A continuous, non-invasive blood chemistry monitor

Similar projects worth following
The goal of the BloodWatch is to monitor blood chemistry continuously and non-invasively via dialectic spectroscopy. It will initially target blood glucose and oxygenation, but monitoring of other blood chemistry components should be easily added. Analysis of the spectrographic to extract relevant measures will be via a deep learning neural network, trained using reference measurements from standard blood glucose/oxygenation meters.

At this time this is NOT for diagnostic use, and the project is as much about learning what the barriers are to making this practical as anything (there have been several similar blood glucose monitors announced over the past decade, but all have failed to come to market and disappeared).

Keen to find people with experience of fine SMD work, dialectic spectroscopy, analog/digital circuits in the 200MHz bandwidth range.

Still very much in the research part of this project, and have a lot to learn before the end. For those interested, some background reading on Dielectric Spectroscopy can be found at:

Will be dealing with waveforms in 1 to 200 MHz range, probably using either a DAC or a waveform generator, and either an ADC or a gain-phase detector. Currently expect this to be a wrist-worn device (subject to bench testing - finger/ring style sensor might be better, or a patch sensor against abdomen) and bluetooth LE to interface to phone or smart-watch.

Unsure if data-analysis will be performed on-device or not. In either case I will try using an artificial neural network to extract the desired information from the raw data in the hope that this will be able to cope with the noise/unknown variables inherent in the data.

Diabetes is a well known medical condition that affects an estimated 387 million people worldwide, or 8.3% of the adult population. It contributes to many deaths and has a huge negative impact on the world economy. The health and hence economic impacts can be reduced my improved management of the disease. Key to this is monitoring of blood glucose levels. At the moment the best available options are either testing via blood-prick tests a few times a day, or use of an invasive blood glucose monitor (which have high ongoing costs due to expensive sensors which generally last only a week).

The BloodWatch aims to help address these problems by providing continuous blood glucose readings without the need for an invasive or disposable sensor. Hence the ongoing costs to the user are practically zero. It can monitor blood glucose levels when the user can't (e.g. when asleep) and alert the user if a problem is detected (low blood sugar events during the night art particularly dangerous).

Beyond its initial targeted blood glucose monitoring functionality, I expect the BloodWatch will be able to monitor a number of health related blood factors and provide assistance to users other than diabetics. For example:

  • Blood oxygenation
  • Pulse rate
  • Cholesterol
  • Fat levels (possibly tracking from your most recent meal)
  • Ketones
  • Alcohol
  • Salts
  • Signature molecules for various diseases

  • Back!

    Linus Dillon12/01/2015 at 22:49 0 comments

    Right, just starting to get back to this now after being pulled away by family stuff. Have picked up a Tsunami (basically an Arduino coupled with a basic DDS waveform generator and frequency/phase/amplitude detector). It is almost certainly too slow to pick up a decent glucose signal (or possibly anything) but it will let me get some of the basic ideas straight before I start designing PCBs, wrestling with high frequency signals, buying possibly pricey DDS chips and so on. Regardless I'll keep posting here.

    I've had one or two offers of help/interest and I will be following these up too.

  • Lack of progress

    Linus Dillon07/13/2015 at 06:15 0 comments

    Been fairly busy recently with other obligations. As a result not a lot of progress has been made. What I HAVE managed is:

    • Gotten the waveform generator and phase/amplitude detector soldered not breakout boards.
    • Wired everything up on a breadboard with an Arduino clone to test things out.
    • Gotten a basic heartbeat (sine wave) using a 16MHz clock generated by a CMOS Schmitt trigger chip driving the waveform generator - note not being controlled by Arduino at this point.
    • Switched to a clock generator breakout from Adafruit so I could drive the waveform generator at 75MHz; thought this would give better results than the jury rigged crystal+CMOS chip.
    • Struggled to get ANYTHING to work with this setup. Unable to get waveform generator to respond to Arduino or generate clean output.
    • Realised that breadboarding circuits running at 75MHz is PROBABLY NOT going to be successful.

    So there we are. I think I'll try dropping everything down to maybe 1MHz max and see how that goes. The breadboarding is largely to validate how things are wired up, although I had hoped to do some measurements too. Have to see on that. This being my first project involving high frequencies AND a significant analog signal portion, I knew there would always be a sizeable learning curve. As for when I'll have time, that's less clear.

  • Parts

    Linus Dillon05/21/2015 at 11:20 0 comments

    Well, I finally have the parts to proceed to the breadboard stage; waveform generators (both low and high speed), phase/gain detectors and breakout boards to match. I'll be soldering them up this weekend and start working towards having the low speed waveform generator and gain phase detector on a breadboard talking to an Arduino (at this stage I'm after easy to work with; speed, features, form factor and power usage will come when I proceed to prototype).

  • Initial Plans

    Linus Dillon04/30/2015 at 10:58 0 comments

    I've spent a little time looking at what I can find on the web about Dielectric Spectroscopy, in particular in relation to glucose. I've not yet found the ideal range of frequencies for glucose detection; one source suggests as much as 80MHz, while another as low as 40kHz. The higher the frequency the harder things get. Also, I need to get some experience of the technique and tech in general. So, plan is:

    • Start with breadboarding/test equipment to create a minimal dielectric spectroscopy setup.
    • Start at low frequencies (tens of Hz) and work up (to tens of MHz).
    • Test different electrode styles and evaulate their response (I can think of three styles already).
    • Test using known materials in solution (plain water, saline, glucose, possible blood itself) and characterise their signatures across different frequency ranges.
    • Test using a running solution (flow through tubing) at different rates and characterise impact this has on the signature. Include case where some material within scope of electrode is NOT affected by flow.

    I have more or less settled on an initial architecture for the device; a discrete waveform generator coupled with a discrete phase/amplitude detector, both of which are coupled to a basic micro controller. At this stage this doesn't need to be that powerful; a simple Arduino will even do (I can solve for what micro controller and bluetooth hardware to use latter).

    If I move forward then the waveform generator may be replaced with something more basic (all I need is a sine wave with a frequency I can control and which covers the desired range).

    So, I've ordered the following bits to start with. Will be a few weeks before they arrive.

    • AD8302 Phase/amplitude detector (small and cheap)
    • AD9954 High-end DDS waveform generator (large and pricey, 400MSPS, 200MHz max output)
    • AD9834 Mid-range DDS waveform generator (small and cheap, 75MSPS, 37.5MHz max output)
    • Signal Generator Module (0-40MHz, based on AD9850 - saves time during the bench testing phase)

    In the meantime I'll use my other projects to learn more about SMD design/assembly so when the time comes I'll be able to make this into something reasonably compact and wearable!

View all 4 project logs

Enjoy this project?



Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates