check_probe_do_not_overlap only checks x and y

[AOONIH]

Hi, I am using a double sided probe so my coordianates are in 3D. I make sure to sort the probe by group, however, I run into the issue when trying to load the recording with the 3D probegroup as check_probe_do_not_overlap raises the exception "Exception: Probes are overlapping! Retrieve locations of single probes separately". Further inspetion shows that the function only checks the first 2 dimensions and not the 3rd for a 3D probe.

Is there away to bypass this when reading in the recordings?

[zm711]

I'm not sure about this. I don't know of any spike sorters that explicitly handle 3d probes so normally we would say sort by probegroup. In your case sorting by probegroup wouldn't quite be fair because it is likely that your contacts will detect the exact same units. What is your 3d dimension in size (um's)? I can't think of a truly fair solution. You could offset one of your sides by like .2 microns so that they don't show up the same, but that also isn't fair.

Unless, by double sided you mean multishanked with a spread on the order of 100s of microns in which case there are work arounds. Could you provide a little more info about what the probe looks like?

[AOONIH]

I'm using this probe from cambridge neurotech: https://www.cambridgeneurotech.com/assets/files/ASSY-325D-P-1_P-2-map.pdf To generate the pribe I create a probegroup where the sites on the 2 sides are offset by 30 um in the z dim and rotated (this is more of a probeinterface topic). Regarding the chek overlapping func, I think the function should use as all the probe dimensions to determining whether the site are overlapping and not just the first 2.

I slightly modified the function and can now load the recording: def check_probe_do_not_overlap(probes): """ Ensure that multiple probes do not have overlapping contact positions in N-dimensional space by using axis-aligned bounding boxes.

    Parameters
    ----------
    probes : list
        A list of probes, each with a `contact_positions` attribute 
        as a NumPy array of shape (n_contacts, n_dims)

    Raises
    ------
    Exception
        If any probes have overlapping contact positions
    """
    for i in range(len(probes)):
        probe_i = probes[i]
        positions_i = probe_i.contact_positions
        min_i = np.min(positions_i, axis=0)  # (n_dims,)
        max_i = np.max(positions_i, axis=0)  # (n_dims,)

        for j in range(i + 1, len(probes)):
            probe_j = probes[j]
            positions_j = probe_j.contact_positions  # shape (n_contacts, n_dims)

            # Check if each contact in probe_j is inside the bounding box of probe_i
            in_bounds = np.all((positions_j >= min_i) & (positions_j <= max_i), axis=1)
            if np.any(in_bounds):
                raise Exception(
                    f"Probes {i} and {j} are overlapping! Retrieve locations of single probes separately.")

As an aside, I thought trisdeclous or mountatinsort5 would allow for 3D positions. I don't know if there is any specific workflow for this

[samuelgarcia]

Hi Dammy. Zach is right no sorter handle correctly the double side probes geometry And it even worst : probeinterface do not handle it! If you keep contact at the same position you will have the warning you already experience. The only way is to have the B side virtualy in the same plan and far away enough from the A side. The Z axis is really poorly handled. Sorry for that.