A user has identified a problem with SMS alerts not sending on a Motorola phone – I have tested it and it is also failing on a Samsung A17, so is probably related to permissions on newer versions of Android. I saw the issue when I used the ‘Test SMS Alarm Notification” menu item in the main app screen – you see a popup saying “ERROR – FAILED TO SEND SMS MESSAGE”.
I will aim to do a fix and publish it over the next couple of days.
In the meantime, if you need to use SMS alerts, you can work around the problem by granting ‘Phone’ and ‘SMS’ permissions in the main phone settings screens:
Go into the main phone settings menu, then select Apps, then OpenSeizureDetector. This will give you the phone level settings for the apps – one of which is Permissions.
Click on Permissions and it will show you which permissions are allowed and Not allowed.
Enable Phone and SMS permission from the list.
Once you have done that, SMS alerts should work again in V5.0.8 so no need to downgrade to get SMS alerts working again.
We are pleased to (finally!) release the new Version 5 of the OpenSeizureDetector Android App – it should be on Google Play Store shortly.
There are several major changes in this version:
The ability to download Machine Learning (“AI”) models that we are making with the data provided by users via our ‘Data Sharing’ programme.
A first-use setup “Wizard” to guide new users through setting up the system for the first time (using either a PineTime or Garmin watch)
A simpler user interface – it defaults to a “Basic Mode” with no graphs, just a simple OK/ALARM status and Raise/Mute alarms buttons. You can disable Basic Mode in the General Settings page to get something that looks more like the current version.
A lot of ‘behind the scenes’ improvements to improve the stability of the system and make it easier to maintain in the future.
To simplify the user interface, it only lists PineTime, Garmin, Network and Phone data sources by default. If you want to select one of the more ‘experimental’ data sources, there is a ‘Developer Mode’ that can be enabled in the settings.
The Machine Learning model that the system recommends during installation is the ‘Run24’ one that I made in January and have been testing since. I am very pleased with its performance for or son, with it having >90% detection reliability. When I analyse the recently provided data from users, I am seeing just over 80% reliability, so it is comparable to the existing algorithm with ‘flap’ detection – but it should give fewer false alarms. Note though that it does still false alarm for doing things like using hand tools and travelling in a vehicle. Once the release of this version of the Android App has settled down, I will get back to working on a better model and making that available to users.
Thank you to everyone who has taken part in the beta testing of this app – you have found issues and made some very helpful suggestions to improve it.
When you find issues with this version, please report them via our bug tracker, or email graham@openseizuredetector.org.uk.
Apologies to the beta testers, but the ‘final’ test version that I published last night (V5.0.7) is partially broken – The Basic Mode user interface works, but the Advanced Mode crashes because of an issue with the graph drawing library.
I would suggest not upgrading if you haven’t done so yet.
If you see the issue, you get a crash when in advanced mode. To get it working again:
Do a ‘Force Stop’ (or re-boot the phone) to make sure OpenSeizureDetector is properly shutdown.
Start OpenSeizureDetector
As soon as the start-up screen appears, press the ‘Edit Settings’ button to bring up the settings screen.
Go to General -> Basic Mode and enable the Basic Mode checkbox.
Then when you exit the settings screen, the system will re-start into Basic Mode and should work correctly.
I’ll look into what caused this and issue a new ‘final’ test version later this evening.
I have updated the web site at https://openseizuredetector.org.uk to try to make it simpler for new users to find information. There are is a new home page that links to installation instructions, a user guide page for existing users and a developers page. I’m afraid the developers section is still very sparse – detail to follow.
The old web site pages are still there, so if you see something that is missing, please let me know and I’ll move it across.
Please contact graham@openseizuredetector.org.uk with any comments or suggestions.
My name is Jannie, and I am currently working on a graduation project focused on improving the design and user experience of the Open Seizure Detector app.
To better understand real user experiences, I am looking to speak with current users of Open Seizure Detector. This could be through a short online interview or chat. Depending on what you are comfortable with.
Topics may include:
first-time setup
daily use of the app
notifications and alerts
trust and reliability
features that work well or could be improved
The project is purely educational and focused on improving the app experience for users.
If you would be interested in participating or sharing your experiences, please feel free to contact me at:
Is anyone using a Garmin Forerunner 55 with OpenSeizureDetector? A user has reported a strange behaviour where the first time they used it, it worked correctly, but since then it has only been sending data every 20 seconds rather than every 5 seconds. (https://github.com/OpenSeizureDetector/Garmin_SD/issues/82)
If you are using a Forerunner 55, please will you send me a log file off the watch? (you need to put an empty text file in GARMIN/APPS/LOGS called GarminSD.txt, then run the app for a couple of minutes, then when you re-connect the watch to the computer, GarminSD.txt, and possibly GarminSD.bak will contain data that we can analyse).
With the release of Version 5 of OpenSeizureDetector approaching, I wanted to re-visit the logo for the project. At the moment we use a public domain ‘Star of Life’ image which Jozefus Gaarthuis kindly drew for us. But this does make the app look rather medical, and I see it as being more related to supporting care for people than a medical tool, so I’d rather move away from the star-of-life.
So in my completely non-artistic way, I have been trying to keep the basic shape (6 arms) and blue colour, but make it softer so more related to care and warning people, rather than medicine. I have tried to turn the arms of the star into petals, add a ‘bell’ alarm icon, and some purple lightning to try to give a hint that it is to do with epilepsy and generating alarms…..And I ended up with this:
I’m fairly happy with it, but it does look a bit busy – would it be better without the lighter set of petals maybe? I don’t think this clashes with other organisations’ logos, but I could be wrong….
Thank you to everyone who has helped with testing the draft Version 5 of OpenSeizureDetector. As a result of this testing we have added a few more features to the app before it is released. These include:
User Selectable alarm sounds – If you enable ‘MP3 alerts’ you can select audio files from your phone to use for the Warning, Alarm, and Fault alert sounds.
Cancel SMS Alerts from the watch – if a false alarm is triggered, there is a delay to allow you to cancel it before an SMS alert is sent. It can be fiddly to open the phone and get to the app before the alert is sent, so now if you are using a Garmin watch you can select the Mute function and this will cancel any pending SMS alerts without you opening the phone app.
Fall Detection Algorithm display – there is now a page in the app which contains the maximum and minimum acceleration values in each analysis period, to help with setting up the fall detection algorithm.
Plus there were quite a few minor bugs fixed as a result of the testing.
I should be publishing Version 5 for all users in about a week (once I have produced some documentation and example videos for it….), so if you have any other features you would like to see added to the app, please get in touch as soon as possible so I can see if we can implement them before V5.0.0 is published.
I am keen to move over to a Machine Learning algorithm to reduce the number of false alarms from the OpenSeizureDetector system. To do this I have had to make some quite significant changes to the Android app, so I have taken the opportunity to modernise the code and user interface. There are so many changes I am calling it a new major version (Version 5).
Alpha testers have been testing draft versions of Version 5 of the phone app for several weeks now, and it is ready for testing by more users. I will be publishing it for anyone who has volunteered to be a beta tester by registering here: https://play.google.com/store/apps/details?id=uk.org.openseizuredetector in a few days time.
The things that are changing in this major release are:
There is an installation ‘wizard’ to guide new users through setting up the system (setting up the PineTime watch and selecting which seizure detection algorithms to use). It will still work with Garmin watches, but the set up is not automated like it is for PineTime. The wizard will also help set up the Network data source to generate alarms over WiFi. There is a video walk-through of the full installation process available here: https://www.youtube.com/watch?v=v9GzcKG0mVc
The user can select one or more Machine Learning (“AI”) seizure detection models that can be used in addition to, or instead of our original, deterministic ‘OSD’ algorithm. The machine learning algorithm works well for our son’s tonic-clonic seizures (we had a live test of it a couple of days ago), but you might want to run the original ‘OSD’ algorithm in parallel with it for a while to see if it works for your seizures too.
There is a very simple default user interface to avoid overwhelming new users with too many graphs. The normal user interface with graphs etc. can be enabled in the settings.
The ‘advanced’ user interface has been modernised to make it compliant with modern user interface expectations (I think!).
There are a lot of ‘behind the scenes’ improvements to the code structure to make it easier to maintain (thanks to contributor AroonPro for a lot of these suggestions).
It is possible that you might need to do a fresh install for this version, so please record the settings that you are using before doing the upgrade, so you can restore them if you need to.
When you do the upgrade, please provide feedback, in particular on the following:
Did the upgrade work ok, or did you have to re-install the app (this is useful to know for when we roll it out to all users).
Does the installation wizard make sense to you? Should we change anything in it?
Would you like any changes to the user interface (either the basic or the normal one)? Now is a good time for this feedback because we can easily do a few tweaks before we roll it out to everyone.
How do you find the new Machine Learning (ML) algorithm? Does it give fewer false alarms? Is it detecting your seizures ok?
Have any features that you used to use stopped working?
If you have any questions, please contact graham@openseizuredetector.org.uk.
