Merge pull request #96 from catalystneuro/custom_subconverter

Add custom spikeglx subconverter

README.md

@@ -30,6 +30,8 @@ An `Interface` reads a single data stream (such as DLC pose estimation) and crea
 
 The `Converter` orchestrates the conversion by combining multiple interfaces, and can also be used to add additional metadata to the NWB file. It is responsible for creating the NWB file saved to disk.
 
+Occasionally, a sub-`Converter`, such as the `IBLSpikeGLXConverter`, will be used to handle the conversion of multiple data streams that is more complex than a single interface can handle; though these behave like other `Interfaces` with respect to the main orchestrating `Converter`.
+
 ## Metadata
 
 Anywhere you see handwritten text in the NWB files that is meant to be human-readable, it is likely that it was copied from the public Google IBL documents and written in the `.yaml` files found in `src/ibl_to_nwb/_metadata`.

src/ibl_to_nwb/_scripts/convert_brainwide_map_raw_only.py

@@ -1,10 +1,8 @@
 from pathlib import Path
 
-from brainbox.io.one import EphysSessionLoader, SpikeSortingLoader
-from neuroconv.datainterfaces import SpikeGLXRecordingInterface
 from one.api import ONE
 
-from ibl_to_nwb.converters import BrainwideMapConverter
+from ibl_to_nwb.converters import BrainwideMapConverter, IblSpikeGlxConverter
 from ibl_to_nwb.datainterfaces import RawVideoInterface
 
 session_id = "d32876dd-8303-4720-8e7e-20678dc2fd71"
@@ -28,39 +26,13 @@
 )
 
 # Specify the path to the SpikeGLX files on the server but use ONE API for timestamps
-probe_1_source_folder_path = Path("D:/example_data/ephy_testing_data/spikeglx/Noise4Sam_g0")
-probe_2_source_folder_path = Path(
-    "D:/example_data/ephy_testing_data/spikeglx/multi_trigger_multi_gate/SpikeGLX/5-19-2022-CI0/5-19-2022-CI0_g0/"
-)
-
-
-# Initialize interfaces one probe at a time and properly align raw timestamps
-ephys_session_loader = EphysSessionLoader(one=ibl_client, eid=session_id)
-
-probe_to_imec_map = {
-    "probe00": 0,
-    "probe01": 1,
-}
-
-data_interfaces = list()
-for probe_name, pid in ephys_session_loader.probes.items():
-    spike_sorting_loader = SpikeSortingLoader(pid=pid, one=ibl_client)
-
-    probe_index = probe_to_imec_map[probe_name]
-    for band in ["ap", "lf"]:
-        file_path = probe_1_source_folder_path / f"Noise4Sam_g0_imec0/Noise4Sam_g0_t0.imec{probe_index}.{band}.bin"
-        interface = SpikeGLXRecordingInterface(file_path=file_path)
-
-        # This is the syntax for aligning the raw timestamps; I cannot test this without the actual data as stored
-        # on your end, so please work with Heberto if there are any problems after uncommenting
-
-        # band_info = spike_sorting_loader.raw_electrophysiology(band=band, stream=True)
-        # aligned_timestamps = spike_sorting_loader.samples2times(numpy.arange(0, band_info.ns), direction='forward')
-        # interface.set_aligned_timestamps(aligned_timestamps=aligned_timestamps)
+data_interfaces = []
 
-        data_interfaces.append(interface)
+spikeglx_source_folder_path = Path("D:/example_data/ephy_testing_data/spikeglx/Noise4Sam_g0")
+spikeglx_subconverter = IblSpikeGlxConverter(folder_path=spikeglx_source_folder_path, one=ibl_client)
+data_interfaces.append(spikeglx_subconverter)
 
-# Raw video take some special handling
+# Raw video takes some special handling
 metadata_retrieval = BrainwideMapConverter(one=ibl_client, session=session_id, data_interfaces=[], verbose=False)
 subject_id = metadata_retrieval.get_metadata()["Subject"]["subject_id"]
 

src/ibl_to_nwb/converters/__init__.py

@@ -1,7 +1,9 @@
 from ._brainwide_map_converter import BrainwideMapConverter
 from ._iblconverter import IblConverter
+from ._ibl_spikeglx_converter import IblSpikeGlxConverter
 
 __all__ = [
     "BrainwideMapConverter",
     "IblConverter",
+    "IblSpikeGlxConverter",
 ]

src/ibl_to_nwb/converters/_ibl_spikeglx_converter.py

@@ -0,0 +1,44 @@
+from neuroconv.converters import SpikeGLXConverterPipe
+from one.api import ONE
+from pydantic import DirectoryPath
+from pynwb import NWBFile
+
+
+class IblSpikeGlxConverter(SpikeGLXConverterPipe):
+
+    def __init__(self, folder_path: DirectoryPath, one: ONE) -> None:
+        super().__init__(folder_path=folder_path)
+        self.one = one
+
+    def temporally_align_data_interfaces(self) -> None:
+        """Align the raw data timestamps to the other data streams using the ONE API."""
+        # This is the syntax for aligning the raw timestamps; I cannot test this without the actual data as stored
+        # on your end, so please work with Heberto if there are any problems after uncommenting
+        # probe_to_imec_map = {
+        #     "probe00": 0,
+        #     "probe01": 1,
+        # }
+        #
+        # ephys_session_loader = EphysSessionLoader(one=self.one, eid=session_id)
+        # probes = ephys_session_loader.probes
+        # for probe_name, pid in ephys_session_loader.probes.items():
+        #     spike_sorting_loader = SpikeSortingLoader(pid=pid, one=ibl_client)
+        #
+        #     probe_index = probe_to_imec_map[probe_name]
+        #     for band in ["ap", "lf"]:
+        #         recording_interface = next(
+        #             interface
+        #             for interface in self.data_interface_objects
+        #             if f"imec{probe_index}.{band}" in interface.source_data["file_path"]
+        #         )
+        #
+        #         band_info = spike_sorting_loader.raw_electrophysiology(band=band, stream=True)
+        #         aligned_timestamps = spike_sorting_loader.samples2times(numpy.arange(0, band_info.ns), direction='forward')
+        #         recording_interface.set_aligned_timestamps(aligned_timestamps=aligned_timestamps)
+        pass
+
+    def add_to_nwbfile(self, nwbfile: NWBFile, metadata) -> None:
+        self.temporally_align_data_interfaces()
+        super().add_to_nwbfile(nwbfile=nwbfile, metadata=metadata)
+
+        # TODO: Add ndx-extracellular-ephys here