Working with sensors to measure health

/ Author: Miriam van der Sangen
Sensors have become an essential tool of everyday life. Think for example about the blood pressure monitor inside a smartwatch or the step counter on a smartphone. Sensors are accurate, which is why Statistics Netherlands (CBS) is exploring various ways to implement them – e.g. as a substitute for survey questionnaires. A number of experiments to this end were conducted at the second Sensor Data Challenge event in January 2019. The event was co-hosted by Statistics Netherlands, The Hague University of Applied Sciences, the National Institute for Public Health and the Environment (RIVM) and Utrecht University.

Spread of infectious diseases

Winner of this challenge was the Team Intelligence (including Utrecht University, University of Amsterdam and Carlos III). Winning team member Mathijs Nelemans: ‘Our assignment - contributed by RIVM - was to recognise human contact patterns. These are an important factor in the spread of infectious diseases.’ The team built a small model to find out the extent to which certain individuals come into contact with others. RIVM collects such data by asking people to complete surveys. The disadvantage is that these are conducted afterwards and people may make a mistake or forget data. Nelemans: ‘To collect data on hand touches and estimate the gender and age of these individuals, we made use of the bluetooth feature on smartphones as well as camera glasses. Producing a model which could determine the age of people was not an easy job.’



It was the first time to join this type of large data challenge event for Mathijs Nelemans, a student of Geographical Information Management and Applications (GIMA) at Utrecht University and employee at GeoPhy, a technology company based in Delft. He was persuaded to join this event by a fellow student who attended the previous Sensor Data Challenge. ‘As a team, we did do some preparing and we studied the different assignments beforehand. We also brought our own router and electronics.’ The challenge involved hard work – even until after midnight. ‘Things went reasonably well until about 4 AM. That’s when participants were hit by fatigue. But by 8 AM we went straight back to work. When our demo was succesful we thought we would qualify for a prize.’

Working conditions

Second runner-up Feline Vis van Heemst was working with three fellow students in the L!nk team, to collect objective, reliable data on labour conditions. For her as well, this data challenge was a new experience. Her assignment was based on the National Survey on Working Conditions NEA - produced jointly by CBS and TNO. ‘First, we looked up the different things that are monitored in the NEA - for instance lifting, or noise pollution - and checked whether these could be properly measured with sensors. We used an ear sensor to measure ambient noise, heart rate and harmful gases, and a pressure sensor under the feet for the number of lifting movements and the lifting weight.’ Vis van Heemst, in her fifth year of Industrial Product Design (IPO) at The Hague University of Applied Sciences, thought that programming all the sensors was a pretty time-consuming job. ‘Luckily we had a team member with lots of experience in this field.’ They were surprised to win the second prize for their contribution. ‘We did not know each other as team members and this form of cooperation was new to us. But even though we only slept three hours, we achieved fine results.’

To CBS, the most interesting experiences are those related to measurement of working conditions

Out of the box

CBS also sent four team members, who represented the disciplines of methodology, business analysis and IT. One of them was Sapphire Han. She has worked as a methodologist at CBS since September 2018 and is currently writing her thesis at the Netherlands Interdisciplinary Demographic Institute (NIDI). Before that she did a Master’s in biomedical technology at the University of Twente in Enschede. Han found the ‘out of the box’ thinking at the data challenge especially stimulating. ‘We had the same assignment as the team that won second prize. We also studied the NEA survey questions first. We brought our own sensors, but we also assembled a sensor hub during the challenge. That sensor hub consisted of sensors that measure humidity, map body movements, and register walking and sitting. Using machine learning techniques we performed a data analysis. Approximately 90 percent of all movements could be mapped with the help of our sensors. That was a fine result and it earned us the third spot.’

Translation into practice

This 24-hour Data Challenge was co-organised by CBS, Utrecht University (the innovative survey network WIN), the Smart Sensor Systems lectorate at The Hague University of Applied Sciences and the National Institute for Public Health and the Environment (RIVM). Altogether eight teams from government, academia and universities participated in this challenge. Barry Schouten is a Senior Methodologist at CBS and professor in Methods and Techniques at Utrecht University. ‘We look back on a highly successful data challenge. Everything went extremely well. Together with RIVM, The Hague University of Applied Sciences, Utrecht University and the teams, we are going to study how we can use the ideas in actual practice. To CBS, the most interesting experiences are those related to measurement of working conditions.’ Another enthusiastic response comes from Rowan Voermans, data scientist and specialised in the application of sensors at CBS. ‘The teams partly consisted of volunteers and partly of students who could earn credits by participating. Most teams were notably well-prepared. Each team also had a unique approach.’