def prepare_annotations(
xdffile: FileName,
channel: str,
pre_in_ms: float,
post_in_ms: float,
xmlfile: FileName = None,
event_stream: str = "localite_marker",
event_name: Union[str, int] = "coil_0_didt",
) -> Annotations:
"""load a documentation.txt and cnt-files and distill annotations from them
args
----
xdffile: FileName
the :code:`.xdf`-file with the recorded streams, e.g. data and markers
channel: str
which channel to pick
pre_in_ms: float
how many ms to cut before the tms
post_in_ms: float
how many ms to cut after the tms
returns
-------
annotation: Annotations
the annotations for this origin files
"""
stream_of_interest = channel # rename to have same function signature
streams = XDFFile(xdffile)
if stream_of_interest in streams:
datastream = streams[stream_of_interest]
else:
raise KeyError(
f"Stream {stream_of_interest} was not found in the data")
e_stream = streams[event_stream]
time_stamps = [ts for ts in yield_timestamps(e_stream, event_name)]
event_count = len(time_stamps)
if "localite_flow" in streams or "localite_marker" in streams:
loc_stream = streams["localite_marker"]
coords = list(yield_loc_coords(loc_stream, time_stamps))
stimulation_intensity_didt = list(
yield_loc_didt(loc_stream, time_stamps))
stimulation_intensity_mso = list(yield_loc_mso(loc_stream,
time_stamps))
else:
coords = list_nan_coords(event_count)
stimulation_intensity_didt = list_nan(event_count)
stimulation_intensity_mso = list_nan(event_count)
print(f"Found {event_count} events")
if "reiz_marker_sa" in streams:
comments = [
c for c in yield_comments(
streams["reiz_marker_sa"],
time_stamps=time_stamps,
identifier="stimulus_idx",
relative="earlier",
)
]
else:
comments = ["" for c in time_stamps]
# global fields
fs = datastream.nominal_srate
anno = AnnotationFactory(readin="tms",
readout="erp",
origin=Path(xdffile).name)
anno.set("filedate", time.ctime(Path(xdffile).stat().st_mtime))
anno.set("subject", "") # TODO parse from somewhere
anno.set("samplingrate", fs)
anno.set("samples_pre_event", int(pre_in_ms * fs / 1000))
anno.set("samples_post_event", int(post_in_ms * fs / 1000))
anno.set("channel_of_interest", datastream.name)
anno.set("channel_labels", datastream.channel_labels)
# trace fields
event_samples = find_closest_samples(datastream, time_stamps)
event_times = [
float(t) for t in datastream.time_stamps[event_samples] -
datastream.time_stamps[0]
]
time_since_last_pulse = [inf] + [
a - b for a, b in zip(event_times[1:], event_times[0:-1])
]
for idx in range(event_count):
tattr = {
"id": idx,
"event_name": e_stream.name + "-" + str(event_name),
"event_sample": event_samples[idx],
"event_time": event_times[idx],
"xyz_coords": coords[idx],
"time_since_last_pulse_in_s": time_since_last_pulse[idx],
"stimulation_intensity_mso": stimulation_intensity_mso[idx],
"stimulation_intensity_didt": stimulation_intensity_didt[idx],
"comment": comments[idx],
}
anno.append_trace_attr(tattr)
return anno.anno
def prepare_annotations(
xdffile: FileName,
channel: str,
pre_in_ms: float,
post_in_ms: float,
xmlfile: FileName = None,
event_name="coil_0_didt",
event_stream="localite_marker",
comment_name=None,
) -> Annotations:
"""load a documentation.txt and cnt-files and distill annotations from them
args
----
xmlfile: FileName
an option xml file with information about the target coordinates
readout: str
which readout to use
channel: str
which channel to pick
pre_in_ms: float
how many ms to cut before the tms
post_in_ms: float
how many ms to cut after the tms
xdffile: FileName
the :code:`.xdf`-file with the recorded streams, e.g. data and markers
returns
-------
annotation: Annotations
the annotations for this origin files
"""
# ------------------
streams = XDFFile(xdffile)
datastream = pick_stream_with_channel(channel, streams)
event_stream = streams[event_stream]
print(f"Reading events from {event_stream.name} using {event_name}")
time_stamps = [ts for ts in yield_timestamps(event_stream, event_name)]
event_count = len(time_stamps)
print(f"Found {event_count} events")
if "localite_flow" in streams or "localite_marker" in streams:
loc_stream = streams["localite_marker"]
print(f"Reading information from {loc_stream.name}")
coords = list(yield_loc_coords(loc_stream, time_stamps))
stimulation_intensity_didt = list(yield_loc_didt(loc_stream, time_stamps))
stimulation_intensity_mso = list(yield_loc_mso(loc_stream, time_stamps))
else:
coords = list_nan_coords(event_count)
stimulation_intensity_didt = list_nan(event_count)
stimulation_intensity_mso = list_nan(event_count)
if "reiz_marker_sa" in streams and comment_name is not None:
print("Reading comments from reiz_marker_sa")
comments = [
c
for c in yield_comments(
streams["reiz_marker_sa"],
time_stamps=time_stamps,
identifier="stimulus_idx",
relative="earlier",
)
]
else:
comments = ["" for c in time_stamps]
if "BrainVision RDA Markers" in streams:
rda_stamps = list(yield_timestamps(streams["BrainVision RDA Markers"], "S 2"))
print(f"Found {len(rda_stamps)} 'S 2' for {event_count} events")
if len(rda_stamps) >= len(time_stamps):
time_stamps = [find_closest(ts, rda_stamps) for ts in time_stamps]
print("Corrected event timestamps for RDA 'S 2'")
else:
print("Count mismatch between RDA and Localite events")
if "BrainVision RDA" in streams:
bvr = streams["BrainVision RDA"]
time_stamps = correct_tkeo(bvr, time_stamps)
print("Corrected event timestamps for TMS artifact")
# global fields
fs = datastream.nominal_srate
anno = AnnotationFactory(readin="tms", readout="cmep", origin=Path(xdffile).name)
anno.set("filedate", time.ctime(Path(xdffile).stat().st_mtime))
anno.set("subject", "") # TODO parse from correctly organized file
anno.set("samplingrate", fs)
anno.set("samples_pre_event", int(pre_in_ms * fs / 1000))
anno.set("samples_post_event", int(post_in_ms * fs / 1000))
anno.set("channel_of_interest", channel)
anno.set("channel_labels", [channel])
# trace fields
event_samples = find_closest_samples(datastream, time_stamps)
event_times = [
float(t)
for t in datastream.time_stamps[event_samples] - datastream.time_stamps[0]
]
time_since_last_pulse = [inf] + [
a - b for a, b in zip(event_times[1:], event_times[0:-1])
]
for idx, t in enumerate(event_samples):
tattr = {
"id": idx,
"event_name": event_stream.name + "-" + str(event_name),
"event_sample": event_samples[idx],
"event_time": event_times[idx],
"xyz_coords": coords[idx],
"time_since_last_pulse_in_s": time_since_last_pulse[idx],
"stimulation_intensity_mso": stimulation_intensity_mso[idx],
"stimulation_intensity_didt": stimulation_intensity_didt[idx],
}
anno.append_trace_attr(tattr)
return anno.anno
def prepare_annotations_multfile(
files,
origin,
filedate,
channel: str,
pre_in_ms: float,
post_in_ms: float,
comment_name=None,
):
# we assume that all mapping was done after any spongebob assessments,
# therefore everything earlier than the latest spongebob trigger is
# considered irrelevant
# alternatively, everything in an xdf-file earlier than a reiz_marker of
# "" or earlier than any other marker is irrelevant. This is because every
# start of a reiz-marker sends out an "" to see whether the marker-server
# is running, while for all other assessments but mapping, the reiz-marker
# was used to send markers
for streams in files:
if streams["reiz_marker_sa"].time_series[-1] == [""]:
iu1 = 0
else:
iu1 = streams["reiz_marker_sa"].time_stamps[-1]
spbob = np.where(streams["Spongebob-Data"].time_series[:,
11] == 1.0)[0]
if len(spbob) == 0: # we never triggered Spongebob in this file
iu2 = 0
else:
idx = spbob[-1]
iu2 = streams["Spongebob-Data"].time_stamps[idx]
if max((iu1, iu2)) == 0:
continue
else:
irrelevant_until = max((iu1, iu2))
print(irrelevant_until)
# we concatenate the time_series and time_stamps from the multiple xdf \
# files. Because the clock is continuous and monotonic, we do not need to
# correct for any possible reset between xdf files, and gaps are jumped
# later
time_stamps = []
data_series = None
data_stamps = None
for streams in files:
datastream = pick_stream_with_channel(channel, streams)
if data_series is None:
data_series = datastream.time_series
data_stamps = datastream.time_stamps
else:
data_series = np.concatenate((data_series, datastream.time_series),
axis=0)
data_stamps = np.concatenate((data_stamps, datastream.time_stamps),
axis=0)
for event in streams["BrainVision RDA Markers"].time_stamps:
if event > irrelevant_until:
time_stamps.append(event)
event_count = len(time_stamps)
coords = list_nan_coords(event_count)
stimulation_intensity_didt = list_nan(event_count)
stimulation_intensity_mso = list_nan(event_count)
print(f"Found {event_count} events")
if event_count == 350:
grid_layout = "5x7"
print("This corresponds to a subject with a 5x7 grid")
elif event_count == 360:
grid_layout = "6x6"
print("This corresponds to a subject with a 6x6 grid")
else:
grid_layout = "Unknown"
print("This does not correspond to a known grid layout")
# global fields
fs = datastream.nominal_srate
anno = AnnotationFactory(readin="tms", readout="cmep", origin=origin)
anno.set("filedate", filedate)
anno.set("subject", "") # TODO parse from correctly organized file
anno.set("samplingrate", fs)
anno.set("samples_pre_event", int(pre_in_ms * fs / 1000))
anno.set("samples_post_event", int(post_in_ms * fs / 1000))
anno.set("channel_of_interest", channel)
anno.set("channel_labels", [channel])
anno.set("global_comment", f"grid_layout={grid_layout}")
# trace fields
event_samples = []
for ts in time_stamps:
idx = int(np.argmin(np.abs(data_stamps - ts)))
event_samples.append(idx)
# New Implementation
# gmfp = np.std(data_series[:, 0:64], 1)
# artifact = tkeo(gmfp)
# aptp = []
# tp = []
# for onset in event_samples:
# hood = gmfp[onset - 50 : onset + 50]
# aptp.append(np.max(hood))
# tp.append(int(np.argmax(hood) - 50 + onset))
# event_samples = tp
gmfp = np.std(data_series[:, 0:64], 1)
aptp = []
tp = []
for onset in event_samples:
artifact = gmfp[onset - 25:onset + 25]
aptp.append(np.ptp(artifact))
tp.append(int(np.argmax(artifact) - 25 + onset))
event_samples = tp
event_times = [
float(t) for t in data_stamps[event_samples] - data_stamps[0]
]
time_since_last_pulse = [inf] + [
a - b for a, b in zip(event_times[1:], event_times[0:-1])
]
for idx, t in enumerate(event_samples):
tattr = {
"id": idx,
"comment": f'{{"artifact_amplitude":{aptp[idx]:3.2f}}}',
"event_name": "BrainVision RDA Markers - 'S 2'",
"event_sample": event_samples[idx],
"event_time": event_times[idx],
"xyz_coords": coords[idx],
"time_since_last_pulse_in_s": time_since_last_pulse[idx],
"stimulation_intensity_mso": stimulation_intensity_mso[idx],
"stimulation_intensity_didt": stimulation_intensity_didt[idx],
}
anno.append_trace_attr(tattr)
return anno.anno
def prepare_annotations(
streams,
origin,
filedate,
channel: str,
pre_in_ms: float,
post_in_ms: float,
comment_name=None,
) -> Annotations:
"""load a documentation.txt and cnt-files and distill annotations from them
args
----
xmlfile: FileName
an option xml file with information about the target coordinates
readout: str
which readout to use
channel: str
which channel to pick
pre_in_ms: float
how many ms to cut before the tms
post_in_ms: float
how many ms to cut after the tms
xdffile: FileName
the :code:`.xdf`-file with the recorded streams, e.g. data and markers
returns
-------
annotation: Annotations
the annotations for this origin files
"""
# ------------------
datastream = pick_stream_with_channel(channel, streams)
iu1 = streams["reiz_marker_sa"].time_stamps[-1]
idx = np.where(streams["Spongebob-Data"].time_series[:, 11] == 1.0)[0][-1]
iu2 = streams["Spongebob-Data"].time_stamps[idx]
irrelevant_until = max((iu1, iu2))
time_stamps = []
for event in streams["BrainVision RDA Markers"].time_stamps:
if event > irrelevant_until:
time_stamps.append(event)
event_count = len(time_stamps)
coords = list_nan_coords(event_count)
stimulation_intensity_didt = list_nan(event_count)
stimulation_intensity_mso = list_nan(event_count)
comments = ["" for c in time_stamps]
print(f"Found {event_count} events")
if event_count == 350:
grid_layout = "5x7"
print("This corresponds to a subject with a 5x7 grid")
elif event_count == 360:
grid_layout = "6x6"
print("This corresponds to a subject with a 6x6 grid")
else:
grid_layout = "Unknown"
print("This does not correspond to a known grid layout")
# global fields
fs = datastream.nominal_srate
anno = AnnotationFactory(readin="tms", readout="cmep", origin=origin)
anno.set("filedate", filedate)
anno.set("subject", "") # TODO parse from correctly organized file
anno.set("samplingrate", fs)
anno.set("samples_pre_event", int(pre_in_ms * fs / 1000))
anno.set("samples_post_event", int(post_in_ms * fs / 1000))
anno.set("channel_of_interest", channel)
anno.set("channel_labels", [channel])
anno.set("global_comment", f"grid_layout={grid_layout}")
# trace fields
event_samples = find_closest_samples(datastream, time_stamps)
# shift onset on Peak of artifact
ephys = streams["BrainVision RDA"]
gmfp = np.std(ephys.time_series[:, 0:64], 1)
aptp = []
tp = []
for onset in event_samples:
artifact = gmfp[onset - 25:onset + 25]
aptp.append(np.ptp(artifact))
tp.append(int(np.argmax(artifact) - 25 + onset))
event_samples = tp
event_times = [
float(t) for t in datastream.time_stamps[event_samples] -
datastream.time_stamps[0]
]
time_since_last_pulse = [inf] + [
a - b for a, b in zip(event_times[1:], event_times[0:-1])
]
for idx, t in enumerate(event_samples):
tattr = {
"id": idx,
"comment": f'{{"artifact_amplitude":{aptp[idx]:3.2f}}}',
"event_name": "BrainVision RDA Markers - 'S 2'",
"event_sample": event_samples[idx],
"event_time": event_times[idx],
"xyz_coords": coords[idx],
"time_since_last_pulse_in_s": time_since_last_pulse[idx],
"stimulation_intensity_mso": stimulation_intensity_mso[idx],
"stimulation_intensity_didt": stimulation_intensity_didt[idx],
}
anno.append_trace_attr(tattr)
return anno.anno
def prepare_annotations_multifile(
files,
origin,
filedate,
channel: str,
pre_in_ms: float,
post_in_ms: float,
comment_name=None,
):
# we assume that all mapping was done after any spongebob assessments,
# therefore everything earlier than the latest spongebob trigger is
# considered irrelevant
# alternatively, everything in an xdf-file earlier than a reiz_marker of
# "" or earlier than any other marker is irrelevant. This is because every
# start of a reiz-marker sends out an "" to see whether the marker-server
# is running, while for all other assessments but mapping, the reiz-marker
# was used to send markers
for streams in files:
if streams["reiz_marker_sa"].time_series[-1] == [""]:
iu1 = 0
else:
iu1 = streams["reiz_marker_sa"].time_stamps[-1]
spbob = np.where(streams["Spongebob-Data"].time_series[:,
11] == 1.0)[0]
if len(spbob) == 0: # we never triggered Spongebob in this file
iu2 = 0
else:
idx = spbob[-1]
iu2 = streams["Spongebob-Data"].time_stamps[idx]
if iu1 > iu2:
print(
"WARNING: Spongebob data was not transmitted after NMES-IOC. Possible reason: Death of Outlet?"
)
else:
print("Spongebob data was transmitted after NMES-IOC")
if max((iu1, iu2)) == 0:
continue
else:
irrelevant_until = max((iu1, iu2))
print(irrelevant_until)
# we concatenate the time_series and time_stamps from the multiple xdf \
# files. Because the clock is continuous and monotonic, we do not need to
# correct for any possible reset between xdf files, and gaps are jumped
# later
time_stamps = []
data_series = None
data_stamps = None
for streams in files:
datastream = pick_stream_with_channel(channel, streams)
if data_series is None:
data_series = datastream.time_series
data_stamps = datastream.time_stamps
else:
data_series = np.concatenate((data_series, datastream.time_series),
axis=0)
data_stamps = np.concatenate((data_stamps, datastream.time_stamps),
axis=0)
for event in streams["BrainVision RDA Markers"].time_stamps:
if event > irrelevant_until:
time_stamps.append(event)
event_count = len(time_stamps)
coords = list_nan_coords(event_count)
stimulation_intensity_didt = list_nan(event_count)
stimulation_intensity_mso = list_nan(event_count)
print(f"Found {event_count} events")
if event_count == 350:
grid_layout = "5x7"
print("This corresponds to a subject with a 5x7 grid")
elif event_count == 360:
grid_layout = "6x6"
print("This corresponds to a subject with a 6x6 grid")
else:
grid_layout = "Unknown"
print("This does not correspond to a known grid layout")
# global fields
fs = datastream.nominal_srate
anno = AnnotationFactory(readin="tms", readout="cmep", origin=origin)
anno.set("filedate", filedate)
anno.set("subject", "") # TODO parse from correctly organized file
anno.set("samplingrate", fs)
anno.set("samples_pre_event", int(pre_in_ms * fs / 1000))
anno.set("samples_post_event", int(post_in_ms * fs / 1000))
anno.set("channel_of_interest", channel)
anno.set("channel_labels", [channel])
anno.set("global_comment", f"grid_layout={grid_layout}")
# trace fields
event_samples = []
for ts in time_stamps:
idx = int(np.argmin(np.abs(data_stamps - ts)))
event_samples.append(idx)
# New Implementation
# gmfp = np.std(data_series[:, 0:64], 1)
# artifact = tkeo(gmfp)
# aptp = []
# tp = []
# for onset in event_samples:
# hood = gmfp[onset - 50 : onset + 50]
# aptp.append(np.max(hood))
# tp.append(int(np.argmax(hood) - 50 + onset))
# event_samples = tp
gmfp = np.std(data_series[:, 0:64], 1)
aptp = []
tp = []
for onset in event_samples:
artifact = gmfp[onset - 25:onset + 25]
aptp.append(np.ptp(artifact))
tp.append(int(np.argmax(artifact) - 25 + onset))
event_samples = tp
event_times = [
float(t) for t in data_stamps[event_samples] - data_stamps[0]
]
time_since_last_pulse = [inf] + [
a - b for a, b in zip(event_times[1:], event_times[0:-1])
]
# sometimes, we record way too many pulses for a mapping
max_allowed_counts = int(360 * 1.2)
if event_count > max_allowed_counts:
print(
"WARNING: There are too many pulses, throwing away possible faulty ones"
)
template = np.ones(180)
template[0::5] = 1.5
template[0] = 100
selection = []
old_xcorr = 0
new_xcorr = 0
to = 1
fr = 181
while fr < max_allowed_counts:
new_xcorr = np.correlate(time_since_last_pulse[-fr:-to], template)
print(fr, to, new_xcorr)
if new_xcorr < old_xcorr / 2:
# the last was a good fit, at least twice better than now
winner = (fr - 1, to - 1)
selection.extend(list(range(to + 1, fr + 1)))
fr += 180
to += 180
old_xcorr = 0
new_xcorr = 0
print("We found a winner", winner)
else:
old_xcorr = new_xcorr
fr += 1
to += 1
if len(selection) != 720:
raise Exception("Did not find a good selection of good events")
event_samples = drop_selection(event_samples, selection)
event_times = drop_selection(event_times, selection)
coords = drop_selection(coords, selection)
time_since_last_pulse = drop_selection(time_since_last_pulse,
selection)
stimulation_intensity_mso = drop_selection(stimulation_intensity_mso,
selection)
stimulation_intensity_didt = drop_selection(stimulation_intensity_didt,
selection)
aptp = drop_selection(aptp, selection)
for idx, t in enumerate(event_samples):
tattr = {
"id": idx,
"comment": f'{{"artifact_amplitude":{aptp[idx]:3.2f}}}',
"event_name": "BrainVision RDA Markers - 'S 2'",
"event_sample": event_samples[idx],
"event_time": event_times[idx],
"xyz_coords": coords[idx],
"time_since_last_pulse_in_s": time_since_last_pulse[idx],
"stimulation_intensity_mso": stimulation_intensity_mso[idx],
"stimulation_intensity_didt": stimulation_intensity_didt[idx],
}
anno.append_trace_attr(tattr)
return anno.anno