Wrong probe map from open ephys and unexpected position from peak location
Closed this issue · 6 comments
Hi all,
Here is my issue (with some plot). Also open in spikeinterface SpikeInterface/spikeinterface#1546 (comment)
I started using a 4 shanks neuropixel probe for my recording and when I try to load the open ephys (version 0.64) recording into spikeinterface (version 0.96.1), my channel location are different from the one I have in my settings.xml
from probeinterface import Probe, ProbeGroup
from probeinterface import generate_linear_probe, generate_multi_shank
directory = '/camp/home/combadk/working/raw_data/raw_NP/NP_230315_A/2023-03-15_18-07-02'
raw_rec = si.read_openephys(directory, stream_id ='0',block_index=1)
raw_rec
fig, ax = plt.subplots(figsize=(15, 10))
si.plot_probe_map(raw_rec, ax=ax, with_channel_ids=True)
ax.set_ylim(-800, 750)
It give this result
When I would expect this probe map, that I created from the settings.xml from open ephys.
I changed the probe with those line and did some preprocessing.
df_probe = raw_rec.get_probe().to_dataframe()
chan_map = io.loadmat( '/camp/home/combadk/working/combadk/Kilosort/configFiles/chanMap.mat')
chan_location = pd.DataFrame(chan_map['xcoords'][0,:],chan_map['ycoords'][0,:])
df_probe['x'] = chan_map['xcoords'][0,:]
df_probe['y'] = chan_map['ycoords'][0,:]
df_probe
probe = Probe.from_dataframe(df_probe)
probe.set_device_channel_indices(df_probe.index)
rec = raw_rec.set_probe(probe)
rec1 = si.highpass_filter(rec, freq_min=400.)
rec4 = si.common_reference(rec1, operator="median", reference="global")
preprocess_recording = rec4
job_kwargs = dict(n_jobs=40, chunk_duration='1s', progress_bar=True)
peaks = detect_peaks(preprocess_recording, method='locally_exclusive', noise_levels=noise_levels_int16,
detect_threshold=5, local_radius_um=50., **job_kwargs)
peak_locations = localize_peaks(preprocess_recording, peaks, method='center_of_mass', **job_kwargs)
# check for drifts
fs = rec.sampling_frequency
fig, ax = plt.subplots(figsize=(10, 8))
ax.scatter(peaks['sample_ind'] / fs, peak_locations['y'], color='k', marker='.', alpha=0.002)
# we can also use the peak location estimates to have an insight of cluster separation before sorting
fig, ax = plt.subplots(figsize=(15, 10))
si.plot_probe_map(preprocess_recording, ax=ax, with_channel_ids=True)
ax.set_ylim(-800, 800)
ax.scatter(peak_locations['x'], peak_locations['y'], color='purple', alpha=0.002)
However the result from me checking on the drift looks like my first probe map which I expect to be wrong.
Any help would be appreciated! Thank you in advance!
@Kayv-cmb Could you send us the settings.xml file? It would help us find and fix the bug!
settings_2.xml.zip
Here a compressed version of my xml file(compressed just to upload on github)
Also I upgraded my version of spikeinterface to the latest one 0.97.1 and it seems to have solve the problem. I think the issue can be close.
Ah! Thats' good to hear! :) Closing then!
Do you mind if we use the settings file for internal testing? We don't have a multi-shank NP2 test file for Open Ephys
I think it should be ok if you use it for internal testing.
If it's ok, we would like to include it in the set of test files that we test against on GitHub: https://github.com/SpikeInterface/probeinterface/tree/main/tests/data/openephys
It's only the settings.xml file, so there is no ephys data associated to it. Let me knoe
Hi,
Feel free to use it.