Subscripted Assignment Dimension Mismatch (error in line 369)

[sebastianmaruri]

When trying to open a .mat file containing micro electrode recording data using wave_clus I sometimes get the following error message:

Subscripted assignment dimension mismatch.

Error in wave_clus>load_data_button_Callback (line 369) clustering_results(:,2) = classes'; % GUI classes

Error in gui_mainfcn (line 95) feval(varargin{:});

Error in wave_clus (line 63) gui_mainfcn(gui_State, varargin{:});

Error while evaluating UIControl Callback

Does anyone know what might be missing from my file or any other way to fix this? Thanks a lot.

[ferchaure]

That looks like the algorithm isn't finding valid clusters, can you share with me the data? maybe after detecting spikes with Get_spikes()

[sebastianmaruri]

Sorry for taking so long. Here is one of the .mat files that caused that error message and its detected spikes (using get_spikes()) micro EEG_sorted_PC_links_2_CRAW_01___Medial.mat_channel21_depth =+2.zip Thanks for your help.

[ferchaure]

Hi sebastian, you have just 18 spikes, for that reason the algorithm don't work, if you try using using the batch file Do_clustering you will get a clear error message.

[sebastianmaruri]

Hi Fer. I get that, but I thought the toolbox was unable to open the file only if it had less than 15 spikes (which usually leads to the display of an error in line 230). Is there any difference between both errors (in line 369 and line 230)? Or would the solution be lowering the detection threshold for this one as well?

[ferchaure]

They are the same error. A fast fix could be change that 15 value to something like 50. Sometimes SPC can handle small amounts of spikes but the results won't be great.

Lowering the detection threshold, isn't always a good idea... otherwise you are just detecting peaks in the background noise and calling them spikes. If the channel has that small amount of spikes don't use it.

[sebastianmaruri]

Great, thanks for the advice. Do you think a detection threshold of 4 standard deviations is still acceptable, or should I leave it at 5 always?

[ferchaure]

4.5 is an acceptable value if you want to see enough multiunit activity or if the recording has low noise. If you only care only about nice single units choose 5.

[sebastianmaruri]

Thanks a lot. Regarding the cluster size, you suggested that any cluster with less than 50 spikes is no longer acceptable, right?

[ferchaure]

Doing clustering for 50 spikes as the total number in the recording is kind of useless. 50 spikes in a cluster after clustering, could be a sparse neuron in a short recording

[sebastianmaruri]

This has been really helpful, thanks you. I have one final question, are these numbers (detection threshold and spike count) based on literature or on your experience with spike sorting?

[ferchaure]

detection threshold: they are the typical values in the literature if you check that the threshold is related to the MAD and not to the standard deviation.

spike count: experience I guess, and the exact number could change with your objective and the length of your recording

[sebastianmaruri]

great, thanks