Hi there! I’m Famke, a bioinformatics student that recently completed her minor research project at Utrecht University. As part of this project, I’ve been doing an internship at The Hyve’s RADAR-base team. My project involved the integration of several Polar devices – such as the Polar Verity Sense, the Polar Vantage V3, and the Polar H10 – to the RADAR-base platform, by developing a new plugin to Passive Mobile App (pRMT).

The RADAR-base pRMT app serves as a central platform for collecting sensor data, streaming information from both phone sensors and various wearable devices. This app depends on plugins to efficiently collect and process the data.

While plugins already exist for wearables like the Empatica E4, Pebble 2, Fitbit Charge 2, Biovotion, and Faros, my job was to develop one specifically for Polar devices. To accomplish this, I utilized Polar’s SDK. The developed Polar plugin for the pRMT app is available in the RADAR-commons-android repository.

Analyzing Polar’s Heart Rate measurements

We integrated these Polar devices into the platform to solve our research question:

‘How do these Polar devices perform in terms of consistency, robustness and quality of their data collected within the RADAR-base environment, compared to other, previously integrated PPG devices?’

As a test case we selected the Polar devices and a Fitbit Charge 2, a device that was easily accessible and has already been widely used in combination with the RADAR-base platform.

To this end, heart rate measurements of 13 participants were taken during different activity phases, using the newly integrated devices and a Fitbit Charge 2 for comparison. Measurements were compared to the Polar H10, as this chest strap uses ECG, which is reported to be a standard in measuring heart rate.

Results

Bland-Altman plots, which are used to visualize the differences between the two measurements against their averages, were created in which HR measurements of the different devices (the Fitbit Charge 2 and the Polar Vantage V3) were compared to HR measurements of an ECG sensor (the Polar H10). Also, Lin’s concordance correlation coefficients (r_c) were calculated:

For the Polar Vantage V3, the Bland-Altman plots revealed a mean bias of 1,21 bpm, of which 95% of values fell within -8.57 and 10.98 bpm. For the Fitbit Charge 2, a mean bias of 5,64 bpm was found, with 95% of values within -14,42 and 25,69 bpm. High agreement was found as the Lin’s concordance correlation coefficients (r_c) were 0,98 (substantial) and 0,91 (moderate) for the Polar Vantage V3 and Fitbit Charge 2, respectively.

Based on these results, we concluded that while each of the analyzed devices is suitable for use in research, the Vantage V3 appears to be the most suitable in terms of consistency, ease of use and accuracy.

Nevertheless, we believe it is crucial for researchers to consider the differences in usage and data processing methodologies of these devices when designing their patient monitoring studies, and that they limit HR data collection throughout larger studies to a single device as much as possible.

If you are interested in reading more details about this research, the full thesis can be found here.

My internship experience

I’m thrilled that I could share here the work I completed during my internship. This project has been an incredible learning experience, and I’m proud to say that our implementation of the Polar devices was a success. For this, I would really like to thank my daily supervisor Bastiaan, as well as the entire RADAR-base development team at The Hyve and King’s College London (KCL). Their support and guidance have been invaluable throughout this project.