forked from cortex-lab/Suite2P
-
Notifications
You must be signed in to change notification settings - Fork 1
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
defaulting to soft signal extraction without overlaps. Updated readme…
… file.
- Loading branch information
Showing
7 changed files
with
244 additions
and
40 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,182 @@ | ||
| function [ops, stat, Fcell, FcellNeu] = extractSignalsNoOverlaps(ops, m, stat) | ||
|
|
||
| ops.saveNeuropil = getOr(ops, 'saveNeuropil', 0); | ||
|
|
||
| Nk = numel(stat); | ||
|
|
||
| Ly = numel(ops.yrange); | ||
| Lx = numel(ops.xrange); | ||
|
|
||
| % make new set of basis functions (larger coverage) | ||
| ops.neuropilRange = 10; | ||
|
|
||
| S = getNeuropilBasis(ops, Ly, Lx, 'raisedcosyne'); % 'raisedcosyne', 'Fourier' | ||
| S = normc(S); | ||
|
|
||
| % S = m.S; | ||
|
|
||
| S = reshape(S, [], size(S, ndims(S))); | ||
| nBasis = size(S,2); | ||
|
|
||
| % initialize mask | ||
| maskNeu = ones(size(S,1), 1); | ||
|
|
||
| stat = getNonOverlapROIs(stat, Ly, Lx); | ||
|
|
||
| for k = 1:Nk | ||
| ix = stat(k).ipix(~stat(k).isoverlap); | ||
| maskNeu(stat(k).ipix)= 0; | ||
| if numel(ix)==0 | ||
| LtS(k,:) = 0; | ||
| else | ||
| LtS(k,:) = stat(k).lam(~stat(k).isoverlap)' * S(ix, :); | ||
| end | ||
| end | ||
|
|
||
| % add all pixels within X um | ||
| if isfield(ops, 'exclFracCell') && ops.exclFracCell>0 | ||
| H = fspecial('disk', round(ops.diameter * ops.exclFracCell)); | ||
| maskNeu = reshape(maskNeu, Ly, Lx); | ||
| maskNeu = imfilter(maskNeu, H, 'replicate'); | ||
| maskNeu = single(maskNeu(:) > 1-1e-3); | ||
| end | ||
| %% get signals | ||
| S = bsxfun(@times, S, maskNeu(:)); | ||
| StS = S' * S; | ||
| StS = StS + 1e-2 * eye(size(StS)); | ||
|
|
||
| nimgbatch = 2000; | ||
|
|
||
| ix = 0; | ||
| fclose all; | ||
| fid = fopen(ops.RegFile, 'r'); | ||
|
|
||
| tic | ||
| F = zeros(Nk, sum(ops.Nframes), 'single'); | ||
| Fneu = zeros(Nk, sum(ops.Nframes), 'single'); | ||
| if ops.saveNeuropil | ||
| Ntraces = zeros(nBasis, sum(ops.Nframes), 'single'); | ||
| end | ||
| % S = bsxfun(@times, S, maskNeu(:)); | ||
|
|
||
| [Ly Lx] = size(ops.mimg1); | ||
|
|
||
| mimg1 = ops.mimg1(ops.yrange, ops.xrange); | ||
| % find indices of good clusters | ||
| while 1 | ||
| data = fread(fid, Ly*Lx*nimgbatch, '*int16'); | ||
| if isempty(data) | ||
| break; | ||
| end | ||
| data = reshape(data, Ly, Lx, []); | ||
| data = data(ops.yrange, ops.xrange, :); | ||
| data = single(data); | ||
| NT = size(data,3); | ||
|
|
||
| % process the data | ||
| data = bsxfun(@minus, data, mimg1); | ||
| data = my_conv2(data, ops.sig, [1 2]); | ||
| data = bsxfun(@rdivide, data, m.sdmov); | ||
| data = single(reshape(data, [], NT)); | ||
|
|
||
| % | ||
| Ftemp = zeros(Nk, NT, 'single'); | ||
| for k = 1:Nk | ||
| ipix = stat(k).ipix(~stat(k).isoverlap)'; | ||
| if ~isempty(ipix) | ||
| Ftemp(k,:) = stat(k).lam(~stat(k).isoverlap)' * data(ipix,:); | ||
| end | ||
| end | ||
| F(:,ix + (1:NT)) = Ftemp; | ||
|
|
||
| Tneu = StS\(S' * data); | ||
| Ftemp2 = LtS * Tneu; | ||
| Fneu(:,ix + (1:NT)) = Ftemp2; | ||
|
|
||
| % Fneu(:,ix + (1:NT)) = m.LtS * Fdeconv(1+Nk:end, :); % estimated neuropil | ||
| % F(:,ix + (1:NT)) = Fneu(:,ix + (1:NT)) + Fdeconv(1:Nk, :); % estimated ROI signal | ||
|
|
||
| if ops.saveNeuropil | ||
| Ntraces(:,ix + (1:NT)) = Tneu; | ||
| end | ||
|
|
||
| ix = ix + NT; | ||
| if rem(ix, 3*NT)==0 | ||
| fprintf('Frame %d done in time %2.2f \n', ix, toc) | ||
| end | ||
| end | ||
| fclose(fid); | ||
| %% add the means back in to both neuropil and total | ||
| data = my_conv2(mimg1, ops.sig, [1 2]); | ||
| data = bsxfun(@rdivide, data, m.sdmov); | ||
| data = single(reshape(data, [], 1)); | ||
|
|
||
| scalefactors = nan(numel(stat),1); | ||
| Ftemp = zeros(Nk, 1, 'single'); | ||
| for k = 1:Nk | ||
| ipix = stat(k).ipix(~stat(k).isoverlap)'; | ||
| if ~isempty(ipix) | ||
| Ftemp(k,:) = stat(k).lam(~stat(k).isoverlap)' * data(ipix,1); | ||
| scalefactors(k) = mean(m.sdmov(ipix)); | ||
| end | ||
| end | ||
|
|
||
| Tneu = StS\(S' * data); | ||
| Ftemp2 = LtS * Tneu; | ||
|
|
||
| Fneu = bsxfun(@plus, Fneu, Ftemp2); % estimated neuropil | ||
| F = bsxfun(@plus, F, Ftemp); | ||
|
|
||
| Fneu = bsxfun(@times, Fneu, scalefactors); % estimated neuropil | ||
| F = bsxfun(@times, F, scalefactors); | ||
|
|
||
| %% get activity stats | ||
| indNoNaN = find(~ops.badframes); | ||
| ix = cumsum(~ops.badframes) + 1; | ||
| ix = ix(ops.badframes); | ||
| ix(ix>numel(indNoNaN)) = numel(indNoNaN); | ||
|
|
||
| F(:, ops.badframes) = F(:, indNoNaN(ix)); | ||
| Fneu(:, ops.badframes) = Fneu(:, indNoNaN(ix)); | ||
|
|
||
| % figure out the ICA coefficients here | ||
| ops.fs = getOr(ops, 'fs', ops.imageRate/ops.nplanes); | ||
| % | ||
| [coefNeu, inomax] = my_ica(F', Fneu', ops.fs, 0.7); | ||
| dF = F - bsxfun(@times, Fneu, coefNeu(:)); | ||
|
|
||
| % dF = F - Fneu; | ||
|
|
||
| sd = std(dF, [], 2); | ||
| sdN = std(Fneu, [], 2); | ||
|
|
||
| sk(:, 1) = skewness(dF, [], 2); | ||
| sk(:, 2) = sd./sdN; | ||
| sk(:, 3) = (max(dF, [], 2)-median(dF, 2))./sd; | ||
| sk(:, 4) = (prctile(dF, 95, 2)-median(dF, 2))./sd; | ||
|
|
||
| for j = 1:numel(stat) | ||
| stat(j).dFstat = sk(j,:); | ||
| stat(j).skew = sk(j,1); | ||
| stat(j).std = sk(j,2); | ||
| stat(j).maxMinusMed = sk(j,3); | ||
| stat(j).top5pcMinusMed = sk(j,4); | ||
| stat(j).blockstarts = [0 cumsum(ops.Nframes)]; | ||
| stat(j).iplane = ops.iplane; | ||
| stat(j).neuropilCoefficient = coefNeu(j); | ||
| end | ||
|
|
||
| %% | ||
| csumNframes = [0 cumsum(ops.Nframes)]; | ||
| Fcell = cell(1, length(ops.Nframes)); | ||
| FcellNeu = cell(1, length(ops.Nframes)); | ||
| for i = 1:length(ops.Nframes) | ||
| Fcell{i} = F(:, csumNframes(i) + (1:ops.Nframes(i))); | ||
| FcellNeu{i} = Fneu(:, csumNframes(i) + (1:ops.Nframes(i))); | ||
| end | ||
|
|
||
| if getOr(ops, 'saveNeuropil', 0) | ||
| S = reshape(S, numel(ops.yrange), numel(ops.xrange), Nbasis); | ||
| save(sprintf('%s/NEU_%s_%s_plane%d.mat', ops.ResultsSavePath, ... | ||
| ops.mouse_name, ops.date, ops.iplane), 'ops', 'S', 'Ntraces', '-v7.3') | ||
| end |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,12 @@ | ||
| function stat = getNonOverlapROIs(stat, Ly, Lx) | ||
|
|
||
|
|
||
| Mask = zeros(Ly, Lx); | ||
|
|
||
| for k = 1:numel(stat) | ||
| Mask(stat(k).ipix) = Mask(stat(k).ipix) + 1; | ||
| end | ||
|
|
||
| for k = 1:numel(stat) | ||
| stat(k).isoverlap = Mask(stat(k).ipix)>1; | ||
| end |