seiz_eeg.clips

Functions to extract clips from seizure datasets

seiz_eeg.clips.make_clips(segments_df: DataFrame[ClipsDF], clip_length: float, clip_stride: float | str | tuple, overlap_action: str = 'ignore', sort_index: bool = False) → DataFrame[ClipsDF]

Split annotations dataframe in dataframe of clips

Parameters:

• segments_df (DataFrame[ClipsDF]) – Dataframe containing annotations for EEG segments

• clip_length (float) – Lenght of the output clips, in same unit as start_time and end_time of segments_df. A negative value returns the segments unchanged, but sort the dataset by index.

• clip_stride (Union[float, str, tuple]) –

  Stride to extract the start times of the clips. Integer or real values give explicit stride. If string or tuple, must be one of the following, with respective parameters.: - start: extract one clip per segment, starting at onset/termination label. - pre-ictal: for each onset time extract the beginning of ictal

  segment and the preictal clip ending clip_lenght, or specified sec before onset time. Only pairs of pre-ictal/ictal clips are returned

  • random: For each segment, extact a random clip of desired length.

• overlap_action (str) –

  What to do with clips overlapping segments. Options: - ignore: do not include any crossing clips - left: the label of crossing clips is given by the left

  (preceding) segment

  • right: the label of crossing clips is given by the right

    (ending) segment

  • seizure: the label of crossing clips is given by the ictal

  segment. (No more than two segments should be crossing) - bkgd: set the label of crossing clips to be 0

Raises:

• ValueError – If clip_stride is negative, or an invalid string

• ValueError – If overlap_action is an invalid string

Returns:

Clips dataframe

Return type:

DataFrame[ClipsDF]