question on findWheelMoves3

[yilmazduygu]

Hi,
I'm using this script to extract movements from my wheel data, first of, thanks a lot for putting it here! I wanted to get precise movement offsets so I attempted to adapt what you've done in onsets:

`offsetSamps = find(isMoving(1:end-1) & ~isMoving(2:end));

wheelToepOffsetsDev = zeros(length(offsetSamps), tThreshSamps);
c = 0; cwt = 0;
while isempty(onsetSamps)
i2proc = (1:p.batchSize) + c;
[icomm] = intersect( i2proc(1:end-1), offsetSamps );
[, itpltz] = intersect( i2proc(1:end-1), offsetSamps );
i2proc(i2proc > length(t)) = [];
if ~isempty(icomm)
w2e = hankel(pos(i2proc), nan(1,tThreshSamps));
w2e = abs(bsxfun(@minus, w2e, w2e(:,1)));
wheelToepOnsetsDev(cwt + (1:length(icomm)), :) = w2e(itpltz, :);
cwt = cwt + length(icomm);
end
c = c + p.batchSize;
if i2proc(end) >= onsetSamps(end), break, end
end
warning('on', 'MATLAB:hankel:AntiDiagonalConflict')

hasOffset = wheelToepOffsetsDev > p.posThreshOffset;
[a, b] = find(~fliplr(hasOffset));
offsetLags = tThreshSamps+accumarray(a(:), b(:), [], @min);
offsetSamps = offsetSamps + offsetLags;
offsets = t(offsetSamps);
`

This gives me lots of 'MATLAB:hankel:AntiDiagonalConflict' warnings. As it might show, I don't know much math, so I don't know how ok it is to ignore it. I'd appreciate any help

Thanks in advance!

[nsteinme]

Hm, I think @k1o0 would have to comment on the details here, this has been too long ago for me to remember. However I wanted to just comment that you'd have to be very careful in interpreting wheel offsets, since mice can often spin the wheel and let go, letting it spin for a while after they stopped moving their limbs. You might be better off in general looking at video of the mice if you're interested in this.

[yilmazduygu]

I see, that makes sense. I will figure out an alternative confirmatory measure using the videos. But yes, if @k1o0 could comment on the details, it would help a lot.

Thank you!

[yilmazduygu]

Btw is there a publication that I could cite for this?

[nsteinme]

It was used in Steinmetz 2019 - https://www.nature.com/articles/s41586-019-1787-x - and I believe there's a supplemental figure panel showing how well the algorithm worked on the wheel traces - so that would probably be best. Thanks -

…

On Thu, Jan 26, 2023 at 9:28 PM G. Duygu Yilmaz ***@***.***> wrote: Btw is there a publication that I could cite for this? — Reply to this email directly, view it on GitHub <#6 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABZ5IP3RLVVA3AYLHRDCBA3WUNMG7ANCNFSM6AAAAAAUG7QU3A> . You are receiving this because you commented.Message ID: ***@***.***>