mne.Annotations#

class mne.Annotations(onset, duration, description, orig_time=None, ch_names=None)[source]#

Annotation object for annotating segments of raw data.

Note

To convert events to Annotations, use annotations_from_events. To convert existing Annotations to events, use events_from_annotations.

Parameters:

onsetarray of float, shape (n_annotations,)

The starting time of annotations in seconds after orig_time.

durationarray of float, shape (n_annotations,) | float

Durations of the annotations in seconds. If a float, all the annotations are given the same duration.

descriptionarray of str, shape (n_annotations,) | str

Array of strings containing description for each annotation. If a string, all the annotations are given the same description. To reject epochs, use description starting with keyword ‘bad’. See example above.

orig_timefloat | str | datetime | tuple of int | None

A POSIX Timestamp, datetime or a tuple containing the timestamp as the first element and microseconds as the second element. Determines the starting time of annotation acquisition. If None (default), starting time is determined from beginning of raw data acquisition. In general, raw.info['meas_date'] (or None) can be used for syncing the annotations with raw data if their acquisition is started at the same time. If it is a string, it should conform to the ISO8601 format. More precisely to this ‘%Y-%m-%d %H:%M:%S.%f’ particular case of the ISO8601 format where the delimiter between date and time is ‘ ‘.

ch_nameslist | None

List of lists of channel names associated with the annotations. Empty entries are assumed to be associated with no specific channel, i.e., with all channels or with the time slice itself. None (default) is the same as passing all empty lists. For example, this creates three annotations, associating the first with the time interval itself, the second with two channels, and the third with a single channel:

Annotations(onset=[0, 3, 10], duration=[1, 0.25, 0.5],
            description=['Start', 'BAD_flux', 'BAD_noise'],
            ch_names=[[], ['MEG0111', 'MEG2563'], ['MEG1443']])

New in v0.23.

Attributes:

orig_time: The time base of the Annotations.

Methods

`__add__`(other)	Add (concatencate) two Annotation objects.
`__getitem__`(key, *[, with_ch_names])	Propagate indexing and slicing to the underlying numpy structure.
`__iter__`()	Iterate over the annotations.
`__len__`()	Return the number of annotations.
`append`(onset, duration, description[, ch_names])	Add an annotated segment.
`copy`()	Return a copy of the Annotations.
`count`()	Count annotations.
`crop`([tmin, tmax, emit_warning, ...])	Remove all annotation that are outside of [tmin, tmax].
`delete`(idx)	Remove an annotation.
`rename`(mapping[, verbose])	Rename annotation description(s).
`save`(fname, *[, overwrite, verbose])	Save annotations to FIF, CSV or TXT.
`set_durations`(mapping[, verbose])	Set annotation duration(s).
`to_data_frame`([time_format])	Export annotations in tabular structure as a pandas DataFrame.

See also

mne.annotations_from_events
mne.events_from_annotations

Notes

Annotations are added to instance of mne.io.Raw as the attribute raw.annotations.

To reject bad epochs using annotations, use annotation description starting with ‘bad’ keyword. The epochs with overlapping bad segments are then rejected automatically by default.

To remove epochs with blinks you can do:

>>> eog_events = mne.preprocessing.find_eog_events(raw)  
>>> n_blinks = len(eog_events)  
>>> onset = eog_events[:, 0] / raw.info['sfreq'] - 0.25  
>>> duration = np.repeat(0.5, n_blinks)  
>>> description = ['bad blink'] * n_blinks  
>>> annotations = mne.Annotations(onset, duration, description)  
>>> raw.set_annotations(annotations)  
>>> epochs = mne.Epochs(raw, events, event_id, tmin, tmax)  

ch_names

Specifying channel names allows the creation of channel-specific annotations. Once the annotations are assigned to a raw instance with mne.io.Raw.set_annotations(), if channels are renamed by the raw instance, the annotation channels also get renamed. If channels are dropped from the raw instance, any channel-specific annotation that has no channels left in the raw instance will also be removed.

orig_time

If orig_time is None, the annotations are synced to the start of the data (0 seconds). Otherwise the annotations are synced to sample 0 and raw.first_samp is taken into account the same way as with events.

When setting annotations, the following alignments between raw.info['meas_date'] and annotation.orig_time take place:

----------- meas_date=XX, orig_time=YY -----------------------------

     |              +------------------+
     |______________|     RAW          |
     |              |                  |
     |              +------------------+
 meas_date      first_samp
     .
     .         |         +------+
     .         |_________| ANOT |
     .         |         |      |
     .         |         +------+
     .     orig_time   onset[0]
     .
     |                   +------+
     |___________________|      |
     |                   |      |
     |                   +------+
 orig_time            onset[0]'

----------- meas_date=XX, orig_time=None ---------------------------

     |              +------------------+
     |______________|     RAW          |
     |              |                  |
     |              +------------------+
     .              N         +------+
     .              o_________| ANOT |
     .              n         |      |
     .              e         +------+
     .
     |                        +------+
     |________________________|      |
     |                        |      |
     |                        +------+
 orig_time                 onset[0]'

----------- meas_date=None, orig_time=YY ---------------------------

     N              +------------------+
     o______________|     RAW          |
     n              |                  |
     e              +------------------+
               |         +------+
               |_________| ANOT |
               |         |      |
               |         +------+

            [[[ CRASH ]]]

----------- meas_date=None, orig_time=None -------------------------

     N              +------------------+
     o______________|     RAW          |
     n              |                  |
     e              +------------------+
     .              N         +------+
     .              o_________| ANOT |
     .              n         |      |
     .              e         +------+
     .
     N                        +------+
     o________________________|      |
     n                        |      |
     e                        +------+
 orig_time                 onset[0]'

Warning

This means that when raw.info['meas_date'] is None, doing raw.set_annotations(raw.annotations) will not alter raw if and only if raw.first_samp == 0. When it’s non-zero, raw.set_annotations will assume that the “new” annotations refer to the original data (with first_samp==0), and will be re-referenced to the new time offset!

Specific annotation

BAD_ACQ_SKIP annotation leads to specific reading/writing file behaviours. See mne.io.read_raw_fif() and Raw.save() notes for details.

__add__(other)[source]#: Add (concatencate) two Annotation objects.

__getitem__(key, *, with_ch_names=None)[source]#: Propagate indexing and slicing to the underlying numpy structure.

__iter__()[source]#: Iterate over the annotations.

__len__()[source]#

Return the number of annotations.

Returns:

n_annotint: The number of annotations.

append(onset, duration, description, ch_names=None)[source]#

Add an annotated segment. Operates inplace.

Parameters:

onsetfloat | array_like

Annotation time onset from the beginning of the recording in seconds.

durationfloat | array_like

Duration of the annotation in seconds.

descriptionstr | array_like

Description for the annotation. To reject epochs, use description starting with keyword ‘bad’.

ch_nameslist | None

Annotations(onset=[0, 3, 10], duration=[1, 0.25, 0.5],
            description=['Start', 'BAD_flux', 'BAD_noise'],
            ch_names=[[], ['MEG0111', 'MEG2563'], ['MEG1443']])

New in v0.23.

Returns:

selfmne.Annotations: The modified Annotations object.

Notes

The array-like support for arguments allows this to be used similarly to not only list.append, but also list.extend.

Examples using append:

Visualise NIRS artifact correction methods

copy()[source]#

Return a copy of the Annotations.

Returns:

instinstance of Annotations: A copy of the object.

count()[source]#

Count annotations.

Returns:

countsdict: A dictionary containing unique annotation descriptions as keys with their counts as values.

crop(tmin=None, tmax=None, emit_warning=False, use_orig_time=True, verbose=None)[source]#

Remove all annotation that are outside of [tmin, tmax].

The method operates inplace.

Parameters:

tminfloat | datetime | None: Start time of selection in seconds.
tmaxfloat | datetime | None: End time of selection in seconds.
emit_warningbool: Whether to emit warnings when limiting or omitting annotations. Defaults to False.
use_orig_timebool: Whether to use orig_time as an offset. Defaults to True.
verbosebool | str | int | None: Control verbosity of the logging output. If None, use the default verbosity level. See the logging documentation and mne.verbose() for details. Should only be passed as a keyword argument.

Returns:

selfinstance of Annotations: The cropped Annotations object.

Examples using crop:

Sleep stage classification from polysomnography (PSG) data

delete(idx)[source]#

Remove an annotation. Operates inplace.

Parameters:

idxint | array_like of int: Index of the annotation to remove. Can be array-like to remove multiple indices.

property orig_time#: The time base of the Annotations.

rename(mapping, verbose=None)[source]#

Rename annotation description(s). Operates inplace.

Parameters:

mappingdict: A dictionary mapping the old description to a new description, e.g. {‘1.0’ : ‘Control’, ‘2.0’ : ‘Stimulus’}.
verbosebool | str | int | None: Control verbosity of the logging output. If None, use the default verbosity level. See the logging documentation and mne.verbose() for details. Should only be passed as a keyword argument.

Returns:

selfmne.Annotations: The modified Annotations object.

Notes

New in v0.24.0.

save(fname, *, overwrite=False, verbose=None)[source]#

Save annotations to FIF, CSV or TXT.

Typically annotations get saved in the FIF file for raw data (e.g., as raw.annotations), but this offers the possibility to also save them to disk separately in different file formats which are easier to share between packages.

Parameters:

fnamepath-like: The filename to use.
overwritebool: If True (default False), overwrite the destination file if it exists.

New in v0.23.
verbosebool | str | int | None: Control verbosity of the logging output. If None, use the default verbosity level. See the logging documentation and mne.verbose() for details. Should only be passed as a keyword argument.

Notes

The format of the information stored in the saved annotation objects depends on the chosen file format. .csv files store the onset as timestamps (e.g., 2002-12-03 19:01:56.676071), whereas .txt files store onset as seconds since start of the recording (e.g., 45.95597082905339).

Examples using save:

Annotating continuous data

set_durations(mapping, verbose=None)[source]#

Set annotation duration(s). Operates inplace.

Parameters:

mappingdict | float: A dictionary mapping the annotation description to a duration in seconds e.g. {'ShortStimulus' : 3, 'LongStimulus' : 12}. Alternatively, if a number is provided, then all annotations durations are set to the single provided value.
verbosebool | str | int | None: Control verbosity of the logging output. If None, use the default verbosity level. See the logging documentation and mne.verbose() for details. Should only be passed as a keyword argument.

Returns:

selfmne.Annotations: The modified Annotations object.

Notes

New in v0.24.0.

to_data_frame(time_format='datetime')[source]#

Export annotations in tabular structure as a pandas DataFrame.

Parameters:

time_formatstr | None: Desired time format. If None, no conversion is applied, and time values remain as float values in seconds. If 'ms', time values will be rounded to the nearest millisecond and converted to integers. If 'timedelta', time values will be converted to pandas.Timedelta values. If 'datetime', time values will be converted to pandas.Timestamp values, relative to raw.info['meas_date'] and offset by raw.first_samp. Default is None.

New in v1.7.

Returns:

resultpandas.DataFrame: Returns a pandas DataFrame with onset, duration, and description columns. A column named ch_names is added if any annotations are channel-specific.

Examples using `mne.Annotations`#

Parsing events from raw data

Working with CTF data: the Brainstorm auditory dataset

Importing Data from Eyetracking devices

Annotating continuous data

Rejecting bad data spans and breaks

Working with eye tracker data in MNE-Python

Auto-generating Epochs metadata

Sleep stage classification from polysomnography (PSG) data

Automated epochs metadata generation with variable time windows

Visualise NIRS artifact correction methods

Annotate movement artifacts and reestimate dev_head_t

Annotate muscle artifacts

mne.Annotations#

Examples using mne.Annotations#

Examples using `mne.Annotations`#