rbroc / reading_brain_prep Goto Github PK

At each run, participants answer a post-experiment multiple choice questionnaire after the self-paced reading task.
The duration of BOLD time series shared on OpenNeuro suggests that fMRI acquisition also covers that.
We should discuss what to do (truncate BOLD, or how to code for the questionnaire events in the events files...?)

• Currently, consecutive fixations are aggregated if the gap between them is < .400 ms (TR)
• In the original files, there are fixations extending beyond the button press (which should trigger the end of the trial and transition to next sentence) - we're currently keeping them as they are (i.e., not capping durations)
• button press is inferred from the value of TRUEOffset in raw events file (seems to make sense, but keeping a note of this just in case)
• button presses are given fixed duration of 100ms
• there are no explicit events for fixation crosses in the original event files, so no events for those in processed files
• in the processed file, sentence is a binary regressor (but we may consider storing information on the full sentence in case these are later supported in Neuroscout)
• we could also consider adding / replacing sentence regressor coding for the offset of a sentence, or the onset of a sentence as impulse events

I was looking at the agg_ transcripts and just found a couple issues:

• The last two items in this file: https://github.com/rbroc/reading_brain_prep/blob/master/processed_data/sub-01/func/agg_sub-01_task-read_run-1_events.tsv (and maybe other files) have an onset of 0. Any idea why?
• There are a few numbers as words, any idea why? Can we strip these?
• There are also a few blank lines, can we also strip these?

Otherwise, looking good and I think we can try to ingest soon.

Going forward, I think we should host a forked version of the dataset on Neuroscout and on github (w/ datalad), to keep track of the changes we've made. Probably there will be other BIDS non-complicance issues when I try to ingest.