ignore wrong triggers from data collection

💬 Support & Discussions
,
1

Hi,

I have a problem to find my correct events, because during data collection, the Biosemi system sent wrong extra triggers for some reasons, please see the screenshot. My correct triggers were sent correctly in the first line, but it kept sending 202020 triggers in the second line.

202020
2020201920×1080 289 KB

In ActiRead902, it shows this, the red line here is the wrong 202020 triggers.

image
image3101×917 212 KB

During preprocessing, I cannot find my correct triggers due to this error. By using find_events, it can find events, but then I want to print the events, it gives “You have 16 events shorter than the shortest_event…”.

I heard this error can be ignored in eeglab, can we do this in MNE? Could you please help me ignore the triggers in the second line. Thank you very much!

  • operating system: Windows 10

Thank you again!

Yunr

2

You can extract all events with events = mne.find_events(raw) and then subset only those you actually need (events is a NumPy array).

3

Thank you for your reply~

I did this, then I tried to print the events to check, but I cannot print it, it gives error: “You have 16 events shorter than the shortest_event…”

Do you know how to deal with this?

4

Can you share a minimal example what you did?

5

Thank you! Here is the code I am using to find events:

events = mne.find_events(raw, shortest_event=1)
events[:200]

image
image701×1030 68.3 KB

I don’t have a 512 event, instead I have 51. I guess the 202020 influences the 51 trigger.

raw.copy().pick_types(stim=True).plot() # try to find the STIM channels
events = mne.find_events(raw)
print(events[:20])

“ValueError: You have 16 events shorter than the shortest_event. These are very unusual and you may want to set min_duration to a larger value.”

If I set min_duration=0.140 which is the minimum duration between my correct triggers, it returns no events (“array(, shape=(0, 3), dtype=int32)”).

6

This error originates from your events having a duration of less than 2 samples. You can change that with shortest_event. However, I think you need to take a look at other parameters as well to tailor it to your needs: mne.find_events — MNE 1.3.1 documentation

7

Thank you for your suggestion!

I have tried different parameters and combinations. But it looks like MNE merges the 2 lines of triggers and the 202020 contaminates the correct ones. I don’t have trigger named 512, and I barely see my correct triggers, so it seems that it doesn’t matter how do I select events, no correct ones would show up.

image
image3820×733 460 KB

8

Are there two separate trigger channels or just one? Can you not manually filter the events you want? If they stack (i.e. add up when occurring at the same time), maybe you can still find a criterion which tells you which events to keep.

9

Hi Yunr,

The way to fix your problem is to do:

find_events(…, min_duration=1. / raw.info[‘sfreq’])

Hope this helps,
Mainak

10

Hi Mainak,

Thank you very much for the help~ I tried it, but still got error “You have 16 events shorter than the shortest_event. These are very unusual and you may want to set min_duration to a larger value e.g. x / raw.info[‘sfreq’]…”.

I tried to set min_duration=2, it can run, but returns event ‘512’ which is the wrong event. I guess the system may merge events somehow.

Thank you again!
Yunr

11

if there are 2 separate stim channels, and one of them has the correct events and the other one is garbage, then try passing mne.find_events(raw, stim_channel='name_of_the_good_stim_channel')

12

Thank you very much for the suggestion!

I found another solution. In case other people would have the similar problem. I put it here:

We can select or remove the extra triggers with the “Muter” tool in ActiTools902 ( http://www.biosemi.com/download/ToolsFo … 02-Win.zip ). It will generate a new BDF file without the wrong triggers.

14

The BioSemi system uses the lower 8 bits for actual trigger codes and the upper 8 bits for status updates about the system. This is what is generating these weirdly high event codes. In my opinion (yes, let’s add yet another opinion to this thread :wink: ) is to use the mask parameter of the find_events function to have it only pay attention to the bottom 8 bits of the status channel:

events = mne.find_events(raw, mask=0b11111111)  # that's 8 ones
1 Like