Prototype Behavior Interface

src/schneider_lab_to_nwb/schneider_2024/schneider_2024_behaviorinterface.py

@@ -1,25 +1,115 @@
 """Primary class for converting experiment-specific behavior."""
 from pynwb.file import NWBFile
+from pydantic import FilePath
+import numpy as np
+from h5py import File
+from hdmf.common.table import DynamicTableRegion
+from pynwb.behavior import BehavioralTimeSeries, TimeSeries
+from ndx_events import EventTypesTable, EventsTable, Task, TimestampVectorData
 
 from neuroconv.basedatainterface import BaseDataInterface
 from neuroconv.utils import DeepDict
+from neuroconv.tools import nwb_helpers
+
 
 class Schneider2024BehaviorInterface(BaseDataInterface):
     """Behavior interface for schneider_2024 conversion"""
 
     keywords = ["behavior"]
-
-    def __init__(self):
-        # This should load the data lazily and prepare variables you need
-        pass
+
+    def __init__(self, file_path: FilePath):
+        super().__init__(file_path=file_path)
 
     def get_metadata(self) -> DeepDict:
         # Automatically retrieve as much metadata as possible from the source files available
-        metadata = super().get_metadata()   
+        metadata = super().get_metadata()
 
         return metadata
 
     def add_to_nwbfile(self, nwbfile: NWBFile, metadata: dict):
-        # All the custom code to add the data the nwbfile
+        # Read Data
+        file_path = self.source_data["file_path"]
+        with File(file_path, "r") as file:
+            behavioral_time_series, name_to_times, name_to_values = [], dict(), dict()
+            for time_series_dict in metadata["Behavior"]["TimeSeries"]:
+                name = time_series_dict["name"]
+                timestamps = np.array(file["continuous"][name]["time"]).squeeze()
+                data = np.array(file["continuous"][name]["value"]).squeeze()
+                time_series = TimeSeries(
+                    name=name,
+                    timestamps=timestamps,
+                    data=data,
+                    unit="a.u.",
+                    description=time_series_dict["description"],
+                )
+                behavioral_time_series.append(time_series)
+            for event_dict in metadata["Behavior"]["Events"]:
+                name = event_dict["name"]
+                times = np.array(file["events"][name]["time"]).squeeze()
+                name_to_times[name] = times
+            for event_dict in metadata["Behavior"]["ValuedEvents"]:
+                name = event_dict["name"]
+                times = np.array(file["events"][name]["time"]).squeeze()
+                values = np.array(file["events"][name]["value"]).squeeze()
+                name_to_times[name] = times
+                name_to_values[name] = values
+
+        # Add Data to NWBFile
+        behavior_module = nwb_helpers.get_module(
+            nwbfile=nwbfile,
+            name="behavior",
+            description="Behavioral data from the experiment.",
+        )
+
+        # Add BehavioralTimeSeries
+        behavioral_time_series = BehavioralTimeSeries(
+            time_series=behavioral_time_series,
+            name="behavioral_time_series",
+        )
+        behavior_module.add(behavioral_time_series)
+
+        # Add Events
+        event_types_table = EventTypesTable(name="event_types", description="Metadata about event types.")
+        event_type_name_to_row = dict()
+        i = 0
+        for event_dict in metadata["Behavior"]["Events"]:
+            event_type_name_to_row[event_dict["name"]] = i
+            event_types_table.add_row(
+                event_name=event_dict["name"],
+                event_type_description=event_dict["description"],
+            )
+            i += 1
+        for event_dict in metadata["Behavior"]["ValuedEvents"]:
+            event_type_name_to_row[event_dict["name"]] = i
+            event_types_table.add_row(
+                event_name=event_dict["name"],
+                event_type_description=event_dict["description"],
+            )
+            i += 1
+        events_table = EventsTable(
+            name="events_table",
+            description="Metadata about events.",
+            target_tables={"event_type": event_types_table},
+        )
+        for event_dict in metadata["Behavior"]["Events"]:
+            event_times = name_to_times[event_dict["name"]]
+            event_type = event_type_name_to_row[event_dict["name"]]
+            for event_time in event_times:
+                events_table.add_row(timestamp=event_time, event_type=event_type)
+        valued_events_table = EventsTable(
+            name="valued_events_table",
+            description="Metadata about valued events.",
+            target_tables={"event_type": event_types_table},
+        )
+        valued_events_table.add_column(name="value", description="Value of the event.")
+        for event_dict in metadata["Behavior"]["ValuedEvents"]:
+            event_times = name_to_times[event_dict["name"]]
+            event_values = name_to_values[event_dict["name"]]
+            event_type = event_type_name_to_row[event_dict["name"]]
+            for event_time, event_value in zip(event_times, event_values):
+                valued_events_table.add_row(timestamp=event_time, event_type=event_type, value=event_value)
+        behavior_module.add(events_table)
+        behavior_module.add(valued_events_table)
 
-        raise NotImplementedError()
+        task = Task(event_types=event_types_table)
+        nwbfile.add_lab_meta_data(task)

src/schneider_lab_to_nwb/schneider_2024/schneider_2024_convert_session.py

@@ -18,6 +18,7 @@ def session_to_nwb(data_dir_path: Union[str, Path], output_dir_path: Union[str,
     output_dir_path = Path(output_dir_path)
     recording_folder_path = data_dir_path / "Raw Ephys" / "m69_2023-10-31_17-24-15_Day1_A1"
     sorting_folder_path = data_dir_path / "Processed Ephys" / "m69_2023-10-31_17-24-15_Day1_A1"
+    behavior_file_path = data_dir_path / "Behavior" / "m69_231031" / "raw_m69_231031_001.mat"
     if stub_test:
         output_dir_path = output_dir_path / "nwb_stub"
         recording_folder_path = recording_folder_path.with_name(recording_folder_path.name + "_stubbed")
@@ -38,9 +39,9 @@ def session_to_nwb(data_dir_path: Union[str, Path], output_dir_path: Union[str,
     source_data.update(dict(Sorting=dict(folder_path=sorting_folder_path)))
     conversion_options.update(dict(Sorting=dict()))
 
-    # # Add Behavior
-    # source_data.update(dict(Behavior=dict()))
-    # conversion_options.update(dict(Behavior=dict()))
+    # Add Behavior
+    source_data.update(dict(Behavior=dict(file_path=behavior_file_path)))
+    conversion_options.update(dict(Behavior=dict()))
 
     converter = Schneider2024NWBConverter(source_data=source_data)
 

src/schneider_lab_to_nwb/schneider_2024/schneider_2024_metadata.yaml

@@ -30,27 +30,27 @@ Ecephys:
   ElectricalSeries:
     - name: ElectricalSeries
       description: Recording of AC neural responses in mice performing this behavioral task will utilize dense 128-channel recording probes (Masmanidis Lab). These recording probes span a depth ~1mm, allowing for sampling of all layers of cortex. Electrophysiology data will be recorded using OpenEphys Acquisition Board v2.4 and associated OpenEphys GUI software.
-  # UnitProperties:
-  #   - name: n_spikes
-  #     description: Number of spikes recorded from each unit.
-  #   - name: fr
-  #     description: Average firing rate of each unit.
-  #   - name: depth
-  #     description: Estimated depth of each unit in micrometers.
-  #   - name: Amplitude
-  #     description: Per-template amplitudes, computed as the L2 norm of the template.
-  #   - name: ContamPct
-  #     description: Contamination rate for each template, computed as fraction of refractory period violations relative to expectation based on a Poisson process.
-  #   - name: KSLabel
-  #     description: Label indicating whether each template is 'mua' (multi-unit activity) or 'good' (refractory).
-  #   - name: original_cluster_id
-  #     description: Original cluster ID assigned by Kilosort.
-  #   - name: amp
-  #     description: For every template, the maximum amplitude of the template waveforms across all channels.
-  #   - name: ch
-  #     description: The channel label of the best channel, as defined by the user.
-  #   - name: sh
-  #     description: The shank label of the best channel.
+
+Behavior:
+  TimeSeries:
+    - name: encoder
+      description: Sampled values for entire duration of experiment for lever pressing/treadmill behavior read from a quadrature encoder.
+    - name: lick
+      description: Samples values for entire duration of experiment for voltage signal readout from an infrared/capacitive) lickometer sensor.
+  Events:
+    - name: target
+      description: Time at which the target zone is entered during a press.
+    - name: targetOUT
+      description: Time at which the target zone is overshot during a press.
+    - name: toneIN
+      description: Time at which target entry tone is played.
+    - name: toneOUT
+      description: Time at which target exit tone is played.
+    - name: valve
+      description: Times at which solenoid valve opens to deliver water after a correct trial.
+  ValuedEvents:
+    - name: tuningTones
+      description: Times at which tuning tones are played to an animal after a behavioral experiment during ephys recording sessions.
 
 Sorting:
   units_description: Neural spikes will be sorted offline using Kilosort 2.5 and Phy2 software and manually curated to ensure precise spike time acquisition.

src/schneider_lab_to_nwb/schneider_2024/schneider_2024_notes.md

@@ -1,9 +1,15 @@
 # Notes concerning the schneider_2024 conversion
 
+## Behavior
+- opto events is just [0, 1] like empty variables (lick events), and opto_time in 'push' is all NaNs.  Where is the actual opto stim times recorded?
+- TuningTones has a values field with integers, what does this correspond to?
+- 'push' is basically some kind of trials table, but need descriptions for variables ex. ITI_respect?
+
 ## Data Requests
 - Mice sexes
 - Remaining data for Grant's project
 - More detailed position info for recording probe
     - Subfield of auditory cortex: A1? A2? AAF? etc.
     - stereotactic coordinates of the whole probe
-- 
+- Detailed description of the behavioral paradigm
+- Description of lickometer and lever/treadmill quadrature encoder.

src/schneider_lab_to_nwb/schneider_2024/schneider_2024_nwbconverter.py

@@ -14,4 +14,5 @@ class Schneider2024NWBConverter(NWBConverter):
     data_interface_classes = dict(
         Recording=OpenEphysRecordingInterface,
         Sorting=PhySortingInterface,
+        Behavior=Schneider2024BehaviorInterface,
     )