Custom Hardware Seizure Detector

Contents

Objetives

Commercially available seizure detectors / alarms remain expensive and many require monthly subscriptions, which will be difficult to afford for people on low incomes.

I am therefore looking into the possibility of running the OpenSeizureDetector system on very low cost hardware.   This, coupled with the free and open source nature of OpenSeizureDetector should make seizure detectors / alarms available to people on lower incomes.

Hardware used for Feasibility Study

The version we are considering at the moment uses a computer costing £2.50 and an accelerometer chip costing £1.40.   If we can run the OpenSeizureDetector software on this, and package it with a battery into a wrist-worn device, it could make a very low cost seizure detector.

Development will take place in a number of stages, where we consider the feasibility of continuing at each stage:

  1. Feasibility Study – does it even look possible
  2. Proof of Concept – get something working as a seizure detector, and look at battery requirements – we need to get it to have at least 12 hours battery life to be useful.
  3. Prototype – Produce miniature hardware and 3D printed wrist watch case and see how it works.
  4. Production ?  If there is interest we could look at small scale manufacturing to make them for end-users.

I would very much welcome thoughts on this – is anyone interested?   How much would you be willing to pay for the hardware (because although the basic hardware is cheap, paying someone to assemble it will cost money).

Graham (graham@openseizuredetector.org.uk)

Progress Updates

Feasibility Study 2 (November 2018)

I parked this idea in the middle of 2017 because I was struggling to get the current consumption of the ESP8266 system down low enough to meet my target of over 12 hours battery life.

I have had a few discussions with different people lately and there seems to be interest in a custom device that measures a number of  different parameters, so I am resurrecting the project.      What we are aiming to do is:

  • Build a watch-sized device with over 12 hours battery life (preferably 24 hours)
  • It will communicate with an android mobile phone over bluetooth (to make set-up easier and reduce watch power requirements compared to wifi).   It will also save a lot of development effort because we can use the OpenSeizureDetector Android App as the basis of the phone software.
  • It will contain several sensors such as:
    • Accelerometer (adxl345)
    • Optical heart rate and blood oxygen saturation monitor (MAX30102)
    • Skin conductivity (home made analogue electronics?)
  • It will also contain a small LCD or OLED screen – there is no good reason for this other than Benjamin really likes his Pebble Seizure Detector screen where he can tap his arm and make the light go on – but this will cost me both device size (thickness) and battery life.

Status as of 20 November is:

  • Targetting ESP32 platform because It has built in bluetooth, so saves having a separate IC on the board, and the SDK seems much nicer than the earlier ESP8266 SDK – and I think that we can switch off unused peripherals and CPU cores to get the battery usage down.
  • Toolchain is working – can compile software and flash it onto the device using a pre-built module with USB connection.   Just working on a breadboard version first for feasibility study.
  • Starting to get the i2c communication working – I can detect the ADXL345 chip, but need to convert my library from using the ESP8266 SDK to the ESP32 one.
  • Started on the electrical schematic and PCB layout – intention is to get the PCBs made by a low cost supplier in china and I will populate the prototypes myself.

Feasibility Study (June 2017)

We can:

  • Build software for the ESP8226 microcontroller board
  • Interface the board to an ADXL345 accelerometer chip.  (we are using a Wemos D1 mini board to and an ADXL345 module, both off ebay for the feasibility study, to make the soldering easier).
  • Collect accelerometer data
  • Process the accelerometer data using the same algorithm (and pretty much the same code) as the Pebble Watch Seizure Detector.
  • Write the seizure detector data to a computer screen connected via USB.
  • Put the device into setup mode using a switch attached to the Wemos D1 board (acts as a wifi access point and web server to receive basic info on actual wifi and server configuration for operation).
  • Write data to the on-board flash memory for persistent storage of settings.
  • Send data to a server using http GET and POST requests.

See https://github.com/OpenSeizureDetector/ESP8266_SD for more information.

The above is all of the software components we need to declare the project feasible.

The only outstanding issue is power consumption – the board currently consumes about 80mA at 5V.   We intend to use a 150 mAh battery, so at that consumption rate it would be drained in about 2 hours, which is no good.   I believe the problem is that the wifi radio is on all the time.  The ESP8266 should consume more like 20mA if it is in modem_sleep mode….but modem-sleep mode does not seem to work using the esp-open-rtos sdk – the solution may be to go back to using the Espressif SDK instead, but need to do a simple test to make sure that we can reduce power consumption that way before trying to convert.

Update 24 July 2017 – if we switch to using the openRTOS SDK provided by Espressif rather than esp-open-sdk we can get the device into MODEM_SLEEP mode – a brief test showed that this drops the power consumption to 25mA.  This would give a battery drain time of about 6 hours for a 150 mAh battery.  This is a bit short, but the batteries are pretty small so we could use two to give a 12 hour battery life (and there are more tricks we can look at to make the device sleep as much as possible to improve that).  So I think the power consumption issue is solveable, but I will need to convert the feasibility study code to the Espressif SDK to confirm.

So the feasibility study stage is nearly complete, pending sorting out of the power consumption problem.

Case

Laura has been working on designing a 3d model of a case for the watch that can be printed on a 3d printer – looking promising – a bit chunky at the moment but that is because we have two circuit boards stacked on top of each other – once we go to attaching the accelerometer chip directly onto the ESP8266 board it will be neater.