mne_hfo.match_detected_annotations¶
-
mne_hfo.
match_detected_annotations
(ytrue_annot_df, ypred_annot_df, ch_names=None, label=None, sec_margin=1.0, method='match-true')[source]¶ Given two annotations.tsv DataFrames, match HFO detection overlaps.
- Parameters
ytrue_annot_df : pd.DataFrame
The reference annotations DataFrame containing the HFO events that are considered “ground-truth” in this comparison.
ypred_annot_df : pd.DataFrame
The estimated annotations DataFrame containing the HFO events that are estimated using a
Detector
.ch_names : list | str | None
Which channels to match. If None (default), then will match all available channels in both dataframes. If str, then must be a single channel name available in the
ytrue_annot_df
. If list of strings, then must be a list of channel names available in theytrue_annot_df
.label : str | None
The HFO label to use. If None (default) will consider all rows in both input DataFrames as an HFO event. If a string, then it must match to an element of
label
column in the dataframes.sec_margin : float
Number of seconds to consider a valid checking window. Default = 1.
method : str
Type of strategy for matching HFO events. Must be one of
match-true
,match-pred
, ormatch-total
. If “match-true”, will return a dataframe of all true indices and matching predicted indices if they exist. If “match-pred”, will return a dataframe of all predicted indices and matching true indices if they exist. If “match-total”, will return the concatenation of the two. See Notes for more information.- Returns
matched_df : pd.DataFrame
A DataFrame with the columns
pred_index
andtrue_index
, which corresponds to indices,