Opinions/Experience re the ZEO?

Nate, Here is an image I had posted in Jay's sleep lab thread. I took the mask off as you can see. The light green is REM, dark green=deep, grey =light, orange=wake. The O2 drops in REM are consistant with my PSG results.
The Blue graph is based on 30 second data points. The website graph is 5 minute data points. The 30 second graph just gives you a better idea how disrupted the sleep stages are.

Lizistired wrote: The Blue graph is based on 30 second data points. The website graph is 5 minute data points. The 30 second graph just gives you a better idea how disrupted the sleep stages are.

What software did you use to draw the blue graph with the 30 second data points?

Any studies of correlations between Zeo and EEG, PSG?

Only for those interested


The Zeo is mentioned here from pg 10

http://robertdickerson.net/papers/phdproposal.pdf

System for the Collection, Storage, Analysis, and Reporting
of Objective Behavioral Measures

Ph.D. Dissertation Proposal
Robert F. Dickerson

Timeline: The last three months of the summer 2012 will be used for thesis writing, with
an intended graduation date of August 2012.

rfdickerson@cs.virginia.edu
Department of Computer Science
University of Virginia
Charlottesville, VA, USA


The following is from Robert Dickerson's paper:


2.4.1 State-of-the-Art


SNAPPy mote

Three tri-axis accelerometers are taped onto the mattress and connected
to a wireless Synapse mote.

Self-report questionnaires such as the Pittsburgh Sleep
Quality Index (PSQI) [4], are answered by patients
every two weeks, however, studies have shown that
subjects with insomnia are not accurate in their subjective
report of variables such as sleep latency, sleep
duration and number of disturbances and overall tend
to underestimate their ability to sleep. There is a
need for objective instruments that can measure sleep
quality where the subjective ratings fail. Because of
its importance, many sleep-monitoring systems have
been developed. These systems attempt to recognize
sleeping disorders by providing healthcare providers
with quantitative data about irregularity in sleeping
periods and durations or the amount of agitation and
restlessness experienced during the night. These solutions
vary in cost, comfort, and accuracy. However
to date, there are few low-cost and unobtrusive sleep
monitoring systems.

The most accurate are polysomnography devices, but there are major drawbacks to using
them since they need to be worn, and require professional monitoring and thus expensive to use. The actigraph
is an accelerometer device that can be attached to any of the limbs (e.g. wrist) to provide data on movement,
however they still need to be worn. Pressure pads such as the bladder can be used, but they are uncomfortable
and moderately expensive. The Zeo is a headband that measures electrical signals on the scalp to estimate the
stage in sleep, however they need to be worn and their accuracy has not been evaluated.

2.4.2 Approach

Solutions for providing sleep quality are still invasive or costly. The WISP receiver costs over $600 each. We
built a custom solution using a Synapse SNAPpy RF motes for wirelessly transmitting data. We attached three
independent tri-axis accelerometers to the mote and they are sampled at 1 Hz as shown in Figure 5. Data
is processed on the client PC which determines the amount of deflection since the last sample the side of the
mattress when a person’s weight is applied. Since the accelerometers indicate the direction of the force due to
gravity as a vector, we take the dot product of the last sampled vector and new to determine the amount of
deflection since the last sampling. If that deflection exceeds a threshold, a movement event has occurred. The
advantage of this approach is that the true orientation of the accelerometer does not have to be established to
measure movements, which allows us to do detection without calibration, and it allows continued operation even
when the sensors may have been knocked out of place, also since we only store events and not raw measurements,
we eliminate noise and lower the storage requirements and increased scalability.

Comparison with Zeo

The Zeo device has been compared against polysomnography in several clinical studies (Wright 2008, Fabregas 2009, Shambroom 2009). The Zeo, via a headband, monitors electrical signals around the head to infer the sleep stage of the wearer. Zeo’s SoftWave algorithm uses a neural network to classify each 30 seconds into the sleep stages ‘wake’, ‘REM’, ‘light sleep’ (stages 1-2), and ‘deep sleep’ (stages 3-4).

We collected preliminary data to explore the relationships between the data collected by the Zeo and the WISP tags. For this experiment, we used a single sleeper on a twin-sized mattress. For seven nights, the subject wore both the Zeo headband and slept on a WISP instrumented bed. Eight WISP tags were used on the bed. The Zeo reported the sleep stage prediction results every 30 seconds, so correspondingly, we used a 30 second sliding time window on the WISP data and the movement levels on each accelerometer to extract the variation of the time signal. After the signal has been Z-normalized, values over a threshold of 2-sigma were classified as a ‘movement event’. The result was binary data recognizing periods of movement in the sleep as shown in Figure 20.


Figure 20 – One night’s sleep period comparing the movement data with the Zeo

Next we use a Naïve Bayesian classifier to classify the sleep stages. 15% of the data was used for testing, and the other 85% was used for training. We received poor classification accuracy especially with the REM, light sleep, and deep sleep. Next we trained a dynamic neural network using 70% as training data, and 20% for testing. The performance of the regression R values was 0.28 during testing with a mean squared error (MSE) of 0.64. In this very preliminary study with a single subject, there seems to be little correlation between the four sleep stages Zeo predicted and the features from the WISP tags.

Next, we limited the scope from predicting four sleep stages, to just two, sleeping vs. wake. Our dataset had 4285 frames for all the sleeping periods and 189 frames of wake periods. Our hypothesis was that wake periods receive more movement than sleep stages. We performed a one-tailed t-test on the two distributions assuming unequal variances. The null hypothesis was rejected with a p of 0.0017 that wake stages receive more movement than sleep stages with a mean of 1.59 vs. -0.07 standardized movements per frame. When the person first enters the bed and just prior to leaving the bed there is large period of movement. These preliminary results show that there could be a relationship with movement and consciousness. However, accurate segmentation of the time regions where someone drifted into sleep from being awake is a challenging and open problem. Future research once addressing this issue could then determine the ratio of the time sleeping vs. time spent in bed (sleep efficiency), which is an important metric when determining sleep quality. It will also be important to study both good and poor sleepers.

p.s. I am still trying to extract the results for the Zeo. I need more time.

avi123 wrote:Any studies of correlations between Zeo and EEG, PSG?

Only for those interested


The Zeo is mentioned here from pg 10

http://robertdickerson.net/papers/phdproposal.pdf

System for the Collection, Storage, Analysis, and Reporting
of Objective Behavioral Measures

Ph.D. Dissertation Proposal
Robert F. Dickerson

Timeline: The last three months of the summer 2012 will be used for thesis writing, with
an intended graduation date of August 2012.
••••••••••••••••••••••••••••••••••••••••••

p.s. I am trying to extract the results for the Zeo. Need more time.

Very interesting. Apparently they have not yet performed the study or released its results?
I was looking for an evaluation or opinion of the efficacy of the Zeo, but couldn't find any within the proposal.
Something to look forward to I guess.

Regards, Nate

Here's something from their own webpage FWIW:

http://www.myzeo.com/sleep/validation

Nate

...A smattering of returns using Google Scholar:
http://scholar.google.com/scholar?start ... s_sdt=0,14

Here is a brain twister for you... It comes from about 30 years experience in data analytics and can be really confusing to some people but it's a real thing...

If I use the Zeo information as an input to my analysis, the accuracy of the unit no longer matters, just as long as it is repeatable (consistent reading under same conditions)
...however...
If I use the Zeo directly as a performance measure, then it must be both accurate and repeatable.

As I use the Zeo as an input, technically, I don't care how accurate it is.

Braintwister for today !

MaxDarkside wrote:If I use the Zeo information as an input to my analysis, the accuracy of the unit no longer matters, just as long as it is repeatable (consistent reading under same conditions)

And that's why (how) it's useful in establishing trends.

Jay

As many other CPAPers I would like to have a single device next to my XPAP that will tell me my treatment's values as close as possible
to the standards of all sleep clinics PSGs in the U.S. as established by the AASM.

The one that I am familiar with, that almost does it, is the Watch PAT 200 from Itamar Medical Co, in Caesarea, Israel.

Notice that the Watch PATs do not use EEG for establishing sleep architecture but use Actigraphy instead.

Re-edit here. It's too soon.

I have no grudge against ZEO as long as it would provide all the data that I mentioned (i.e. provide data like the WatchPat 200), be as accurate, and sell, to any one below $300. As to interpretation I am still debating.

See here what the Watch PAT provides:

http://www.gds-medtech.com/watchpat-pro ... =2&index=6

And:

http://www.centerforsoundsleep.com/blog ... onitoring/

p.s. MaxDarkside, is it you in your Avatar showing off?

robysue wrote:

Lizistired wrote: The Blue graph is based on 30 second data points. The website graph is 5 minute data points. The 30 second graph just gives you a better idea how disrupted the sleep stages are.

What software did you use to draw the blue graph with the 30 second data points?

I use Open Office Calc (Free Excel). After ! export my data from the website, I open the .csv in a spreadsheet. Then copy the "detailed sleep graph" column (bx) to a new worksheet. Then select that column and format, "text to columns", (select space delimited).
Then select the night you want, usually last night so the last row. Select Insert Chart.
Let me know if you have a problem. I think I found the instructions on their website somewhere or Jay sent me the link.

avi123 wrote:Any studies of correlations between Zeo and EEG, PSG?
Only for those interested
. . .
http://robertdickerson.net/papers/phdproposal.pdf

System for the Collection, Storage, Analysis, and Reporting
of Objective Behavioral Measures

Ph.D. Dissertation Proposal
Robert F. Dickerson

Thanks for the post, Avi. Very interesting reading. I send him an inquiry about his take on the Zeo to seek further comments.

There needs to be a category for voting: "Interested in the Zeo and purchased one today a result of this thread."

Mine will be here Thursday....ordered the bedside model after reading about issues with the travel version. Plus I have been meaning to get a new clock.

I'm looking forward to giving it a whirl.

avi123 wrote:As many other CPAPers I would like to have a single device next to my XPAP that will tell me my treatment's values as close as possible to the standards of all sleep clinics PSGs in the U.S. as established by the AASM.

The one that I am familiar with, that almost does it, is the Watch PAT 200 from Itamar Medical Co, in Caesarea, Israel.

Check this site out as well: http://sleepgroupsolutions.com/modules/ ... x.php?id=5

Are there any cpaptalkers trying this out yet?

Papit wrote:

avi123 wrote:As many other CPAPers I would like to have a single device next to my XPAP that will tell me my treatment's values as close as possible to the standards of all sleep clinics PSGs in the U.S. as established by the AASM.

The one that I am familiar with, that almost does it, is the Watch PAT 200 from Itamar Medical Co, in Caesarea, Israel.

Check this site out as well: http://sleepgroupsolutions.com/modules/ ... x.php?id=5

Are there any cpaptalkers trying this out yet?

Reply,

The Watch PATs 100, and 200 are available from sellers that are connected to the manufacturer Itamar Medical Co. They are like DMEs but are different kind of organizations. They get the Watch PATs for several thousands dollars each from Itamar Medical and rent them out to individuals for nightly "at home testing for sleep disorders". If you are interested to be tested they prefer that you get a physician or a dentist involved. It is covered by insurance and Medicare.

As to costs for the nightly testing, it is about $130 for checking if the result for OSA is positive. If yes, then there is an extra cost of about $90 for interpretation report to be sent to your physician.

Check here more about it:

http://www.1stlinemedical.com/

avi123 wrote: . . . The Watch PATs 100, and 200 are available from sellers that are connected to the manufacturer Itamar Medical Co. They are like DMEs but are different kind of organizations. They . . . rent them out to individuals for nightly "at home testing for sleep disorders". If you are interested to be tested they prefer that you get a physician or a dentist involved. It is covered by insurance and Medicare. . . about $130 for checking if the result for OSA is positive. If yes, then there is an extra cost of about $90 for interpretation report to be sent to your physician.
Check here more about it:

http://www.1stlinemedical.com/

Fascinating. Looks like this company has made a good deal of progress in the 'PAT' technology approach to sensing and gauging the symptoms of sleep disorders. It's a whole new approach (at least to me) with apparently superior accuracy. Of course, the above link is off their own website so I'll be looking for corroboration from independent sources.

At $130 (or rather $220 including interpretation), it's for one night, but they do all the work and what would be very time-consuming analysis for most of us to attempt on our own. And it's a tiny fraction of what a new PSG lab test would cost us, especially if insurance covers this.

Good find. Thanks.