Sleep 2.1 Intro to Sleep Actigraphy

Sleep actigraphy signal processing: transforming raw acceleration data into sleep metrics, its limitations, and the future of wearable sleep research.

Video Breakdown:

00:00 - Introduction & Why This Matters

03:00 - Intro to Sleep Actigraphy

03:42 - Clinical Applications & Research Insights

05:47 - Challenges with Existing Algorithms

07:03 - Labfront & Garmin’s Approach to Enhanced Actigraphy

Today we'll be diving deep into advanced sleep actigraphy signal processing. And this is an aspect of sleep monitoring that has often remained sort of behind the scenes.
And while many of you, I'm sure, are familiar with sleep actigraphy, we rarely discuss what happens under the hood, like how raw acceleration signals from our devices transform into clinically relevant sleep metrics. In the next 50 minutes or so, I'll unpack this black box of signal processing, and to warn you in advance, the talk will be quite technical, but this was very deliberative on my part.
It is because sleep actigraphy really has not advanced over the past 30 years, despite advances in hardware and signal processing methods, and I feel that only until we understand our existing methodologies and recognize their limitations and assumptions that I feel that we could truly explore, um, sort of advance the field forward.
And so an extension of this in this talk, we'll also explore some untapped research opportunities in this field and demonstrate how Labfront's capabilities can support you in pushing the boundaries of sleep research.
and before we delve into this topic, let me briefly situate this talk in our broader webinar series. This series spans the full spectrum of sleep science, from the fundamentals of actigraphy, um, to PPG signals, to the complexity of circadian rhythms and sleep disorders, all from a wearable perspective.
And note, each of these subcategories will be explored separately in a dedicated webinar session. and we've designed this series with three core objectives in mind. First, we want to deepen your understanding of the science, particularly through the lens of wearable technology. Second, we'll demonstrate how Labfront's features can be leveraged to support your research endeavors.
And finally, and this is really important, is that we want to hear from you. I think your feedback is invaluable in helping us refine our functionalities and to ensure that we're providing the tools that you need in order to advance your research. So as mentioned, today's talk will be on sensor signal processing, some discussions about algorithms for sleep timing and duration, as well.
And, the next talk will be on, PPGs and biophysics and physiological considerations. And some of the talks, some of the questions that I'll be answering in the upcoming talk, will be, to answer some of these questions, like why PPG with green light is less susceptible to motion artifact compared to red light PPG? What limitations exist for reflective PPG in assessing oxygen saturation? As you may or may not know, our wearable devices use reflective PPG and how does data differ between ring versus wrist devices, and why skin tone, does not typically change the heart rate and heart variability quality, but can diminish quality of oxygen saturations.
So this is the outline for my talk today, and I'm going to first start with a quick intro to sleep actigraphy, and this is for individuals, you know, this, for those individuals who may not be up to date on actigraphy overall. So, quickly, actigraphy involves wearing a watch-like device called an actigraph or a wrist on a wrist or the ankle.
This device contains an accelerometer to monitor motion. The acceleration data is processed and converted by various algorithms into sleep/wake algorithms, classifications, and these dots indicate times of being awake. And this classification is then used to produce sleep quality parameters such as total sleep time, sleep efficiency, et cetera.
And so in 2018, the American Academy of Sleep Medicine released guidelines on the clinical uses for actigraphy, and their comprehensive literature review revealed that there's strong evidence supporting actigraphy's use in insomnia, periodic limb movement, sleep apnea, and circadian rhythm, disorders.
And now, you know, we all can acknowledge that PSGs or polysomnography is our gold standard for sleep evaluation, but its complexity with all those wires and monitors really makes it impractical for long-term monitoring, um, at home particularly. And this is where actigraphy shines. It offers a practical alternative that allows us to track over, uh, sleep patterns over weeks or even months, and make it particularly valuable, to monitor how patients respond to interventions.
And building on this evidence base, the AASM took a significant step forward last year in October of 2023, where their Act on Actigraphy campaign. Their goal is to get payers to reimburse Actigraphy as a service and to move away from its categorization as an experimental diagnostic procedure. And this initiative reflects the AASM position that the scientific evidence has become robust enough to justify broader clinical implementation.
And they also recognize that limited access to actigraphy is creating a real barrier for individuals who could benefit from this valuable sleep tool. Furthermore, you know, the significance of actigraphy has been further highlighted By recent research linking sleep regularity to mortality risk.
This was a publication done by Windred et al. and published in Sleep in 2024. This actually analyzed the UK Biobank data. And apparently sleep regularity was a stronger predictor of mortality risk than sleep duration. And it's important to note that, in order to assess sleep regularity, actigraphy is the primary tool, that's being used for that purpose.
And so, despite actigraphy's growing recognition and clinical importance, there's a surprising stagnation in its technological foundation. I would say in the 30 to 40 years since its introduction, the underlying algorithm have seen remarkably little advancement. And probably for a number of reasons, but first, you know, many of these algorithms remain black boxes protected by proprietary intellectual properties.
We actually did some preliminary evaluation ourselves and found it very difficult to find data regarding what the processing methods that these companies were using. Secondly, and perhaps more concerning, is the poor concordance between these different devices. And you can see the sleep/wake classifications results from our internal testing.
Each row represents the same night for the same individuals recorded simultaneously by an AMI Motionlogger and a Philips ActiWatch 2. And you can see that there are differences, significant differences, between the two outputs. for these two devices. And these significant differences probably raises questions for us about the reliability and standardization of actigraphy in general as a research tool.
And this is precisely why Labfront partnered with Garmin to take a different approach. And rather than offering another black box solution, we recognize that how the data is really a matter significantly. how the data is processed, I should say. And that researchers really need flexibility in this process.
So Garmin Enhanced Actigraphy puts the control back into your hands and allows you to configure Actigraphy settings, which includes, items such as window, epoch window size, the type of filter that's applied, the filter path, which I'll talk about in more detail later. The process data, which includes error crossing mode, energy and time above threshold, threshold settings, etc.
And our hope is that this flexibility and transparency in data processing will fundamentally shift how activity, actigraphy research is conducted. And furthermore, Garmin enables you to apply three different actigraphy settings simultaneously, allowing you to adjust window size, Filter order, filter, path, et cetera.
And you could basically get a number of outputs simultaneously so that you could really, um, sort of use the data appropriately for your research study.

Course episodes

Sleep 2.1 Intro to Sleep Actigraphy

Sleep 2.2 Traditional Signal Processing: Sleep-Wake Classification

Sleep 2.3 Piezoelectric vs Capacitive Accelerometers

Sleep 2.4 Signal Processing Configurability: Parameter Settings

Sleep 2.5 Unexplored Research Opportunities