I test the sleep cycle app on the apple watch against this small sleep, eeg device that’s being used in several research projects. I work both of these for six nights and i will directly compare their results. As always, i do not want to waste your time, so timestamps are in the description below and also on the timeline Music. For those of you who are not familiar with the app called sleep cycle, it was originally developed as an app to be run on an iphone or android device. If you use it on a smartphone, it will use your movements as measured on your smartphone, to detect your sleep stages. However, since may of 2020, sleep cycle has also been available for the apple watch. I tried to find out which sensors the app uses to monitor your sleep, but i could not find out for sure the app does show your heart rate in the morning when you track your sleep but i’m, not sure if heart rate or heart rate variability are Actively used in the algorithm, i suspect, it’s likely just motion, since this will make it easier for them to adapt our original algorithm. But this is pure speculation. The sleep cycle app does not track the classical sleep stages which are rem, sleep, light sleep, deep sleep and awake. Instead, it tracks the depth of your sleep in the form of a graph going from awake at the top to a sleep in the middle and to deep sleep at the bottom.
This makes some sense as the main goals that sleep cycle lists on their website are one to better understand your sleep patterns and two to have your alarm go off the moment. You’Re in light sleep. This would make it easier for you to wake up and make you less groggy. Now let’s get to the test. I will compare the results of the sleep cycle app on apple watch to the sleep stages as i track them using this small scientific eeg device. At the same time, i also recorded myself using an infrared camera, the eeg device measures, brain waves and muscle movements, it’s called the hypnoline z max and is being used by several of my colleagues in scientific studies. If you’re interested in this device, i will link it below now. I manually went through the recording of each of the eegs and scored each part of the night for the different sleep stages. Now analyzing the data from the sleep cycle app was actually not completely straightforward. I could not export the full graph you see in the app in a usable format. I ended up using a really handy online tool called web plot digitizer to extract the raw data. Basically, i had to take screenshots of the graphs and using the tool. I was able to extract the raw data. Finally, with the infrared recording, i can actually check what my movements were like during the night and see if the sleep cycle app correctly predicted the moments that i was awake.
Let’S first have a look at the six individual nights where i compare the graphs from the sleep cycle app to the different sleep stages. I went through each night here we have the first night where, on top, i plotted my sleep stages with the time of night. On the horizontal axis and the line indicating what sleep stage i was in so i started the night being awake, then i had some light sleep, some deep sleep, a little bit of lightly began and then went into rem, sleep and so on and on the bottom. I plodded my knight, according to the sleep cycle app now, supposedly the lower the value in the sleep cycle, app the deeper the sleep so let’s first have a look at deep sleep that’s. What i’ve highlighted in purple here and as you can see in the moments where i had deeper sleep? Indeed, the scores according to the sleep cycle app are lower. However, we can also see that at later points in the night, where i no longer had any deep sleep, i still had lower scores, as you can see here here and here, i’ll get to that in a second. But let’s first have a look at rem, sleep and that’s. What i’ve marked here in red? What you can see is that the moments where i was in rem sleep this night. The app actually gave me a higher score. What you can also see is that it appears at the moment where i was not in rem.
Sleep sleep cycle actually gave me a lower score that’s. What i plotted here in blue green, you see the moments where i was not in rem, sleep and in red. The moments where i was in rem, sleep and, as you can see, there appears to be a peak in the sleep cycle score when i’m in rem and a dip in the score when i’m in non rem. This is actually very interesting, since this is how sleep cycles are defined. You start each cycle in a combination of light and deep sleep together called non rem, sleep and you end each cycle in rem, sleep as you can see. I went through six sleep cycles, each starting with non rem in blue, green and ending with rem in red. The first night makes it seem as though sleep cycle can indeed do what the name suggests. That is pick up on the different sleep cycles. Let’S see if that holds up for the other nights as well. Here i plotted the second night. Let’S first have a look at deep sleep again now again we see that in those moments of deep sleep, the scores tend to be a bit lower. If we look at rem sleep, on the other hand, we see that similar to the first knight scores appear to be higher and if we divide the knight into non ram in blue, green and ram in red, we see a similar pattern to the first knight low Or decreasing scores in non ram and higher or increasing scores in rem? If we look at the third knight, we see something similar again lower scores in deep sleep, but these low scores are not exclusive to deep sleep.
We again see the same patterns for ram and non rem sleep. Interestingly, however, in my second sleep cycle, i had some interrupted sleep. As you can see, i woke up a few times in the non ramp part of the sleep cycle, and this also matches with the sleep score of the sleep cycle, app where my score did not decrease as much as it did for the other non ramp. Parts of the other sleep cycles – as you can see here here and here so it appears that the sleep cycle app is able to pick up on both rem, sleep and the moments that i was awake and it gives me a relatively high score for both here. I mark the awake moments in green and you see that the score did not decrease as much now. At the end, here they were intertwined with rem, sleep, so it’s difficult to say what was the main cause now let’s quickly go through the last few nights, after which i will do an overview analysis here we have another night and similar to before. We see increased values with rem, sleep and decreased values with non rem, sleep for the second to last night, i’m going to show this is also true, though here you can also see a strongly increased peak a moment that i was awake. Finally, for the last night here we generally see the same patterns. Only these two peaks here and here do not fully align with any rem, sleep or awake periods.
Now, let’s look at a bit more of a structural overview analysis, and that is what you see here. What i did is i extracted the scores that were associated with each sleep stage and plotted them here on the horizontal axis. Here we have the different sleep stages, so deep sleep, light sleep, rem, sleep and awake and on the vertical axis we have to score. According to the sleep cycle app now, each dot here is a single moment of measurement, and these boxes here indicate the rough ranges of these points to make it simpler, let’s remove the individual points from the visualization. What you can see here is that deep sleep indeed gives the lowest sleep cycle scores followed by light sleep now. Both of these are considered non rem, sleep which matches with what we saw before in the individual nights. Next rem, sleep gives, on average, higher scores and finally, awake gives the highest scores. What i would conclude from this analysis is that the sleep cycle app can detect your sleep cycles and that if you have a low score, this means that you’re likely in deep sleep or a light sleep, and that if you have a higher score, you’re more likely In rem, sleep or awake, however, what does this mean with regards to the goals of the sleep cycle app? So it seems that, true to its name, the sleep cycle app can detect your sleep cycles, which is kind of cool.
It also seems to detect those moments that you’re awake, so how does this relate to the two goals that they list on their website? Well, the first goal is to understand your sleep patterns, and this seems to be possible with the data they show. You can actually see your sleep cycles and also those moments where your sleep was interrupted. Currently, the app also calculates a sleep score and i would say, based on the data that they have, they should be able to calculate a meaningful score, though i have no idea how they actually calculate it. That brings us to the second goal of the sleep cycle. App – and this is to have your alarm go off during light sleep now, based on what we saw, it seems to be less likely that this is done correctly. I think they assume that a high value in their graph means light sleep. However, based on my tests, i would say this means either rem, sleep or awake, which would mean that often, when they have your alarm go off, you are actually in rem, sleep and not in light sleep. However, i could be wrong in my assumptions and maybe they use their graphs in a different way to wake you up during light sleep. Overall, though, i am impressed with the fact that the sleep cycle app was able to pick up on my sleep cycles now. I also collected data for other apple watch apps, namely the pillow app, the autosleep app and the sleepwatch app and i’ll be releasing videos about these apps in the next two months.
It will be interesting to see how these compared to the sleep cycle app. I should mention some of the limitations of the data that i showed here. First of all, i just tested it on me and only for six nights. Where most of these nights i had pretty decent sleep. It would be interesting to see how it performs on other people and also on nights with worse sleep. Second, to do a full sleep comparison. It would be good to also test the apple watch, apps against a full scientific polysomnography setup. I actually plan to build my own polysonography device with open vci components in the first half of next year. That way, i will not have to rely on sleep labs for my testing, which is especially difficult in these times of corona. Finally, i’m, not a professional sleep stage core on ecgs. I think i did a decent job, but for some parts of the night i might have been a little bit off in my videos. I do scientific tests on different devices like the aura ring the fitbit and the scan watch and, in the end, i hope to use tracking to improve my life. So if you like that, subject and like this video consider subscribing to my channel and also consider giving it a thumbs up, because it makes it easier for other people to find my videos.fjPng54_CtE
