Break detection
find_breaks
module-attribute
¶
find_breaks: bool = False
During an experimental run, the recording might be interrupted by breaks of various durations, e.g. to allow the participant to stretch, blink, and swallow freely. During these periods, large-scale artifacts are often picked up by the recording system. These artifacts can impair certain stages of processing, e.g. the peak-detection algorithms we use to find EOG and ECG activity. In some cases, even the bad channel detection algorithms might not function optimally. It is therefore advisable to mark such break periods for exclusion at early processing stages.
If True
, try to mark breaks by finding segments of the data where no
experimental events have occurred. This will then add annotations with the
description BAD_break
to the continuous data, causing these segments to be
ignored in all following processing steps.
Example
Automatically find break periods, and annotate them as BAD_break
.
find_breaks = True
Disable break detection.
find_breaks = False
min_break_duration
module-attribute
¶
min_break_duration: float = 15.0
The minimal duration (in seconds) of a data segment without any experimental
events for it to be considered a "break". Note that the minimal duration of the
generated BAD_break
annotation will typically be smaller than this, as by
default, the annotation will not extend across the entire break.
See t_break_annot_start_after_previous_event
and t_break_annot_stop_before_next_event
to control this behavior.
Example
Periods between two consecutive experimental events must span at least
15
seconds for this period to be considered a "break".
min_break_duration = 15.
t_break_annot_start_after_previous_event
module-attribute
¶
t_break_annot_start_after_previous_event: float = 5.0
Once a break of at least
min_break_duration
seconds has been discovered, we generate a BAD_break
annotation that does not
necessarily span the entire break period. Instead, you will typically want to
start it some time after the last event before the break period, as to not
unnecessarily discard brain activity immediately following that event.
This parameter controls how much time (in seconds) should pass after the last pre-break event before we start annotating the following segment of the break period as bad.
Example
Once a break period has been detected, add a BAD_break
annotation to it,
starting 5
seconds after the latest pre-break event.
t_break_annot_start_after_previous_event = 5.
Start the BAD_break
annotation immediately after the last pre-break
event.
t_break_annot_start_after_previous_event = 0.
t_break_annot_stop_before_next_event
module-attribute
¶
t_break_annot_stop_before_next_event: float = 5.0
Similarly to how
t_break_annot_start_after_previous_event
controls the "gap" between beginning of the break period and BAD_break
annotation onset, this parameter controls how far the annotation should extend
toward the first experimental event immediately following the break period
(in seconds). This can help not to waste a post-break trial by marking its
pre-stimulus period as bad.
Example
Once a break period has been detected, add a BAD_break
annotation to it,
starting 5
seconds after the latest pre-break event.
t_break_annot_start_after_previous_event = 5.
Start the BAD_break
annotation immediately after the last pre-break
event.
t_break_annot_start_after_previous_event = 0.