Gero Lifespan
Russian researchers have taught artificial intelligence to predict the probability of death according to the indications of a fitness tracker
Interface of the Gero Lifespan application.
Researchers of the Russian company Gero, in collaboration with a group of machine learning specialists from ActivBC, trained neural networks to predict the probability of death based on data from a wearable physical activity tracker. The results of the work were published on March 26 in the journal Scientific Reports (Pyrkov et al., Extracting biological age from biomedical data via deep learning: too much of a good thing?).
The accumulation of a huge amount of human health data contributes to the training of artificial intelligence systems on these databases and the increasingly frequent use of AI for medical purposes. With the help of artificial intelligence, tomograms and cardiograms are already being analyzed, diagnoses are being made, and strategies for treating diseases are being determined.
Recently, there has been a tendency to perceive aging as a potentially curable disease and, accordingly, to use artificial intelligence both in search of significant markers of this process and ways to slow it down, prevent and reverse it.
From a biological point of view, aging is an increase in the risk of disease and death with age. However, the biological age of people of the same calendar age may differ by a dozen years. Therefore, in the prevention and therapy of aging, it is logical to focus not on the formal calendar age, but on timely diagnosed real age-related changes.
One of the important results of the new study is convincing evidence that age–related changes, and therefore an increase in the risk of death ("acceleration of aging"), clearly correlate with changes in the profile of motor activity. The latter is easily fixed by conventional fitness trackers - wearable bracelets with an accelerometer capable of collecting and transmitting motion data.
The authors of the work relied on the medical data of 10,000 people collected in 2003-2006 during the national NHANES study in the USA. The NHANES research database contains information about how people with different health conditions moved while continuously wearing a fitness tracker: how often they switched from movement to rest, how many steps they took per unit of time, what intensity of physical activity was the maximum for them.
Having trained a neural network using deep learning algorithms based on these data, Russian scientists have obtained a system capable of linking certain repetitive sequences of movements with medical history data and analysis indicators and – as a result – determining the risk of death based only on data from fitness trackers with higher accuracy than traditional methods allow.
Based on the obtained algorithm, the researchers created a mobile application Gero Lifespan, the beta version of which can already be downloaded and installed on the iPhone.
Scientists believe that the algorithm they have developed will be useful for monitoring health and mortality risk and timely medical intervention.
In response to the concerns of the correspondent of the portal "XX2 CENTURY" about whether the algorithm will not show an increase in the risk of mortality during periods of brief changes in movement patterns associated, for example, with a mild cold, and thereby contribute to excessive neuroticism, the head of the study is a candidate of physical and mathematical sciences, head of the laboratory for modeling biological systems at MIPT and a scientific Director of Gero Pyotr Fedichev replied:
"Yes, during a cold, life expectancy will decrease, but it will quickly return to the previous level – the same thing happens, for example, with the indicators of blood tests. It will be necessary to pay attention to long-term trends."
The researchers also believe that companies engaged in health and life insurance and medical insurance services will be able to remotely identify people from risk groups using the method they have developed and optimize work with them. When asked by a journalist whether this means that those who have a risk of death, according to the tracker, will be quite large, will not be able to receive the necessary insurance services or will have to pay more for them, a representative of the company Gero replied:
"It's a little more complicated. The fact is that people from a high risk group still pay more for insurance today, it's just that the risk is assessed using a questionnaire and a survey. In this sense, we are changing the instrument, but we are not changing the state of the human population, we are pointing to the same people who can be identified by classical methods, it is simply more difficult. But there are two positive points. Firstly, dynamic monitoring using non-invasive technology allows you to detect the risk earlier, which means that you can take measures earlier and eventually reduce both the cost of treatment and the cost of insurance for such a person. Secondly, already now foreign companies are switching to a model of encouraging people who lead a healthy lifestyle. The usual model in insurance is negative – an increasing coefficient for chronic diseases, age, bad habits – and tools such as ours facilitate the transition to positive reinforcement, allowing you to track a person's progress in maintaining health in dynamics."
Portal "Eternal youth" http://vechnayamolodost.ru