Closed mndiguilmi closed 5 months ago
Hi @mndiguilmi,
I think Kilosort is suited to have linear probes (neuropixels-like). When I use a multi-shank probe, I design the file so that every shank is deeper in the y-axis (ordinate axis).
Look at this example of a probe with four shanks, each containing 16 electrodes:
%=========================================================================
% createE1Probes.m
%
% Creates the coordinates of the ASSY77_E1 probes of Cambridge Neurotech
% It will design it as a linear probe (one probe after each other)
% because both KiloSort2 and Phy are designed better for linear probes.
% it will not affect the sorting since the probes are independent.
%=======================================================================`
%% Four shanks x-separated 250 um
fname = '~/SiliconProbes/Kilosort2/configFiles/64E1.mat';
Nchannels = 67; % Total number of channels
connected = true(Nchannels,1); % zero if bad channel
connected(65:67) = 0; % AUX channels
chanMap = 1:Nchannels;
chanMap0ind = chanMap - 1; % zero-index channel
% x-location of 16 electrodes in shanks
xloc = [0;70;5;65;10;60;15;55;20;50;25;45;30;40;35;35];
xlocAUX = ones(3,1)*35;
xcoords = vertcat(xloc, xloc, xloc, xloc, xlocAUX);
% y-location of 16 electrodes shanks
ylocA = [0;-25;-40;-65;-80;-105;-120;-145;-160;-185;-200;-225;-240;-265;-280;-305];
ylocB = ylocA - 600; % 600 um deeper
ylocC = ylocB - 600;
ylocD = ylocC - 600;
% Auxiliary channels far away
ylocAUX = [-4600;-4650;-4700;];
ycoords = vertcat(ylocA, ylocB, ylocC, ylocD, ylocAUX);
% not used in Kilosort2 (see https://github.com/MouseLand/Kilosort2/issues/155)
kcoords = ones(Nchannels,1);
fs = 30000;
save(fname, ...
'chanMap','connected','xcoords','ycoords','kcoords','chanMap0ind','fs')
fprintf('Creating %s \n', fname)
I don't think Kilosort has had functionality for multiple shanks since v1.0. What you could do is increase the x-distance between each shank to visually separate them in Phy (unless the inter-shank distance truly is 125um, which I would not expect!). This is the approach I take for tetrodes. Or, as @JoseGuzman suggested, orient your shanks in depth, then sort in phy by depth.
Wont assign kcoords as shank ID help? I saw from aonther issue thread and this is suggested by one of the deveopers. However I'm curios to try this way. I'm using ks3.
Hi
I’m currently use a Neuronexus probe A4x4 (16 ch), 4 shanks with 4 channels each one (https://www.neuronexus.com/files/catalog/2021-Probe-Catalog.pdf; page 34).
Assuming the following distribution of the Neuronexus channels: (from the left to the right: shank 1 -> 4; channels distribution from down to top: 6, 2, 3, 1 (shank 1); 5, 8, 4, 7 (shank 2) 9, 12, 10, 13 (shank 3); 15, 11, 16, 14 (shank 4))
I made the following channel map for Kilosort 2
function ops=make_16_A4x4_MD_channelMap(fpath,connectedOriginal) chanMap = [6 2 3 1 5 8 4 7 9 12 10 13 15 11 16 14]; connected=true(16,1);
for i=1:length(connected) connected(i)=connectedOriginal(chanMap(i)); end
xcoords=[0 0 0 0 125 125 125 125 250 250 250 250 375 375 375 375]; ycoords=[0 50 100 150 0 50 100 150 0 50 100 150 0 50 100 150]; kcoords = [1 1 1 1 2 2 2 2 3 3 3 3 4 4 4 4];
fs = 20000;
ops.fs = fs;
ops.NchanTOT = 16;
ops.Nchan = sum(connected);
save(fullfile(fpath, 'chanMap.mat'), 'chanMap', 'connected', 'xcoords', 'ycoords', 'kcoords', 'fs')
When I open the sorted data on Phy, the ProbeView looks OK however all channels appear on the same shank, sh 0 (see below). I’m not really sure if it’s a real problem or not and if it’s a problem in KS or Phy. On the other hand, I’d like to increase the channels for my recording (32-64 ch) using multiple shanks, but I’d like to fix this trouble before.
Thanks!
Mariano