When listing the alarms to qualify, too many results

[pmithrandir]

Hello,

On the data sharing page, the alarm listed are too numerous.

For the same seizure, I'll get one line every 5-10 sec... and for each of them I have to go through the process of explaining that it is a seizure or not.

Is there a way to "group" event happenning around the same time ?

[jones139]

This is a good idea. When I create the anonymised version of the database for publication I group the data into 3 minute periods and use the metadata from the most 'interesting' looking event in that period (either the one that generated a full alarm rather than a warning, or the one with a text description of the event etc.).
Doing something similar on the app would reduce the amount of data the user is presented with, which would help.
One to add to the enhancements list, thanks!

[pmithrandir]

Hi,

Do we have a measure on how many seizure get actually qualified by people ?

The android app is so painful(too many events, sometime multiple occurance...) that I don't even take time to fill them. Usually, I have between 20 and 50 occurrences to fill per day if my son wear the watch, 10 to 20 if it's me.
80% being false alarm, multiplied by 5-10 events per alarm / warning

My understanding is that we want to make the algorithm progress, and we can't do that if we cant' say which data are pointing a seizure and which one is not.

We could either

• change the API to "group" alerts to be enriched in data sharing view
• change the app to be able to multi select elements and update them in bulk (bulk must exist as ther is 2 options in the menu)

I'm sadly not qualified with android dev. I could work on the python api if needed.

[jones139]

20-50 alarms is excessive - I would say it is not working for you - I'll have a look at some of your data and your settings and see if we can adjust something to reduce the number of false alarms.

I do the grouping in the offline processing to produce the Open Seizure Database, so the simplest would be to improve the data sharing page in the app - I'll have a look and see what i can do with that - either simply allow multiple selections, or maybe group them like I do in the database, which would be better.

[pmithrandir]

20 50 is because each one count 10.For real on a normal day it was about 2 to 5 false alarm. Which is not great but not blocking.My usage is now mostly by night or when he is sleeping to make sure I'll be awake in a seizure happen. I realize that during day time he is rarely alone in the end. During night false alarm rate is much lower.PierreSent from Android deviceLe 8 avr. 2024 15:57, Graham Jones ***@***.***> a écrit : 20-50 alarms is excessive - I would say it is not working for you - I'll have a look at some of your data and your settings and see if we can adjust something to reduce the number of false alarms. I do the grouping in the offline processing to produce the Open Seizure Database, so the simplest would be to improve the data sharing page in the app - I'll have a look and see what i can do with that - either simply allow multiple selections, or maybe group them like I do in the database, which would be better. —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: ***@***.***>

[jones139]

ah, ok, so if we group them into 'events' it should not be so bad - I'll try that.