In https://github.com/dutchconnectomelab/CATO/pull/28 we updated the bandpass filter to make it less susceptible to edge artifacts. Although no new issues have been observed, the performance of any filter will depend on the situation (the properties of both the timeseries and the filter) and it is difficult to assess this in general. To avoid potential issues in the future, we can additionally truncate the timeseries.
This PR piggybacks onto the current scrubbing implementation to additionally remove frames at the beginning and end of the timeseries.
The proposed default of 5 is based on the length I've mostly seen for filters (we would like to remove as little as safely possible).
In https://github.com/dutchconnectomelab/CATO/pull/28 we updated the bandpass filter to make it less susceptible to edge artifacts. Although no new issues have been observed, the performance of any filter will depend on the situation (the properties of both the timeseries and the filter) and it is difficult to assess this in general. To avoid potential issues in the future, we can additionally truncate the timeseries.
This PR piggybacks onto the current scrubbing implementation to additionally remove frames at the beginning and end of the timeseries.
The proposed default of
5is based on the length I've mostly seen for filters (we would like to remove as little as safely possible).