Motion correction error: win_step_um=400.0/win_margin_um=-200.0 are too large for the probe size

[desaponte]

Hello!
I'm following the spike interface tutorial on testing different motion correctors on my NP2 4 shank recording.

The kilosort-like correction works..... but when I run dredge , I get a probe size issue.

[alejoe91]

Can you paste the entire error?

[desaponte]

Sure! Please see below, thanks for the help

---

Exception Traceback (most recent call last)
Cell In[17], line 15
13 if folder.exists():
14 shutil.rmtree(folder)
---> 15 recording_corrected, motion, motion_info = si.correct_motion(
16 rec, preset=some_presets, folder=folder, output_motion=True,output_motion_info=True,**job_kwargs)

File ~/miniconda3/envs/spikeinterface/lib/python3.9/site-packages/spikeinterface/preprocessing/motion.py:436, in correct_motion(recording, preset, folder, output_motion, output_motion_info, overwrite, detect_kwargs, select_kwargs, localize_peaks_kwargs, estimate_motion_kwargs, interpolate_motion_kwargs, **job_kwargs)
429 run_times = dict(
430 detect_peaks=t1 - t0,
431 select_peaks=t2 - t1,
432 localize_peaks=t3 - t2,
433 )
435 t0 = time.perf_counter()
--> 436 motion = estimate_motion(recording, peaks, peak_locations, progress_bar=progress_bar, **estimate_motion_kwargs)
437 t1 = time.perf_counter()
438 run_times["estimate_motion"] = t1 - t0

File ~/miniconda3/envs/spikeinterface/lib/python3.9/site-packages/spikeinterface/sortingcomponents/motion/motion_estimation.py:111, in estimate_motion(recording, peaks, peak_locations, direction, rigid, win_shape, win_step_um, win_scale_um, win_margin_um, method, extra_outputs, progress_bar, verbose, margin_um, **method_kwargs)
108 extra = None
110 # run method
--> 111 motion = method_class.run(
112 recording,
113 peaks,
114 peak_locations,
115 direction,
116 rigid,
117 win_shape,
118 win_step_um,
119 win_scale_um,
120 win_margin_um,
121 verbose,
122 progress_bar,
123 extra,
124 **method_kwargs,
125 )
127 if extra_outputs:
128 return motion, extra

File ~/miniconda3/envs/spikeinterface/lib/python3.9/site-packages/spikeinterface/sortingcomponents/motion/dredge.py:116, in DredgeApRegistration.run(cls, recording, peaks, peak_locations, direction, rigid, win_shape, win_step_um, win_scale_um, win_margin_um, verbose, progress_bar, extra, **method_kwargs)
113 except ImportError:
114 raise ImportError("The dredge method require torch: pip install torch")
--> 116 outs = dredge_ap(
117 recording,
118 peaks,
119 peak_locations,
120 direction=direction,
121 rigid=rigid,
122 win_shape=win_shape,
123 win_step_um=win_step_um,
124 win_scale_um=win_scale_um,
125 win_margin_um=win_margin_um,
126 extra_outputs=(extra is not None),
127 progress_bar=progress_bar,
128 **method_kwargs,
129 )
131 if extra is not None:
132 motion, extra_ = outs

File ~/miniconda3/envs/spikeinterface/lib/python3.9/site-packages/spikeinterface/sortingcomponents/motion/dredge.py:310, in dredge_ap(recording, peaks, peak_locations, direction, rigid, win_shape, win_step_um, win_scale_um, win_margin_um, bin_um, bin_s, max_disp_um, time_horizon_s, mincorr, do_window_weights, weights_threshold_low, weights_threshold_high, mincorr_percentile, mincorr_percentile_nneighbs, amp_scale_fn, post_transform, histogram_depth_smooth_um, histogram_time_smooth_s, avg_in_bin, thomas_kw, xcorr_kw, device, progress_bar, extra_outputs, precomputed_D_C_maxdisp)
307 contact_depths = recording.get_channel_locations()[:, dim]
308 spatial_bin_centers = 0.5 * (spatial_bin_edges_um[1:] + spatial_bin_edges_um[:-1])
--> 310 windows, window_centers = get_spatial_windows(
311 contact_depths,
312 spatial_bin_centers,
313 rigid=rigid,
314 win_shape=win_shape,
315 win_step_um=win_step_um,
316 win_scale_um=win_scale_um,
317 win_margin_um=win_margin_um,
318 zero_threshold=1e-5,
319 )
321 # TODO charlie : the count has disapeared
322 # if extra_outputs and count_masked_correlation:
323 # extra["counts"] = counts
324
325 # cross-correlate to get D and C
326 if precomputed_D_C_maxdisp is None:

File ~/miniconda3/envs/spikeinterface/lib/python3.9/site-packages/spikeinterface/sortingcomponents/motion/motion_utils.py:312, in get_spatial_windows(contact_depths, spatial_bin_centers, rigid, win_shape, win_step_um, win_scale_um, win_margin_um, zero_threshold)
309 num_windows = int((max_ - min_) // win_step_um)
311 if num_windows < 1:
--> 312 raise Exception(
313 f"get_spatial_windows(): {win_step_um=}/{win_scale_um=}/{win_margin_um=} are too large for the "
314 f"probe size (depth range={np.ptp(contact_depths)}). You can try to reduce them or use rigid motion."
315 )
316 border = ((max_ - min_) % win_step_um) / 2
317 window_centers = np.arange(num_windows + 1) * win_step_um + min_ + border

Exception: get_spatial_windows(): win_step_um=400.0/win_scale_um=400.0/win_margin_um=-200.0 are too large for the probe size (depth range=705.0). You can try to reduce them or use rigid motion.

[samuelgarcia]

For small probe you should try rigid motion correction, if you think you really need non rigid then you should lower win_step_um and win_scale_um win_margin_um. Theses 3 params define the number of gaussian window for non rigid motion correction on depth axis.

[cwindolf]

Hi @desaponte for NP2.4, I'd recommend a couple of things. First off, I've had the best results when doing motion correction on each shank separately -- you may already be doing that, not sure. Second, if Sam's suggestion of using rigid motion doesn't suit your needs, there are parameters recommended for that probe in the DREDge paper which you can check out -- that would just mean setting win_step_um=100 and win_margin_um=-150. Sometimes trying 250 for the scale can help too. Feel free to send me an email if you want some more detailed advice on parameters! (email on my github profile)