shortened open source data cannot work in kilosort

[xintangg]

hi, I download the neural signal data from neuropixel official website (http://data.cortexlab.net/singlePhase3/data/) and it's a 385/*36000000 matrix. 385 stands for channel and 36000000 stands for timepoints. I was trying to sort these signals and it worked with the original downloaded raw data. Strange things happened when I used the processed raw data: due to the limited computing resource, I 'cut' the original matrix to a 385/*600000 matrix with matlab and then set it as the input of kilosort:

fileID=fopen('F:\neuropixel_dataexample\rawDataSample.bin','r');
rawdata=fread(fileID);
fclose(fileID);
shortdata=reshape(rawdata,[],385);
shortdata1=shortdata(1:600000,:);
shortdata1=int16(shortdata1);
shortdata1=shortdata1';
fileID1=fopen('F:\neuropixel_dataexample\shortrawDataSample.bin','w');
fwrite(fileID1,shortdata1);
fclose(fileID1);

Then the kilosort's GUI finished sorting and reported a error:"error saving data for phy reference to nonexistent field 'wphy'" and the produced rez.mat file contains no st array. (However, the unprocessed/uncut raw data works well with kilosort.

Is there anything wrong?

[marius10p]

The raw binary file is in order chan1 time1, chan2 time1... chan385 time1, chan1 time 2 etc, which Matlab reads in Fortran order. So your reshape operation should be like this

shortdata = reshape(rawdata, 385, []);

and then you don't need the extra transpose in line

shortdata1 = shortdata1';

[marius10p]

Closed due to lack of activity.