor Bassed says. “Hospitals damaged, mortuaries damaged, probably no power and water. There’ll be massive problems in gathering the deceased, recovering them all, and there’ll also be massive problems in working out where to take them to. You need to have one central location where all the deceased people can be taken to, or a few, if they’re widely spaced geographically.”
Identifying the deceased
The next step is figuring out who is missing and figuring out the names of the dead.
“If a whole family is dead under rubble from an apartment, and no one knows them very well, how are you going to even know they’re missing? It’s about trying to get an accurate toll of the number of missing people, and then collecting all the ante-mortem information you can about the deceased to try and identify them, too.”
Both Türkiye and Syria are member countries of Interpol, which designed and implemented the standard operating procedure for disaster victim identification. It has four stages:
- Recovering bodies and gathering any identifying information at the scene
- Examine human remains in the morgue looking at teeth, fingerprints, DNA and distinguishing features such as tattoos
- The ante-mortem phase, gathering dental and medical records and DNA from family
- A “reconciliation” phase of comparing the above to identify victims.
But current news reports say bodies are being buried in mass graves on site in Türkiye-Syria, many without being formally identified, with a shortage of forensic investigators on the ground.
An Australian study published in Forensic Science, Medicine and Pathology journal after the Bali bombings and Black Saturday bushfires states that in the aftermath of a mass disaster, a “critical” issue is:
Professor Bassed says that after a mass disaster, with infrastructure levelled and perhaps hundreds of thousands dead, a confronting question can be asked by governments – “a cold calculus of cost-benefit analysis,” he says, “of ‘Are you better doing things for the living, repairing houses and infrastructure, or do you spend money on a massive recovery and identification process for the deceased?'”
Complex geopolitics can also prevent disaster victim identification.
“The war graves in Iraq and Iran and Syria and everywhere else? People aren’t trying to do anything there, partly because of the political problems in some of those countries, but also because of the huge, huge cost and effort to do it. There’s not enough expertise in the world. When they did Srebrenica, after the Balkans War, the mass-fatality exhumations and identifications took them 10 years to identify 30,000 people. So imagine how long it takes to identify 300,000.”
Utilising facial recognition technology
In the reasonably near future things might be different, which brings us back to the VIFM research collaboration with the Department of Defence and its Defence Science and Technology Group.
The aim is to test the effectiveness of current facial recognition technologies to help identify the deceased.
“At the moment we’re comparing commercial facial recognition systems and how successful they are in recognising the dead. They’re deceased people who are either freshly deceased, so they look the same as they did when they were alive, all the way through to really traumatised or decomposed people. We think what’s going to happen is that we’re going to need to either tweak a current algorithm or create our own facial recognition algorithm that will work reliably with photographs of them when the body is found.”
“Then you can start comparing the photos while alive with the post-mortem images of the deceased through a machine-learning facial recognition algorithm. It won’t work in bushfires as well, because there are often no facial features left.”
Another promising VIFM project involves getting full-body CT scans of every Victorian body that goes to the institute. About 100,000 have already been collected since 2005. The scan shows the bones, but it also shows how a face sits on the bones. The idea is to build an algorithm that can rebuild a face on an unknown skull, and can even be combined with DNA to build hair colour, eye colour or, for example, nose shape.
“When I first started in this business,” says Professor Bassed, “it was a plaster cast of the skull, and you stick pins in it to where you thought the depth of the tissue was over bony points, and you build a face with clay. There was no science in it.”
VIFM is also working on a project with Monash’s Faculty of Information Technology – featured here by the ABC – developing “virtual” autopsies using an augmented reality headset and 3D visualised bodies, reducing the need for invasive physical autopsies.