Radar Sensing for Sleep Research

Lorna Caddick
11 minutes ago
4 min read

Lorna is a Research Assistant at the University of Edinburgh, with experience in radar sensing for sleep measurement. She is interested in longitudinal links between sleep timing, circadian rhythms and mental health. Lorna led the data collection and co-production on the Ambient Teens Sleep Study, and is particularly passionate about co-production and engaging the younger generation in research. Recently, she has extended this work within the Division of Psychiatry at the University of Edinburgh onto the AMBIENT-BD study. Alongside this, Lorna is working to complete a PhD by publication and was previously awarded a Circadian Mental Health Network Early Career Research Award to support her public engagement activities.

This blog post will cover the Ambient Teens Sleep Study which was a feasibility study funded by MRC in the leadup to the Adolescent Health Study to assess if radar sensing is feasible and accepted by adolescents. It will also look at how radar sensing could be used longitudinally in other populations and how co-production strategies have been incorporated to introduce a newer method of monitoring sleep.

Measuring Sleep

Wearables (such as smartwatches like Fitbits or Garmins) have been a great technology advancement for monitoring health and sleep in a home environment. They are much cheaper and less invasive than gold-standard polysomnography and provide additional health measures during 24-hour monitoring such as physical activity levels. However, there are times when wearable devices are not suitable for monitoring sleep. Young people are less likely to tolerate wearables due to sizing and discomfort, especially over longer time frames. It is not uncommon to remove a wearable device and forget to put back on, especially if charging the device is required, or those with Alzheimer’s/Dementia. Other specific populations such as neurodivergent individuals that experience sensory overload and those with skin sensitives are more likely to remove the wearable, especially for sleeping. Even the typical watch-wearer may opt to remove devices before going to bed.

When wearables are used in research it is typically over short durations, 1-2 weeks. However, these short bursts of data collection miss many key life events and the associated changes to sleep timing. In the context of adolescent development, we simply do not fully understand how sleep patterns change over time and the relationship to other health outcomes. Are the links between sleep timing and mental health an association, predictor, cause/consequence? Therefore, we need alternative methods of monitoring sleep that have minimal input (e.g. contactless) and can remain in place continuously for longer time frames.

Radar Sensing

Radar sensors typically sit near the bedside, recording whole-body movement and respiratory rate. They often have other inbuilt sensors for the room environment such as light, sound, temperature and air quality.

As these devices are less familiar, we first need to address if they are feasible for use and accepted by the target populations. An initial response can be privacy concerns, so we need to work with participants to agree the best approach and instructions for using these devices, such as explaining what it can and cannot measure and providing examples of what the data looks like.

An example of what the sleep data would look like

Ambient Teens Sleep Study

In the Ambient Teens Sleep Study, we recruited 48 young people aged between 8-18, split evenly across three age brackets, 8-11, 12-15 and 16-18 years. They placed the radar sensor in their room for 4 weeks and for 2 of those weeks were also asked to wear an activity band and complete a sleep diary. Read the protocol paper for a more detailed description of methods.

Looking at data completeness as a percentage of expected data, the radar sensor collected 90% of expected nights compared to 79% from the accelerometer. Clean usable data from the radar sensor was 87% of expected compared to 61% from the accelerometer. This shows that radar data is mostly lost from nights missing recordings, i.e., not sleeping in their own bed. However, the accelerometer had higher rates of both missing nights as well as invalid nights due to high non-wear time. Therefore, contactless passive radar sensing demonstrates a promising approach for high compliance over longer time frames due to low participant burden. Protocol compliance is further supported by the ability to monitor if radar sensors are online or offline in real-time. 80% of adolescents preferred the radar sensor to the accelerometer.

Percentage of clean (complete) data from the radar sensor and the wearable

Co-production

We included many patient and public involvement and engagement (PPIE) strategies throughout the study to hear from young people and take on board their feedback. These included young person advisory groups (YPAG) and citizen science engagement activities. Here is a summary of what adolescents encouraged and discouraged and how we implemented these changes:

To be engaged more frequently – we introduced check-in calls during data collection.
To see their own data – we included snippets of their data during the calls and provided individualised summary sleep reports at the end of their data collection.
Youtube as a key source of communication – we created video instructions for device setup and a YPAG video series to share study results.
Young people did not enjoy citizen science activities, reporting it felt like ‘more work’, only 13/48 participants completed it – this was made an optional study component.
Sleep diaries were the least favourite part of the study due to remembering to complete each morning and a feeling of stress to log accurate data e.g. how long they were awake during the night – suggestions for future work to use shortened, digital diaries.

Next Steps

Future work is looking to implement radar sensing continuously over longer durations (6 – 18 months) and in other populations. Other future plans include validating the radar sensor against the gold-standard sleep measurement method, polysomnography, in adolescents. It will be important to incorporate radar sensing for sleep measurement alongside other low burden health measurements to explore changes and patterns in sleep timing and health outcomes. Check-out some ongoing studies integrating the use of longitudinal radar sensing:

AMBIENT-BD: https://www.ambientbd.com/
Metabolic Hub for Psychiatry (MetPsy): https://www.metabolicpsychiatryhub.com/takepart
Mental Health in the Moment (MHIM): https://www.mhim-study.ppls.ed.ac.uk/

Please reach out if you would like to discuss in further detail novel radar-based methods of sleep assessment or the above studies and co-production: lcaddick@ed.ac.uk.

Blog by Lorna Caddick