Author: graham

Thanks to those of you who have helped with testing the draft versions of OpenSeizureDetector V4, I think it is stable enough to release to more users (A particular thank you to Jeff Parker who has done a lot of methodical testing and helped find quite a few issues which we have fixed).

The big change in this version is the introduction of the ‘Data Sharing’ feature to allow users to mark false alarms and genuine seizures, and share the data with developers / researchers to help us make better detection algorithms for future versions of OpenSeizureDetector. The main changes are:

• Introduction of Data Sharing feature, with associated user interface for reporting seizures and classifying them as real or false alarms.
• Improved the reliability of sending location in the SMS alarm messages – it should now work on the newer android versions – there are more screens asking for permission to access features to achieve this, which is a requirement of publishing it on Play Store.
• Other general reliability improvements – there should be far fewer issues with the system not shutting down properly with the new version.
• Fixed a calculation issue with the detection algorithm – it is now more sensitive so you may need a higher value of the Alarm Ratio Threshold setting – the default has been increased to 57 to account for this.

OpenSeizureDetector V4 has now been approved for release on Google Play Store (that took a lot more doing than expected!), so I intend to release it for ‘Beta’ testing sometime this weekend (once I have written some user instructions….).

This means it is available on Google Play Store, but you need to actively sign up for the new version if you would like to try it. You do this by:

• Using this link on a web browser: https://play.google.com/apps/testing/uk.org.openseizuredetector
• Or this link on your Android Device: https://play.google.com/store/apps/details?id=uk.org.openseizuredetector

It would be good if a good number of existing users try out this new version now and try signing up to the data sharing system so you can feed back how it works for you before I make it available to new users who may not know what the system does.

I will post again when I press the button to make the Beta test version live, then everyone who has signed up should get the option to upgrade via play store.

Graham (graham@openseizuredetector.org.uk)

Thank you to the OpenSeizureDetector users who have been testing the new Data Sharing system – your feedback has been very useful!

One issue that was identified is that the user registration and password reset system, which relies on emails from the database server was not working very well because emails were often not being delivered.

I think I have fixed that by using the openeizuredetector.org.uk email server instead. I would appreciate it if testers would try creating a new account or updating their passwords at https://osdapi.ddns.net and see if the emails arrive ok now.

My plan is to release Version 4 of OpenSeizureDetector on Play Store to those who have registered as Beta Testers by the weekend – I just want to check the documentation to make sure that users who do not read these posts are pointed in the right direction.

Thanks,

Graham (graham@openseizuredetector.org.uk)

I have updated the Privacy Policy to be more explicit about it accessing your location to include in SMS alerts.

This is to help with getting Version 4 of the app released on Google Play Store as we now have to be more explicit regarding use of location data.

Graham (graham@openseizuredetector.org.uk)

Thank you to the users who have been trying out the draft Version 4 of OpenSeizureDetector and feeding back the issues they have found.

I think I have addressed all of the ‘Must Do’ ones that needed to be fixed to release the new version more widely, so I am now going through the Google review process to try to get it published on Google Play Store. It will initially appear for users who have registered as “Beta Testers” on play store – According to Google there is over 60 of you, so if you do not want to upgrade to this version yet, you might want to unsubscribe as a Beta tester. I’d be grateful if as many of you as possible will try it though and feed back any issues.

The main new thing following Google’s review is that the permission granting to allow SMS location alerts, which requires some ‘dangerous’ permissions (sending SMS messages without user intervention and including the users’s location in the message) is now more obvious with a dialog box explaining why we need the permission before you see the screen asking you to grant it. I also think I have fixed the issue that some users reported that for Android 11 the SMS message did not contain the location. The new permissions screens are shown here:

One outstanding issue is that the sign-up for the Data Sharing database uses email to verify that you really want to create an account. The emails get through to gmail ok (but end up in the spam folder), but other email providers are not delivering the email at all. I can activate accounts manually if you try to set one up but do not receive the email – just drop me a note to graham@openseizuredetector.org.uk.

I have three options for fixing this email issue:

• Switch off the email account verification – I could try this for a short period and see if we get overwhelmed with spam accounts which I would have to deactivate.
• Upgrade the database server to have its own dedicated domain name (I could do this, but I am not 100% confident it will fix it….and it costs more money).
• Switch over to using Google Firebase database – there is a bit of work to convert the app to use this, but I am a bit nervous because it is a ‘pay as you go’ service and it might end up costing more than I want to spend on this project – I’ll have to see how much use the existing database gets so I can see what the cost of using the Google service would be.

I am open to suggestions if anyone has other ideas for this.

I will let you all know when I release it to Beta testers on Play store, which will allow some of you to start using the system.

Graham (graham@openseizuredetector.org.uk)

I have updated the OpenSeizureDetector Privacy Policy to Version 2.0 (February 2022). There are no significant changes to the draft that I published a few weeks ago, because I did not receive any adverse comments on it.

The purpose of the change is to recognise the new Data Sharing feature that will be included in V4.0 of OpenSeizureDetector.

The new Privacy policy is here: OpenSeizureDetector Privacy Policy V2.0 (February 2022). The Google Play Store listing has also been updated to refer to this new policy.
Please let me know if you have any concerns over this.

Graham (graham@openseizuredetector.org.uk)

I am having problems with the account creation part of the data sharing system – users who have tested it have found they can create an account but the email that the system generates for them to verify the account creation is not getting through. I am looking at options of how to fix this, and have listed four so far here: https://github.com/OpenSeizureDetector/webApi/issues/30

If anyone with more experience of server and database management than me has any good ideas, please let me know!

Thanks, Graham (graham@openseizuredetector.org.uk)

Short Version:

We have used the Data Sharing feature to analyse why OpenSeizureDetector did not detect a particular seizure event, and have found a solution…. so it is useful :).

Long Version:

One morning last week Benjamin had a sort of partial seizure – it was not a full blown tonic-clonic seizure – he stayed sat up but his arms were a bit shaky and making a slow flapping motion with his hand too. The addition of the slow movement on top of the shaking meant it was going to be difficult for OpenSeizureDetector to pick it up (because it is tuned to detect shaking), but I was disappointed that it did not even give a warning.

Fortunately I had upgraded our ‘production’ system to use the draft Version 4 of OSD with data sharing enabled, so when we noticed the seizure I pressed the ‘raise alarm’ button on the phone screen to record a seizure event manually. The data was saved to the database and I could mark it as a genuine seizure and at a short description.

We used a suggestion by Jamie Pordoy who is working on the system with me to take the saved data and feed it back into OpenSeizureDetector so I could see what it was doing during the seizure.

Running it through OpenSeizureDetector again meant I could see that during the seizure we did have several periods when the ‘spectrum ratio’ exceeded the set threshold of 50….but for some reason it did not alarm….this was wrong….

I delved through the code and realised that when I had converted the code from the Pebble Version to the Garmin one (or maybe when I tidied it up later)…I put a bracket in the wrong place. This meant that although we can specify the spectrum ratio threshold as single units (51, 52, 55 etc.) the way it was working was that if you set a threshold between 50 and 59, it would only alarm if the actual value exceeded 60. This is fine for full tonic-clonic seizures where I expect the ratio to be over 60, but not so good for less severe ones.

I have put a fix for this in Version 4 of OSD. This means you may initially see more false alarms with Version 4 than you do Version 3, so you may need to increase your Alarm Ratio Threshold a bit to get the same false alarm and detection performance.

It is good to start to see some use for all this ‘Boring Software Infrastructure’!

An example of the analysis of the stored seizure that we can now do easily is shown below

Graham.

The bottom graph shows the problem – the blue line is the ROI ratio with the horizontal line showing the set threshold for alarm. but the only alarm event was the manual alarm that I generated by pressing the ‘raise alarm’ button on the phone. I would have expected it to alarm between 0 and 25 seconds on the graph.

Thank you to those users who have offered to test the draft version 4 of OpenSeizureDetector. I have added some instructions on testing it to the web site here: https://www.openseizuredetector.org.uk/?page_id=1842If anyone else would like to try it, please do, and feed back your experiences (good and bad!). Graham Jones (graham@openseizuredetector.org.uk)

OpenSeizureDetector V4 is about ready for some limited user testing to check it is stable before I release it more widely.

The main change in V4 is the introduction of ‘Data Sharing’ where users (optionally) have the ability to share their data to a central database to allow us to build a large data set of normal activities that appear to be ‘seizure like’ and may cause false alarms, and most importantly real seizures for comparison.

It also introduces a some smaller changes:

• check on whether the system is going to ‘optimise’ OpenSeizureDetector’s battery usage (by shutting it down!), and prompts the user to disable this feature to make sure OpenSeizureDetector works correclty.
• Adds a ‘Report Seizure’ button to record a seizure even if OpenSeizureDetector does not alarm automatically.

So from the user’s perspective the changes are not radical, but there is a lot of extra code to run the user interface for the data sharing (logging on, displaying data, editing), so I’d like a few people to make sure it is stable for them and does not crash before I publish it.

My idea is that as long as that testing is successful, then in a couple of weeks I will publish Version 4 on Google Play Store.

For this initial testing what I need users to do is:

• Run through the registration, user verification and log-in process to the data sharing database.
• Generate some events by doing seizure-like activities and using the ‘report seizure’ function.
• Edit the events to mark them as either genuine seizures or false alarms.
• Check that the system does not crash unexpectedly.

Please drop me an email if you are able to help with this initial testing.

Thanks
Graham (graham@openseizuredetector.org.uk)

A couple of people have mentioned that they do not get notifications when I update this web site….I finally realised that is because there were two lists set up…and the list that gets notifications only contained one person…Me! I think I have fixed it now, so from now on, all subscribers should receive updates.

Sorry for the inconvenience!

Graham (graham@openseizuredetector.org.uk)