quickSort/megaSort.m at master · BellWilson2016/quickSort · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
function megaSort()

    baseName = '~/Desktop/Data/140811/';
    wildcard = 'RL140811_023_0[1,2,3,4]*.mat';

    nClusters = 3;
    negPeaks = true;      % Negative going spikes
    posPeaks = false;
    peakLimit = 1.8;      % SD (abs. value) 1.2 for ab1,
    spikeWidth = .005;    % Seconds
    LPF =  1000;          % Hz
    HPF =  100;           % Hz
    plotOn = true;

    fileList = jdir([baseName,wildcard]);
    dataMatrix = [];
    timeCodes = [];
    for fileN = 1:length(fileList)

        fileName = fileList(fileN).name
        load([baseName,fileName]);

        % Filter data, store
        data.LPF.freq = LPF;
        data.LPF.h = fdesign.lowpass('N,F3dB',4,data.LPF.freq/(data.sampleRate/2));
        data.LPF.d = design(data.LPF.h,'butter');
        data.HPF.freq = HPF;
        data.HPF.h = fdesign.highpass('N,F3dB',4,data.HPF.freq/(data.sampleRate/2));
        data.HPF.d = design(data.HPF.h,'butter');

        lpV = filtfilt(data.LPF.d.sosMatrix,data.LPF.d.ScaleValues,data.V);
        data.fV = filtfilt(data.HPF.d.sosMatrix,data.HPF.d.ScaleValues,lpV);
        data.dVdT = [diff(data.fV);0];
        data.spikeWidth = spikeWidth;


        % Find peaks, only take biggest ones.
        spikeHalfWidth = round(spikeWidth/2*data.sampleRate);
        dataMean = mean(data.dVdT);
        dataStd  = std(data.dVdT);
        peakListN = []; peakListP = []; peakHeightsN = []; peakHeightsP = [];
        ixN = []; ixP = [];
        if negPeaks
            [peakListN,peakHeightsN] = peakFind(data.dVdT,[0,1]);
            ixN = find((peakHeightsN - dataMean)./dataStd < -peakLimit);
        end
        if posPeaks
            [peakListP,peakHeightsP] = peakFind(data.dVdT,[1,0]);
            ixP = find((peakHeightsP - dataMean)./dataStd > peakLimit);
        end
        peakList = [peakListN(ixN);peakListP(ixP)];
        peakHeights = [peakHeightsN(ixN);peakHeightsP(ixP)];

        % Remove peaks too close to start or end
        remIX = find((peakList <= 2*spikeHalfWidth) | (peakList >= (length(data.dVdT) - 2*spikeHalfWidth)));
        peakList(remIX) = [];
        peakHeights(remIX) = [];


        % Assemble trace snippets into a matrix for dimensionality reduction
        for peakN = 1:length(peakList)
            saveSnip = data.dVdT((peakList(peakN) - spikeHalfWidth):(peakList(peakN) + spikeHalfWidth));
            dataMatrix(end+1,:) = saveSnip';
            timeCodes(end+1) = fileN*length(data.dVdT) + peakList(peakN);
        end

        % size(dataMatrix)
    end

%     timeCodes = timeCodes.*10^-7;
%     dataMatrix = cat(2,dataMatrix,timeCodes');
%     size(dataMatrix)

    % Reduce dimensionality
    PCAdim = 50;
    perplexity = getPerplexity(size(dataMatrix,1));
    theta = .3;
    mappedX = fast_tsne(dataMatrix, PCAdim, perplexity, theta);
%     [COEFFS, SCORES] = princomp(dataMatrix);
%     mappedX = SCORES(:,1:2);


    % Cluster
    Z = linkage(mappedX,'ward','euclidean');
    IDX = cluster(Z,'maxclust',nClusters);
%     IDX = kmeans(mappedX,nClusters);

    % Sort clusters by peak values
    for clustN=1:nClusters
        ix = find(IDX == clustN);
        meanWaveform = mean(dataMatrix(ix,:),1);
        maxWave(clustN) = abs(meanWaveform(spikeHalfWidth+1));
    end
    [B,IX] = sort(maxWave,'descend');
    for clustN=1:nClusters
        findClust = IX(clustN);
        ix = find(IDX == findClust);
        newIDX(ix) = clustN;
    end
    IDX = newIDX;


%     % Save cluster info to the data structure
    data.spikeSamples = peakList;
    data.spikeClusters = IDX;
    data.spikeEmbedding = mappedX;
%
%     % Make well-separated spike avg. waveforms.
%     data = makeSpikeAvg(data);

    % Plot if required
    if plotOn
        figure;
        subplot(2,2,1);
        plotEmbeddings(data);


        for clustN=1:nClusters

            ix = find(IDX == clustN);
            meanWaveform = mean(dataMatrix(ix,:),1);
            subplot(2,2,2);
            plot(dataMatrix(ix,:)','Color',pretty(clustN)); hold on;
            subplot(2,2,3);
            plot(meanWaveform,'Color',pretty(clustN)); hold on;

        end
    end