OpenSeizureDetector is a suite of programmes and hardware designs intended to detect a tonic-clonic (grand mal) epileptic seizure (fit).
There are commercial seizure detectors available on the market, but all the ones I have seen requre either a connection to the patient, or installation on or under a mattress. Unfortunately our son, Benjamin is autistic and would not tolerate having anything attached to him, and he refuses to sleep in his bed - instead he sleeps on the floor behind his bedroom door. This means that nothing needing attachment to the patient or installation in a bed will be of use to us. There is some more information about Benjamin and his challenges here.I have experimented with detecting a seizure using vibration (an accelerometer), video (tracking movement), audio (haven't got far with this), and using a depth camera such as a Microsoft Kinect sensor. The Kinect version is the most promising at the moment, and is the subject of most of this site, but you can find information about the other versions on my github repository and my blog. We currently have two working prototypes, one installed in Benjamin's bedroom that is based on a Kinect depth camera, and using a number of Raspberry Pi based Digital Video Monitors to show us he is ok.
I have just completed a version based on a Pebble Smart Watch which connects to an Android mobile phone to share the analysis data. It will be a challenge to persuade Benjamin to wear the watch (we may need to put pockets in his sleep suits to hold it), but it looks like the most reliable way I have of detecting the shaking associated with a tonic-clonic seizure.
Kinect VersionAt the moment it is a functioning prototype breathing detector / alarm - it can detect that I am breathing, and alarm if I can hold my breath long enough (it is using a 30 second time series, and it is hard to stay perfectly still for that long!). The benTV digital video monitors talk to the web interface to display the results and display alarms by chaning colour (I don't trust it enough yet to make an audible alarm in the night...).
Pebble Watch VersionFunctioning prototpype comprising an app running on the watch to do analysis, and send analysis results to android phone. The android app acts as a web server so client devices can see the results and raise alarms as necessary.
- Soak test the current system to see how often it raises false alarms, then modify as necessary.
- Extend the current system to do frequency analysis to see the frequencies of movement - this may allow me to get towards detecting the shaking associated with a seizure.
- Add an audible alarm to the video monitors
- Write an Android Application so it can work on a portable device such as a tablet computer.
Pebble Watch Version
- Add data logging to android app so we can see how many false alarms we get.
- Update android app to provide indication of ip address of phone so users can connect to it easily.