Category Archives: Wearables

Muse Indirectly Crushes Meditation Training

Looking at my morning tracking routine I realized that one of the most impactful wearable devices I have used has been the Muse EEG headband. After using it for more than five months, I think that Muse indirectly crushes meditation training.

I started using the Muse EEG headband in June of this year and have sat with it 152 sessions. The data the product has given me has not been the source of value. The source of value has been that the product has helped me become a habitual meditator.

Muse indirectly crushes meditation training

Muse claims to read your EEG and give you a “calm” score. It also awards a secondary score that is cumulative as a game style mechanic to keep you coming back.

When I started with Muse I took a straightforward sporting approach to the mediation training.  I would practice each morning with a goal to get higher “calm” scores. I saw it similar to training for a 5K where I would look to my speed. The fun would be to see how much higher I could push my calm score with practice.

Unfortunately, it did not work out that way. I did not improve the score even with months of  practice.

Muse indirectly crushes meditation trainingThe calm score itself is of questionable usefulness. It did not correlate with any other physiological factor that I compared it with. For example, when looking at the correlation between the score and heart rate variability (HRV) across 145 readings, there was no relationship (Pearson r = -.016). I had taken HRV readings simultaneous with the Muse readings and they did not track together at all. Blood glucose levels (r= -.17), average resting heart rate (r = .13), blood pressure (r=.21) all were at best a weak relationship.

Thinking that the calm score might somehow be associated with how distracted or stressed I was, I looked at a morning mood score I had been keeping versus the calm score. Oddly, I found a moderate inverse relationship between the Muse calm and my perception of mood (r = -.32). That meant I was more “calm” when I was in a lousier mood that morning. That made no sense at all.

So my original idea of practicing to increase my calm score did not pan out for me.  So why do I believe Muse indirectly crushes meditation training? Because for me, meditation had been boring and numerous attempts in the last 32 years to incorporate it in my daily routine had failed miserably. Chasing the Muse score in a structured way each morning I broke the boredom and acquired the habit of meditating. And meditating has scientifically validated positive benefits.

The Muse basic session is six minutes long with a starting calibration of one minute then a five minute reading. After multiple months of starting doing these simple six minute readings with the headband I found I had started comfortably expanding the amount of time I was sitting quietly.

First, I incorporated a ten-minute session before the Muse session to test the effectiveness of binaural beats, and that ended up with me sitting quietly for fifteen minutes each morning. Then a podcast on HRV inspired me to add twenty minutes of paced breathing later in the day. I was able to expand because I had gotten comfortable with sitting during the initial months of short six minute Muse sessions.

So the paradoxical outcome presents itself. In the past when I had tried to “learn” to meditate I could not do it for long and was unsuccessful. When I introduced the game of chasing the Muse calm score I was able to get enough time sitting quietly to find meditation doable and even pleasurable. And when I had enough data to determine the score I had been chasing was meaningless I had worked my way up to 35 minutes of sitting a day. That is why Muse indirectly crushes meditation training.

I imagine the engineers who created the scoring system and the EEG technology may not appreciate my assessment. It seems better to have the scoring and EEG technology to be a valued feature. However, the product bills itself as a meditation assistant. In that, it performs its job perfectly.

To someone looking for a Quantified Self product review on Muse the answer may sound like something out of a wearables zen koan. To realize the value of the Muse product, diligently try to improve your Muse score until you are sitting in comfortable in daily meditation realizing that the score never was the point.

 

 

Sign up for the QuantXLaFont Newsletter
Get our lifestyle tips and studies delivered to your inbox.
Thank you! We don't spam :)

Why Apple Watch Will Not Be Accurate on Stress Readings

With the imminent launch of the Apple Watch it is clear that everyone is looking to it for new functionality and inevitably there will be claims you can reduce stress using it. I wanted to look into whether the device could reliably deliver on that promise.

Apple Watch Will Not Be Accurate

The watch reportedly can read heart rate by taking pulse readings from the wrist using pulse oximetry, a method that uses light pulses to read heart beats by measuring the change in skin color due to different levels of blood flow. Pulse oximetry is refined enough for reading heart rate, but Heart Rate Variability (HRV) demands precision that pulse oximetry reportedly cannot deliver.

I am interested in using HRV to improve personal performance in working sessions, face to face meetings, negotiations and public speaking. I thought it would be interesting to be able to use an Apple Watch to read HRV and improve those skills. So I wanted to test pulse oximetry myself.
Apple Watch Will Not Be Accurate

I don’t have an Apple Watch, so I connected a Mio Velo wrist worn band to the SweetbeatLife app on an iPhone. The Mio product claims to deliver “EKG-accurate heart rate data” and uses pulse oximetry, so this would be my proxy for an Apple Watch.

Apple Watch Will Not Be Accurate

To compare this setup with readings from electrical signals I connected a Polar H7 to Marco Altini’s Heart Rate Variability Logger on an iPod Touch. This would allow me to take two readings of a single heart beat and compare the methods. After wrangling settings and conflicting signalling I got them both to work.

My goal was to use my Parasympathetic Flatline method when comparing the pulse oximetry with electrical readings using a heart rate belt. This means I am looking for 10 consecutive heart beat intervals that vary less than 17 milliseconds from beat to beat. When I find these strings of beats I am measuring myself in a fight/flight state.

Researching pulse oximetry I found a research paper that said that physical movement introduced errors in readings making pulse oximetry unreliable for measuring HRV when subjects were in motion. The conclusion was that pulse oximetry “is unlikely to prove a practical alternative to the ECG in ambulatory recordings or recordings made during other activities.”

Apple Watch Will Not Be Accurate

My interest is looking at activities like negotiating, meeting and coding where there is not a lot of physical movement. With the study providing some potential for pulse oximetry to provide some value to my area of interest, it seemed reasonable that readings when relaxed would be similar and when walking very different.

I conducted sessions in a relaxed state, working by myself on the computer, in meetings and while walking. I first conducted the dual measurements while in a relaxed state for ten minutes. I sat and did not move and breathed in an even rhythm. Subjectively I think I was in fight/flight for 25% of the time because sitting motionless allowed me to think about all the things I was not getting done. Here is a graph of the two readings:

Apple Watch Will Not Be Accurate

The pulse oximetry reading was that I was in fight/flight 87% of the time and that is way overstated. The P7 said 32% and that was much closer to my experience. So the relaxed state had a completely different outcome than my hypothesis.

Next I measured myself when I was in a working session, which meant I had structured some time to work on my computer without interruption. I was working on some recruiting matters which meant screening resumes. It was very focused work and I felt relaxed. I would have said I was 10% at most in fight/flight. My session was 16 minutes long, and here are the charts:

Apple Watch Will Not Be Accurate

The pulse oximetry reading said I was in fight/flight 57% of the time. This did not remotely match my experience. The H7 reading said I experienced no fight/flight at all. There were accelerations, but none that were more than 9 beats. So though I’m not sure it was a perfect session it was clear that the H7 more closely matched my experience.

I also took readings during the first and second half of a long staff meeting. I was not the host, I was a participant. There were some controversial things being discussed so I would have subjectively said I was in fight/flight 15% of the time. Here is the chart for the first half of the meeting:

Apple Watch Will Not Be Accurate

You can see the pulse oximetry said I was at 62% fight/flight, H7 10%. Here is the chart for the second half:

Apple Watch Will Not Be Accurate

Pulse Oximetry had me at 63% and H7 at 1%. The H7 seemed low because there were a few moments where I was definitely in a heightened state, but an average under 10% is much close to the perceived 15% than a consistent reading by the Apple Watch equivalent of over 60%. That just made no sense.

I took measurements while walking. I had low expectations because I had taken readings when exercising and know that HRV is low when physically active. Here is the chart as I took my first walk to the train from work.

Apple Watch Will Not Be Accurate

I walk briskly so I expected a 60% to 75% reading here. What you see is Oximetry at 91% and H7 at 67%. Again oximetry was high. Here is my reading for leaving the train and going to the pickup point:

Apple Watch Will Not Be Accurate

What is interesting here is my wife picked me up about a third of the way through the reading and I relaxed in the car chatting with her as she drove me home. The H7 clearly shows me moving from an accelerated state to more relaxed, which was my experience. The oximetry reading shows continued stress. Again the H7 reading matched the experience.

So what conclusion can we draw from comparing pulse oximetry as used by Apple Watch to electrical readings from chest worn heart rate belts? In the range of activity from sitting motionless to walking briskly the pulse oximetry method overstates stress.

So your Apple Watch is not the best tool to measure stress response using HRV. When you read, “Physicians and digital health experts are encouraged by the level of accuracy suggested by the Apple Watch’s sensors,” remember that pulse oximetry will overstate your stress. In a world of stress the last thing you need is to have it overstated.

My Quantified Self Lessons Learned in 2014

I started the year wanting to explore how I could use technology to understand when stress was or was not occurring. I was interested in if self reported stress was reliable and if there were techniques I could practice that would reverse stress in increasingly shorter periods.

Looking for technology that could help identify when I was stressed was an exercise in buying a lot of technology and trying to find anything that would actually work. I looked at galvanic skin response, different watch iterations and ended up settling on heart rate variability (HRV) as a way to understand when I was relaxed or stressed.

As I began looking at different states of stress using HRV I measured myself while meditating, getting a tooth drilled at dentist, while giving a public speech, and getting a haircut. Each of these gave me the range of when I was stressed and not and gave me a baseline for further studies. I think the takeaway here is the boring baseline building work is necessary for real insight.

I learned that when it came to returning to poise from a state of upset, I could improve with practice and that a key technique was respiration. The ability breath well, which takes a bit of practice, was the key to busting stress. So stress, like fitness, was a state that could be altered with progressive practice. That was my assumption at the beginning of the work.

What was less obvious was how much thought and belief plays a part in how much stress I experience. Early on during my self reporting studies I found that a surprisingly high percentage of stress was self induced. Most stress was due to a discrepancy between what I thought was proper and what what happening. Even deeper, I found that my reactions were not complex reactions, but that emotion is navigation. Whether I was feeling in the right location or out of place determined whether I was calm or stressed.

I thought I could use technology to measure stress then solve for it through techniques, but that model turned out to be incorrect. It turns out my thoughts drove a stream of stressful reactions (or not) and that knowing when I am in a state of stress or not helped me change the underlying construct. And that is what takes me into the new year.

My Quantified Self Gear 2014

I have steered clear of reviewing products because I think simply buying products has very little to do with Quantified Self. And I thought it good for me to review what I used and how useful some of it was. My premise for my QS work in 2014 was to use technology to train myself to be happier.  I had used several Garmin products to successfully train for a half Ironman. Why couldn’t I train myself to be happier?

I pulled everything out of my wearables storage drawer and took this photo of everything I bought in 2014:

QSGear

The items:

I started with a Pebble smartwatch that my wife had given to me as a birthday gift. $99 from the original Kickstarter campaign. I love it and still use it daily with one app called Motiv8 that tracks activity.

Google Glass. What can I say. I fancied myself as an Explorer with $1500 burning a hole in my pocket. I once looked up the population of the state of New Jersey on it and sent my son an email saying “Hi this is my talking to my Google Glass.” That about sums it up. It has not been charged up for about 8 months now. It was so deep in the drawer it did not make the picture above and I just now remembered having it. Enough said.

Zensorium’s Tinke. Billing itself as a stress and fitness measurement device, I purchased one at the Quantified Self Europe conference in Amsterdam for over $100. Its readings made no sense to me and it went into the drawer pretty quickly.

Heartmath’s emWave2 & emWave Pro. This was over $400 worth of gear and if you follow this blog or my QS speeches at all I did get a lot of use out of both products. I conducted multiple experiments and accrued 183,843 “coherence points” – which is quite a few hours of cardiac coherence. In the end I grew out of it as coherence was not my ultimate goal. I think this product is way overpriced and was useful.

Neurosky Mindwave & Mindwave Mobile. Over $200 in cost, I could never get either headset to work consistently. I took some readings but any attempt to get the devices to reliably produce output was frustrated by bluetooth connectivity issues  of some sort. A big disappointment from Neurosky.

Emfit sleep monitor. I met the Emfit team at the QS EU conference and they helpfully offered me a free trial of their product. A combination of wireless connectivity issues and my move from London to San Francisco resulted in my never getting it working.

Mio heart rate band. Very slick implementation and a comfortable wrist band that uses pulse oximetry. I loved the idea, and it was not useful for heart rate variability experiments. The accuracy was not good enough so into the drawer it went. I paid over $100 for it.

After visiting with a friend who worked at Basis I dutifully bought the first version of the watch for around $150. I liked a lot of the ideas but did not really take to the interface or the gamification element of the online account. By the time I bought it I had eliminated pulse oximetry as reliable source of heart rate data. I gave it to a friend and he likes it.

Fitbit flex. I ended up buying two for $99 each because the first one gave out and stopped charging. The second one was spotty on charging as well. I used the product for 10 months and got a lot of value from it. In the end, the inability to charge it and a policy change that eliminated active minutes as a goal had me put it in the drawer. I replaced it with the $79 Garmin Vivofit because I do like to monitor my daily activity. So far that seems to be working out.

Sweetbeatlife & the VitalConnect Patch. Sweetbeat Life is an app that takes heart rate data from either a belt or the VitalConnect Patch. The patch seemed novel as it was convenient and comfortable. And it did not stay adhered on my chest for more than a few sessions. It was a breathtaking $199 for a set of 10 patches. I did not understand the real cost until the first patch fell off after the second use. Really cool and really expensive. I went back to the old reliable Polar H7 heart rate belt for a nice price of $80.  And one belt will last the whole year.

One thing that is not clearly stated is that you need top end smartphones to use apps associate with all this hardware. Neurosky, Fitbit, VitalConnect Patch and even my much loved Pebble need a phone with Bluetooth LE. I had an older version Android phone without Bluetooth LE so I needed to buy an iPod5 for iOS only apps and devices $199. And for Android I had to buy another device with LE so I bought a Nexus 7 tablet for $245.

So a quick add up gives me approximately $3,500 worth of gear of which 42% of that is the Google Glass. What did that expenditure do for me? It taught me through brute force that picking an area of Quantified Self to study and focussing there is 90% wikipedia work and networking with other people who have knowledge. 10% is hardware. And ultimately the majority of value came from about $500 worth of the gear I bought (Heartmath Pro, Polar H7, iPod5). The rest helped me understand some things but were not good value for money. For the Quantified Self, as in life, money cannot buy you happiness.

V1bes on Indiegogo

I met the founder of V1bes, Gustaf Krank, at a wearables gathering in Helsinki last year. He gave a dynamic presentation to the conference and afterward walked me through the technology with a personal demo. The approach is like no other in that is aims to pull together multiple electromagnetic signals from brain, heart and the environment through a ring.

V1bes has launched an idiegogo campaign. I am going to to get one to see how its measurements correlate with Heart Rate Variability (HRV). The idea of electromagnetic “smog” as an influencer of HRV is something worth looking at. Unlike HRV Gustaf’s invention does not have the large number of medical studies with which to compare but that is part of the fun.

Measuring When With Others

Pretty big design problem to solve is how to collect real time information when sitting with others. When by myself I can record my thoughts by speaking them or logging them when I see the device flashing red. As I described in my post on the respiration study I had a pretty clear system to recover when I was flashed into the Upset state. Yesterday I described how I want to take this work to understanding the Upset vs Poised state when in communication with others. Hard go.

I measured two interactions yesterday, one using Heartmath and the the other using SweetbeatLife. When I did the first interaction I was able to record it as it was by phone, but I was Freakbacking like crazy. I was on a call with the bank and watching the Heartmath monitor and as it flashed red I was trying to correct for it which made the situation even worse. As I was trying to talk to the woman on the phone the reading went haywire and I ended up actually going into a very high state of stress. It was a case of way too many inputs. When I listened to the recording I could hear the stress in my voice. Freakback central!

The second interaction was with a good friend over coffee who is sympathetic to the cause (and who is going to read this) and I just had the SweetbeatLife on monitoring my heart rate and HRV. We talked about a variety of things to include drones, dystopias, quantified self and monitoring oneself when talking to others. At the end I showed him that I was monitoring my own HRV during the discussion and he appropriately asked if I had recorded him, which I hadn’t. And what I got was a contextless HRV and stress line that in no way was helpful because I did not know what was being said or what I was thinking as the line moved during the discussion.

So in both measured interactions where i was looking for ways to be more engaged with people I either reduced engagement by Freakbacking or got a measurement that really was not useful even in review because I could not tag it with what was happening in that moment. Somehow I need to hack together a way to capture thoughts unobtrusively while not violating the privacy of others and transparency in connecting because somehow them knowing they are being recorded or my knowing that I am not being transparent in the measurement would cloud the connection.