A new longitudinal source of health data and facial analytics

Open slide in overlay

Yet, there may be another, simpler means of ascertaining the progress of healthy ageing in individuals. Through clinical testing, it would appear that humans are quite accurate in estimating the biological age of their fellow citizens based just on their facial profile. Just as faces are extraordinarily powerful communication tools in expressing emotion, they also provide a wealth of markers to indicate how healthy an individual is and as a result suggestions as to their biological age.

Figure 1. Environmental and genetic factors in facial aging in twins [1]

In a photo of a pair of twins, one had never smoked, one smoked over many years. With just the naked eye, it is clear the face on the right belonged to the smoker. The eyes are more lined, the skin more flecked, the age lines around the mouth more prominent. If the human eye can tell this, then it is possible to programme computers to code and record such differences. This could be equally achieved through artificial intelligence, where computers have already been trained to learn the differences between facial types. Indeed, facial analytics technology can be used in estimating body mass index (BMI), drug use and a range of other conditions. In tests, facial analytics was able to estimate chronological age within an error rate of +/- 3 years; an 85% accuracy in smoking; and a 79% accuracy rate in BMI – and these accuracy rates are improving as sample sizes increase. It turns out that plastic surgeons are particularly accurate in assessing the age of their patients.

The marriage of facial analytics, biodemography, machine learning, and wearables represent new disruptive technologies for life insurance underwriting. Some life and health insurers are well on the path to finding underwriting metrics in fitness data. Step count measurements from wearable sensors can be changed into quantifiable risk measurements already; and can be used to verify reported levels of physical activity, which can generate a measure of healthy life expectancy. We recently developed an ability to draw on sleep pattern, blood pressure and fasting blood sugar data to generate verifiable measures of risk.

Eventually and foreseeably, wearable sensors will give rise to a new health data economy; data will be used by personal physicians to track and enhance quality of life; and a new form of longitudinal health data will be created. Some of this is already happening among high net worth individuals taking out significant life policies.

References:
[1] cwru.pure.elsevier.com/David J. Rowe, Bahman Guyuron. 2010. Environmental and genetic factors in facial aging in twins. [ONLINE] Available at: https://cwru.pure.elsevier.com/en/publications/environmental-and-genetic-factors-in-facial-aging-in-twins-2. [Accessed 22 December 2016].

 

Summary of Jay Olshansky's presentation at the Centre's Health monitoring event in November 2016. Jay is Co-Founder and Chief Scientist of Lapetus Solutions. Summary by Simon Woodward.