Circadian rhythm dysfunction and mood dysfunction are often interrelated. For this reason, we collect actigraphy data in many of our studies. These data describe a person’s activity throughout the day. They allow us to see when people are most and least active and when they sleep. Fitness trackers such as Fitbit and Apple Watch collect this type of data.
The great thing about actigraphy data is that it can be collected continuously using lightweight monitors, so there is a lot of information available. The challenging aspect is that the data can be noisy - and there is a lot of it!
Because we want to use only the cleanest and most reliable data in our studies, we examine the data from every day the actigraphy monitor was worn. This allows us to assess whether the identified sleep and wake times are reasonable or whether the data may be unreliable, perhaps because the person removed the monitor.
To help with this process, we use two automated methods to estimate sleep and wake times, as well as the most and least active periods of the day. We have created a custom interface, shown below, that allows users to select between the two methods or label the data as unreliable.
We are looking for volunteers to help us perform this quality control. Volunteers who contribute significantly to this work will be acknowledged in publications that use these data. Volunteers may also have opportunities to use the data for their own publications.
The QC system is located here. (You will need an account to log in.) All of the instructions are provided here.
This system helps reviewers compare two ways of measuring daily activity and sleep. Reviewers look at the activity patterns, choose the result that best matches the data, and add notes when something looks unusual.
These graphs show activity and light across the full recording. They help reviewers spot missing data, schedule changes, and days that look different from the usual pattern. (The data shown after the * is collected when the person was in lab, so they moved less and experienced less light and sleep.)
Activity data is also shown in the form of an actogram where activity levels is represented by color (blue to yellow is low to high). Sleep periods are easy to detect (blue patch at the left of the image) and the sleep detection from the two methods can be compared.
Each plot shows one day of activity and the following morning. Reviewers compare the highlighted L5 (blue, lowest levels of activity for 5 hours), M5 (orange, highest levels of activity for 5 hours) and sleep (purple) periods and select the option that best matches the activity pattern. The outlined and non-outlined boxes are from two different methods. Here they match well.