Faceprint founder Erin Smith started working on her diagnostic tech, which tracks the development of Parkinson's through facial expressions, after watching videos of Michael J Fox
Health-tech startup FacePrint aims to diagnose Parkinson’s disease from Facebook photos, but it started with a journey down the YouTube rabbit hole.
Around three years ago, founder Erin Smith, then a high-school student in Kansas, was watching an interview with actor Michael J Fox, the star of Back to the Futurewho was diagnosed with Parkinson’s disease at the age of 29. Smith, who is now 19, noticed something odd about the way Fox smiled. His facial expressions seemed forced and emotionally distant, even in response to genuine emotions.
“Michael J Fox is a really interesting case study, because you can go back to the films he was in, and create kind of a timeline of when these facial differences started to occur,” Smith says. As she clicked through and watched more interviews with people with Parkinson’s – a degenerative brain disorder which can cause tremors and slurred speech – an idea started to form.
A few years prior, Smith had been obsessed with a prime-time detective show called Lie to Me, which starred Tim Roth as Cal Lightman, a scientist who was an expert at reading people’s expressions. The show was loosely based on the work of psychologist Paul Ekman, who has spent decades analysing and cataloguing “micro-expressions” – tiny gestures and movements of the facial muscles which are too quick for most people to notice.
Smith spent hours reading blogs about the show and the scientific research it was based on. “For me, there was just a fundamental curiosity about what else the science behind facial expressions could reveal,” she says.
She decided to run a controlled study. After writing to a local Parkinson’s support group, she spent her spring break running around her hometown with a backpack and two laptops, visiting local people with Parkinson’s as well as healthy control subjects to capture videos of their facial expressions. Smith collected data on both spontaneous expressions (in response to video clips), and posed expressions (where her subjects tried to mirror expressions they were shown).
Once she had collected the data, she needed to analyse it. “I hadn’t coded before, and basically locked myself in my house and got a tonne of different Coding for Dummies books to teach myself the fundamentals,” she says. “And then it was just building and learning as I went, continuously refining the technology and improving my own skills.”
With the help of Affdex, an off-the-shelf facial recognition software package, she was able to demonstrate that there was a measurable difference in facial expressions between people with Parkinson’s and people without. “It was all honestly just driven by curiosity, I was really captivated with this idea,” she says. “After that, it was the human element that motivated me to scale up the process and continue the research.”
Smith contacted the Michael J Fox Foundation to tell them about her work, and they helped her run two more pilot studies. “In order to really get the most out of these differences and develop prediction and detection and monitoring algorithms, I knew I needed to scale up data collection,” she explains.
Her technology is now the basis for FacePrint, which is still based in San Francisco (and which won the startup showcase at WIRED Health in March). Smith has deferred her offer of a place at Stanford University to work on the company, with the help of a two-year, $100,000 (£79,000) fellowship from the Thiel Foundation, funded by the PayPal billionaire Peter Thiel.
FacePrint is now launching clinical trials, and Smith is working with software engineers to create an automated web or smartphone-based detection and monitoring system, which she hopes will help cut down on rates of misdiagnosis – a particular problem in a primary care setting. The company is building a five-minute facial expressions test that can be administered at home or at the doctor’s, via a smartphone, laptop, or pretty much any device with a camera. Eventually, Smith says, doctors could wear Google Glass-style headsets which could detect signs of Parkinson’s during a normal conversation.
The facial recognition software that FacePrint is built on is also compatible with social media platforms. “Ultimately, it would be possible to integrate FacePrint into those systems, and turn social media platforms into a really powerful healthcare tool,” says Smith – although she stresses the need to do so in an ethical way. (You wouldn’t want a sudden push notification: “You might have Parkinson’s.”)
Smith’s ambitions stretch beyond just one disease. By early June, she plans to have launched Project FacePrint, which will extend facial impairment research beyond Parkinson’s through a crowdsourced citizen science effort. “Through the process of designing this, I’ve gained a really immersive view into the healthcare system,” Smith says. “My mission is to redesign different elements of healthcare. For me, I view this technology as one piece of a larger whole.”
https://www.wired.co.uk/article/parkinsons-disease-diagnosis-faceprint?fbclid=IwAR1DN9WsM5H2ANa7VizoHw60SlResxtrCBkDXGYLfgaE5crOMPRL7OQeOr4