Welcome to the Age of Biosensing Wearables

Increasingly sophisticated wearable devices can now measure heart rhythms, glucose, and more, but what will it take to get consumers and health care providers to adopt them?
When I was 10 years old, I had a crush on a girl named Mary. Mary knew my love was real—when I saw her, my mood ring would turn from a swirly green to a wavy, purplish-orange. As an adult, I now have more sophisticated tools—tools that measure information related to my health and fitness. I’ve traded in my mood ring for my Fitbit®, which I have on from the moment I leave the house to walk our dog, Tarot. Tarot has an activity monitor too, which tracks his location by GPS in case he runs off. And, at the end of the day, my wife compares my results to Tarot’s on her smartphone and decides who sleeps on the floor. Wearables have changed our lives.
The past few weeks have brought forth announcements by major technology companies touting new smart watches and digital platforms, while smaller innovators are releasing a range of new wearable (and ingestible) devices. Wearables are now moving beyond the well-established realm of tracking movement, and new entrants are developing devices that continuously monitor a broad range of physiology—from posture to brain activity—and convert the information into a signal output. With greater connectivity and computing power in our pockets and on our wrists, we seem to have entered a new era: Welcome to the age of biosensing wearables.
Given the potential of these devices, one obvious question is if and when will health care take advantage of them. Unlike blood pressure cuffs and other devices that are driven by physician utilization, consumers appear to be driving the growth and use of wearables, and investors are taking note. According to Rock Health, venture funding of biosensing wearables has risen from $20 million in 2011 to $229 million in 2013.¹ Wearable devices are becoming increasingly sophisticated, now measuring heart rhythms, oxygenation, glucose, blood pressure, and more.
What might it take to get consumers to use these devices for their health—not just for fitness and wellness?
Convenience. If it isn’t easy, people probably aren’t going to use it. Many of us have tried a traditional pedometer at one point in our lives, and just as many of us abandoned it within two weeks of starting. Through wireless data collection, made possible with low energy Bluetooth, data appears automatically with these new devices, and tracking is simplified, requiring little to no intervention by the user.
Interoperability. Many consumers want a seamless experience. Having to switch from app to app to see the outputs of various sensors can be a recipe for frustration and confusion.
Privacy. While consumers may be perfectly happy to share their step count with friends, co-workers, and even strangers, they might be reluctant to do so with their employers or their insurance companies. Moreover, they may have no reservations about sharing their physical activity on any given day, but they might feel less inclined to reveal their glucose or blood pressure readings.
Motivation. While each of the prior issues matter, perhaps the most important factor leading to the continued use of a tracker is making it social. Whether users are motivated by sharing, competing, or obtaining rewards, the use of social media platforms has been cited as a strong determinant of ongoing use and achieving goals.²
As consumers get onboard, what might it take for the health care industry to recognize the value in these devices?
Rationale. Just as consumers need to be motivated, health care needs a reason to adopt new technology. While innovation is exciting, unless there is a clear return on investment (ROI), few often embrace it. With the rise of value-based care and its focus on population health, prevention, and outcomes, new opportunities could arise. Consumer engagement for health promotion and medication adherence, management of chronic disease, and remote monitoring now present clearer paths to ROI.
Validity. While the makers of biosensors make many claims about their accuracy, some are no more precise than my old mood ring. If medical decisions are going to be made based on these devices, they will need to truly reflect what they purport to measure. Some companies and developers have gone as far as to get approval from the U.S. Food and Drug Administration (FDA), and many more may follow.
Reliability. The data needs to flow consistently and accurately in order to be useful. Gaps introduce errors and a degree of uncertainty that may often be acceptable to consumers, but can render the data useless to clinicians and researchers.
Evidence. Amid the excitement over the potential of these sensors and companion apps, right now there is scant evidence that they contribute to improved outcomes. And, as I have said before, just because I have a fitness app on my phone, it does not make me an athlete. Thankfully, a number of pilot studies have been launched to understand what devices have an impact under what circumstances.
Finally, as biosensing wearables work their way into health care, industry stakeholders should consider asking if they are measuring the right things. Physiologic signs would seem to provide incredible insights. However, if I wanted a sensor to alert me to potential problems with my aging parents, I might glean more from a sensor on their refrigerator door than a continuous heart rate monitor. Tracking my steps is one thing—but it doesn’t tell you, like my credit card does, that my walk included purchases at a fast food restaurant and a cupcake bakery.
No matter what, these devices are probably here to stay, and more are on the way. How they address the needs of consumers and the health care industry will determine which are truly effective, and which wind up in the doghouse.
— By Harry Greenspun, M.D., Senior Advisor, Deloitte Center for Health Solutions, Deloitte LLP

http://deloitte.wsj.com/cio/2014/07/28/using-sensor-technology-to-lower-elder-care-costs/

To track seniors’ medical conditions and surroundings more closely, researchers are experimenting with more advanced smart sensor networks that provide remote caregivers real-time insight into the health and well-being of in-home patients. For example, research teams led by Diane Cook, Ph.D., of Washington State University and Nirmalya Roy, Ph.D., of the University of Maryland, Baltimore County, supported by a grant from the National Science Foundation (NSF), are exploring how to retrofit homes with sensor networks that monitor a resident’s behavior and activity levels.

These sensor-enabled homes use machine learning to recognize behavior patterns such as eating, sleeping, and movement, and then identify and report any signs of illness or cognitive degeneration to caretakers and physicians via the Web or mobile networks. The monitoring capabilities also alert physicians to changes in the physical and mental health of their senior patients, and allow them to intervene before adverse events occur. Finally, enabling seniors to live safely in their homes for as long as possible will likely help these individuals retain their sense of independence and self-confidence longer.4

Compared to the high cost of traditional assisted-living facilities and nursing homes, sensor-enabled smart homes are relatively inexpensive. Retrofitting a home with sensor technologies costs $2,500, on average, for hardware and installation fees, plus a modest monthly fee for monitoring and analyzing the data.5

Not having to plug in at all would be really impressive.  But it sounds like a helluva stretch goal – on a par with getting it all into a bracelet!

I'm pondering strategy.  I took a helluva long time to get all parts of the hardware working with TS05.  Just getting drivers for the various chips to work took me an embarrassingly long time, and this all has to work before the higher levels of software can be started.  Getting all that goinf for TS05 before I started on the analysis stuff was an error.   I'd like to avoid developing code for all the secondary sensors until we know: 
  1. how many ranging ops we need to do in one day in realistic conditions, 
  2. how much that sucks out the battery,
  3. how repeatable the ranging values we get are.  

For that we need a wearable unit, and some anchors, and the software needs to be getting 

<timestamp><anchor number><range>
records somewhere where we can read them. Almost nothing else matters until that simple pipeline is working and we can see it is going to come in under power budget.  From there on in everything else is just optimization – making Dewbly range more intelligently, report more useful things, and use less power to do it is optional firmware work, and the cloud-based analysis software can proceed independently.
Developing ARM code to get ranges from the DecaWave is going to take me a while.  I've not done drivers for an ARM, and I'm sure there are lots of gotchas.  
But before that even, there's a lot of core (OS) code that has to work before I even get to the ranging stuff. For example I have to get timestamps working, and find a way to initialize the TOD clock with the correct time-of-day for example. 
I'd really like to start with the simplest hand-built OS so the code is small, leaving the possibility of flash space left over to store range data on the ARM itself.  Then I don't have to get the separate flash memory drivers working.
But I do have to figure out how to get the data up to the cloud, or PC.   Maybe it comes over a serial line to start with? I suppose if I have the DecaWave interface working, then I could send all the data over that wireless channel  initially, and if that didn't sink the power budget then hey – we are to first base with the data pipeline, and from there on everything is just 'optimization'.  It's just that the optimization is going to be the bulk of the firmware I expect.
I'm telling you all this because it may have some bearing on your hardware strategy if you know that I'm unlikely to get around to much other than the DW chip for quite a while.  Maybe it buys you some time to figure out some of the more esoteric details of charging, or enclosure?  

//Mik
On Tue, Aug 12, 2014 at 11:50 PM, David Carkeek <dcarkeek@gmail.com> wrote:

It was I who made the mistake in units. 

The Li-Po battery has a lot more usuable energy in it. If we manage energy usage correctly I think we can have a very long battery life. If the solar charging works it may never need to be plugged in.
Which reminds me again that I need to be sure there is a PV cell we will be able to use on the Minitec enclosure.
On Tue, Aug 12, 2014 at 8:31 PM, Mik Lamming <mik@lamming.com> wrote:

Ah.  That showed my complete ignorance didn't it 😀

I always had the model that physical volume was a good estimate of battery capacity, and I was excited to see that my model was violated, but it seems I misunderstood.  Oh well, at least my model is intact.  I wonder when we will get 1000x extractable energy in the same volume.  The harder the challenge the more the triumph.


//Mik

On Tue, Aug 12, 2014 at 11:26 AM, David Carkeek <dcarkeek@gmail.com> wrote:

It's close to the same amount of energy in a CR2032 and fully charged small lithium-polymer pouch battery. I put the CR2032 in W-h but labeled it in mW-h.

CR2032: 0.675 W-h
3mm x 20mm x 30mm: 0.666 W-h
5mm x 30mm x 20mm: 0.888 W-h
4.7mm x 22mm x 29mm: 1.11 W-h

But the advantage of the li-po battery is that large current can be drawn from it without trashing the energy capacity. If button cell is pulsed at high current (high being just a few mA) the energy capacity is much less. I am optimistic that the battery capacity will be adequate.  Making a ring is another matter.

On Tue, Aug 12, 2014 at 9:47 AM, Mik Lamming <mik@lamming.com> wrote:

So a 1000x to 1600x a CR2032.  That a significant difference.  If 1600x lasted one day, then 1000x would be a bit naff.  On the other hand if a 1000x lasted a week, then 1600x is not a significant consideration.

In general, if the difference between one battery and another was the difference between running less-than-a-day and running more-than-a-day then I'd go for more-than-a-day at almost all costs.  If the difference was on a week boundary then I'd go for small.  But since I don't have much clue what our burn rate will be, it's a bit academic.  
At this stage I think I'd maximize your convenience because getting stuff to work ASAP is the priority.  Next we should have a bunch of anchors and a wearable each.  
The first goal is to get daily activity data out of Dewbly onto the screen and into an Excel spreadsheet so we can look at it. It doesn't matter if the devices are big, or small initially.  If we can't recognize patterns in the data, then it will be bloody hard to get software to see them, and we will be done!  
The second goal is to make it easy to gather and upload data every day.  I think the hardware department should take responsibility for making it convenient and reliable for us to gather data day after day, and conveniently recharge with minimal hiccup.   For that reason I think we should try very hard to build two sets of hardware so we can both see the fruits of our labors, and maximize usability – perhaps I mean minimize PITA issues.  
Even if the data presentation software is very primitive, I think it would be very good for you to be able to see what we get.  You'll have different insights and good ideas for how to make things better, or do more interesting things.  I also think it would be good to be able to show early results to potential collaborators if we have any promising results.  Maybe there are retired data analysts, or firmware people who might get excited by good results.
I reckon it will be quite some months before we have anything working well enough to have any insights about battery life.  But firmware work can try and get the burn rate down while hardware can look for ways to get more power into the budget, and recharging schemes.


//Mik

On Tue, Aug 12, 2014 at 7:58 AM, David Carkeek <dcarkeek@gmail.com> wrote:

I have received four of the battery samples I ordered from Ali Express. It looks like a 180mAh battery (3mm x 20mm x 30mm) fits between the PC board posts and leaves enough height for everything else. 3.6V * 0.18Ah = 666 mWh. Recall that a CR2032 button cell is 225 mAH * 3V = 0.675 mWh.

Another larger battery might be made to fit. 4.7mm x 22mm x 29mm, 300mAh * 3.7V = 1110 mWh. One of the posts would have to be cut.
One more sample needs to arrive. It might be the best one if thickness isn't an issue because it will fit between the posts. 240 mAh * 3.7V = 888mWh.