Add event metadata handling (#1285)

* add failing test * code works but should not * we have a failing test * do not use trial type in tsv * my tests are green * make the linter happy * add docstrings * fix docs * add first time stuff --------- Co-authored-by: Stefan Appelhoff <[email protected]>
mne-tools · Aug 23, 2024 · 471bd0c · 471bd0c
1 parent f919962
commit 471bd0c
Show file tree

Hide file tree

Showing 6 changed files with 161 additions and 10 deletions.
diff --git a/CITATION.cff b/CITATION.cff
@@ -187,6 +187,10 @@ authors:
       family-names: Benitez
       affiliation: 'Magnetoencephalography Core, National Institutes of Health, Bethesda, Maryland, USA'
       orcid: 'https://orcid.org/0000-0001-6364-7272'
+    - given-names: Thomas
+      family-names: Hartmann
+      affiliation: 'Paris-Lodron-University Salzburg, Centre for Cogntitive Neuroscience, Department of Psychology, Salzburg, Austria'
+      orcid: 'https://orcid.org/0000-0002-8298-8125'
     - given-names: Alexandre
       family-names: Gramfort
       affiliation: 'Université Paris-Saclay, Inria, CEA, Palaiseau, France'

diff --git a/doc/authors.rst b/doc/authors.rst
@@ -48,3 +48,4 @@
 .. _Julius Welzel: https://github.com/JuliusWelzel
 .. _Kaare Mikkelsen: https://github.com/kaare-mikkelsen
 .. _Amaia Benitez: https://github.com/AmaiaBA
+.. _Thomas Hartmann: https://github.com/thht
diff --git a/doc/whats_new.rst b/doc/whats_new.rst
@@ -19,6 +19,7 @@ The following authors contributed for the first time. Thank you so much! 🤩
 
 * `Kaare Mikkelsen`_
 * `Amaia Benitez`_
+* `Thomas Hartmann`_
 
 The following authors had contributed before. Thank you for sticking around! 🤘
 
@@ -33,6 +34,7 @@ Detailed list of changes
 ^^^^^^^^^^^^^^^
 
 - :meth:`mne_bids.BIDSPath.match()` and :func:`mne_bids.find_matching_paths` now have additional parameters ``ignore_json`` and ``ignore_nosub``, to give users more control over which type of files are matched, by `Kaare Mikkelsen`_ (:gh:`1281`)
+- :func:`mne_bids.write_raw_bids()` can now handle event metadata as a pandas DataFrame, by `Thomas Hartmann`_ (:gh:`1285`)
 
 🧐 API and behavior changes
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^

diff --git a/mne_bids/read.py b/mne_bids/read.py
@@ -169,7 +169,7 @@ def _read_events(events, event_id, raw, bids_path=None):
 
     # If we have events, convert them to Annotations so they can be easily
     # merged with existing Annotations.
-    if events.size > 0:
+    if events.size > 0 and event_id is not None:
         ids_without_desc = set(events[:, 2]) - set(event_id.values())
         if ids_without_desc:
             raise ValueError(
@@ -200,6 +200,22 @@ def _read_events(events, event_id, raw, bids_path=None):
         raw.set_annotations(annotations)
         del id_to_desc_map, annotations, new_annotations
 
+    if events.size > 0 and event_id is None:
+        new_annotations = mne.annotations_from_events(
+            events=events,
+            sfreq=raw.info["sfreq"],
+            orig_time=raw.annotations.orig_time,
+        )
+
+        raw = raw.copy()  # Don't alter the original.
+        annotations = raw.annotations.copy()
+
+        # We use `+=` here because `Annotations.__iadd__()` does the right
+        # thing and also performs a sanity check on `Annotations.orig_time`.
+        annotations += new_annotations
+        raw.set_annotations(annotations)
+        del annotations, new_annotations
+
     # Now convert the Annotations to events.
     all_events, all_desc = events_from_annotations(
         raw,

diff --git a/mne_bids/tests/test_write.py b/mne_bids/tests/test_write.py
@@ -20,6 +20,7 @@
 
 import mne
 import numpy as np
+import pandas as pd
 import pytest
 from mne.datasets import testing
 from mne.io import anonymize_info
@@ -4113,3 +4114,68 @@ def test_write_neuromag122(_bids_validate, tmp_path):
     )
     write_raw_bids(raw, bids_path, overwrite=True, allow_preload=True, format="FIF")
     _bids_validate(bids_root)
+
+
+@testing.requires_testing_data
+def test_write_evt_metadata(_bids_validate, tmp_path):
+    """Test writing events and metadata to BIDS."""
+    bids_root = tmp_path / "bids"
+    raw_fname = data_path / "MEG" / "sample" / "sample_audvis_trunc_raw.fif"
+    raw = _read_raw_fif(raw_fname)
+    events = mne.find_events(raw, initial_event=True)
+    df_list = []
+    for idx, event in enumerate(events):
+        direction = None
+        if event[2] in (1, 3):
+            direction = "left"
+        elif event[2] in (2, 4):
+            direction = "right"
+
+        event_type = "button_press" if event[2] == 32 else "stimulus"
+        stimulus_kind = None
+        if event[2] == 5:
+            stimulus_kind = "smiley"
+        elif event[2] in (1, 2):
+            stimulus_kind = "auditory"
+        elif event[2] in (3, 4):
+            stimulus_kind = "visual"
+
+        df_list.append(
+            {
+                "direction": direction,
+                "event_type": event_type,
+                "stimulus_kind": stimulus_kind,
+            }
+        )
+
+    event_metadata = pd.DataFrame(df_list)
+
+    bids_path = _bids_path.copy().update(root=bids_root, datatype="meg")
+    write_raw_bids(
+        raw,
+        bids_path=bids_path,
+        events=events,
+        event_metadata=event_metadata,
+        overwrite=True,
+        extra_columns_descriptions={
+            "direction": "The direction of the stimulus",
+            "event_type": "The type of the event",
+            "stimulus_kind": "The stimulus modality",
+        },
+    )
+    _bids_validate(bids_root)
+    events_tsv_path = bids_path.copy().update(suffix="events", extension=".tsv")
+    events_json_path = events_tsv_path.copy().update(extension=".json")
+
+    assert events_tsv_path.fpath.exists()
+    assert events_json_path.fpath.exists()
+
+    events_json = json.loads(events_json_path.fpath.read_text())
+    events_tsv = _from_tsv(events_tsv_path)
+
+    assert "trial_type" not in events_tsv
+    assert "trial_type" not in events_json
+
+    for cur_col in event_metadata.columns:
+        assert cur_col in events_tsv
+        assert cur_col in events_json
diff --git a/mne_bids/write.py b/mne_bids/write.py
@@ -266,7 +266,9 @@ def _get_fid_coords(dig_points, raise_error=True):
     return fid_coords, coord_frame
 
 
-def _events_tsv(events, durations, raw, fname, trial_type, overwrite=False):
+def _events_tsv(
+    events, durations, raw, fname, trial_type, event_metadata=None, overwrite=False
+):
     """Create an events.tsv file and save it.
 
     This function will write the mandatory 'onset', and 'duration' columns as
@@ -290,6 +292,9 @@ def _events_tsv(events, durations, raw, fname, trial_type, overwrite=False):
     trial_type : dict | None
         Dictionary mapping a brief description key to an event id (value). For
         example {'Go': 1, 'No Go': 2}.
+    event_metadata : pandas.DataFrame | None
+        Additional metadata to be stored in the events.tsv file. Must have one
+        row per event.
     overwrite : bool
         Whether to overwrite the existing file.
         Defaults to False.
@@ -319,19 +324,30 @@ def _events_tsv(events, durations, raw, fname, trial_type, overwrite=False):
     else:
         del data["trial_type"]
 
+    if event_metadata is not None:
+        for key, values in event_metadata.items():
+            data[key] = values
+
     _write_tsv(fname, data, overwrite)
 
 
-def _events_json(fname, overwrite=False):
+def _events_json(fname, extra_columns=None, has_trial_type=True, overwrite=False):
     """Create participants.json for non-default columns in accompanying TSV.
 
     Parameters
     ----------
     fname : str | mne_bids.BIDSPath
         Output filename.
+    extra_columns : dict | None
+        Dictionary with additional columns to be added to the events.json file.
+    has_trial_type : bool
+        Whether the events.tsv file should contain a 'trial_type' column.
     overwrite : bool
         Whether to overwrite the output file if it exists.
     """
+    if extra_columns is None:
+        extra_columns = dict()
+
     new_data = {
         "onset": {
             "Description": (
@@ -361,9 +377,16 @@ def _events_json(fname, overwrite=False):
                 "associated with the event."
             )
         },
-        "trial_type": {"Description": "The type, category, or name of the event."},
     }
 
+    if has_trial_type:
+        new_data["trial_type"] = {
+            "Description": "The type, category, or name of the event."
+        }
+
+    for key, value in extra_columns.items():
+        new_data[key] = {"Description": value}
+
     # make sure to append any JSON fields added by the user
     fname = Path(fname)
     if fname.exists():
@@ -1378,6 +1401,8 @@ def write_raw_bids(
     bids_path,
     events=None,
     event_id=None,
+    event_metadata=None,
+    extra_columns_descriptions=None,
     *,
     anonymize=None,
     format="auto",
@@ -1463,8 +1488,9 @@ def write_raw_bids(
            call ``raw.set_annotations(None)`` before invoking this function.
 
         .. note::
-           Descriptions of all event codes must be specified via the
-           ``event_id`` parameter.
+           Either, descriptions of all event codes must be specified via the
+           ``event_id`` parameter or each event must be accompanied by a
+           row in ``event_metadata``.
 
     event_id : dict | None
         Descriptions or names describing the event codes, if you passed
@@ -1475,6 +1501,11 @@ def write_raw_bids(
         contains :class:`~mne.Annotations`, you can use this parameter to
         assign event codes to each unique annotation description (mapping from
         description to event code).
+    event_metadata : pandas.DataFrame | None
+        Metadata for each event in ``events``. Each row corresponds to an event.
+    extra_columns_descriptions : dict | None
+        A dictionary that maps column names of the ``event_metadata`` to descriptions.
+        Each column of ``event_metadata`` must have a corresponding entry in this.
     anonymize : dict | None
         If `None` (default), no anonymization is performed.
         If a dictionary, data will be anonymized depending on the dictionary
@@ -1678,8 +1709,24 @@ def write_raw_bids(
             '"bids_path.task = <task>"'
         )
 
-    if events is not None and event_id is None:
-        raise ValueError("You passed events, but no event_id dictionary.")
+    if events is not None and event_id is None and event_metadata is None:
+        raise ValueError(
+            "You passed events, but no event_id dictionary " "or event_metadata."
+        )
+
+    if event_metadata is not None and extra_columns_descriptions is None:
+        raise ValueError(
+            "You passed event_metadata, but no "
+            "extra_columns_descriptions dictionary."
+        )
+
+    if event_metadata is not None:
+        for column in event_metadata.columns:
+            if column not in extra_columns_descriptions:
+                raise ValueError(
+                    f"Extra column {column} in event_metadata "
+                    f"is not described in extra_columns_descriptions."
+                )
 
     _validate_type(
         item=empty_room, item_name="empty_room", types=(mne.io.BaseRaw, BIDSPath, None)
@@ -1974,18 +2021,33 @@ def write_raw_bids(
     # Write events.
     if not data_is_emptyroom:
         events_array, event_dur, event_desc_id_map = _read_events(
-            events, event_id, raw, bids_path=bids_path
+            events,
+            event_id,
+            raw,
+            bids_path=bids_path,
         )
+
+        if event_metadata is not None:
+            event_desc_id_map = None
+
         if events_array.size != 0:
             _events_tsv(
                 events=events_array,
                 durations=event_dur,
                 raw=raw,
                 fname=events_tsv_path.fpath,
                 trial_type=event_desc_id_map,
+                event_metadata=event_metadata,
+                overwrite=overwrite,
+            )
+            has_trial_type = event_desc_id_map is not None
+
+            _events_json(
+                fname=events_json_path.fpath,
+                extra_columns=extra_columns_descriptions,
+                has_trial_type=has_trial_type,
                 overwrite=overwrite,
             )
-            _events_json(fname=events_json_path.fpath, overwrite=overwrite)
         # Kepp events_array around for BrainVision writing below.
         del event_desc_id_map, events, event_id, event_dur