-
Notifications
You must be signed in to change notification settings - Fork 17
/
perceive_ecg.m
175 lines (169 loc) · 6.2 KB
/
perceive_ecg.m
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
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
function ecg=perceive_ecg(data,fs,plotit)
% ecg = perceive_ecg(data,fs,plotit)
% data = raw signal (required)
% fs = sampling rate (default = 250)
% plotit = create plots (default = 1)
%% Defaults
if ~exist('fs','var');fs = 250;end
if ~exist('plotit','var');plotit=1;end
ns = length(data);
%% First pass using cross-correlation
dwindow=round(fs); % segment window size (500 ms)
dmove = fs; % moving window size (100 ms)
i=[1+dwindow:dmove:ns-dwindow-1]'; % segment indices
if ~isempty(i)
for a = 1:length(i)
x(a,:) = data([i(a)-dwindow:i(a)+dwindow]); % epoch data
end
disp(['...running cross-correlation on ' num2str(a) ' segments...'])
ndata = zeros(1,4*dwindow); % template for adjust to xcorr lags (4*segment window size)
nt=linspace(-2*dwindow/fs,size(ndata,2)/fs-2*dwindow/fs,size(ndata,2)); % time axis
ndata(1,perceive_sc(nt,0):perceive_sc(nt,0)+size(x,2)-1)=x(1,:); % initialize with first segment
n=0; % run through remaining segments and find xcorr lags, align data
for a = 2:size(x,1)
[r,l]=xcorr(nanmean(ndata,1),x(a,:),fs);[~,mi]=max(r);tlag = l(mi);
if tlag >0;n=n+1;ndata(n,tlag:tlag+size(x,2)-1)=x(a,:);end
end
mdata = nanmean(ndata,1); % average aligned data
%% find ECG peak characteristics in xcorr aligned data
[absm,imax]=findpeaks(abs(mdata),'SortStr','descend','NPeaks',15); np=0.05;iim=[];iin=[];
while isempty(iim) || isempty(iin)
np=np+.025;pkrange=imax(1)-round(fs*np):imax(1)+round(fs*np);
[~,iim,wm]=findpeaks(mdata(pkrange),'SortStr','descend','NPeaks',1);
[~,iin,wn]=findpeaks(-mdata(pkrange),'SortStr','descend','NPeaks',1);
end
iim=iim+pkrange(1)-1;iin=iin+pkrange(1)-1;pdif=absm(1)./nansum(absm(2:end))*100;
if iin(1)<iim(1)
ii1 = iin(1);ii2 = iim(1);w1=wn(1);w2 = wm(1);
else
ii1=iim(1);ii2=iin(1);w1=wm(1);w2=wn(1);
end
ecg_cut=ii1-round(w1(1)):ii2+round(w2(1));
ecg.proc.template1 = mdata(ecg_cut); % first template
disp('...ecg template 1 generated...')
%% Run temporal correlation across samples
corrdata=[];
for a = 1:ns-size(ecg_cut,2)-1
corrdata(:,a) = data(1,a:a+size(ecg_cut,2)-1)';
end
r = corr(corrdata,ecg.proc.template1').^2;
ecg.proc.r = r;
disp('...first temporal correlation done...')
%% adjust r threshold to maximize HR associated peak identification
h=[max(findpeaks(r))*.95,...
max(findpeaks(r))*.90,...
max(findpeaks(r))*.85,...
max(findpeaks(r))*.80,...
max(findpeaks(r))*.75,...
max(findpeaks(r))*.70,...
max(findpeaks(r))*.65,...
max(findpeaks(r))*.60,...
max(findpeaks(r))*.55,...
max(findpeaks(r))*.50];
thr=[];
for a=1:length(h)
[~,ix]=findpeaks(r,'MinPeakHeight',h(a),'MaxPeakWidth',round(0.1*fs));
if ~isempty(ix);dd=60./(diff(ix)/fs);thr(a)=nansum(dd>55&dd<120)./std(dd);end
end
[~,ithr]=max(thr);ecg.proc.thresh=h(ithr);
disp('...first threshold adjusted...')
%% find peaks from temporal correlations
[~,i]=findpeaks(r,'MinPeakHeight',ecg.proc.thresh,'MinPeakDistance',round(fs/2));
%% start over using the found peaks by aligning to the found peaks
for a = 1:length(i)
try ndata2(a,:)=data(i(a)-round(.05*fs):i(a)+round(.1*fs));end
end
ecg.proc.template2 = nanmean(ndata2,1);
disp('...realigned and generated template 2...')
%% run temporal correlation on second template
corrdata=[];
for a = 1:ns-size(ecg.proc.template2,2)-1
corrdata(:,a) = data(1,a:a+size(ecg.proc.template2,2)-1)';
end
r2 = corr(corrdata,ecg.proc.template2').^2;
ecg.proc.r2 = r2;
disp('...temporal correlation on second template done...')
%% readjust threshold
h=[max(findpeaks(r2))*.95,...
max(findpeaks(r2))*.90,...
max(findpeaks(r2))*.85,...
max(findpeaks(r2))*.80,...
max(findpeaks(r2))*.75,...
max(findpeaks(r2))*.70,...
max(findpeaks(r2))*.65,...
max(findpeaks(r2))*.60,...
max(findpeaks(r2))*.55,...
max(findpeaks(r2))*.50];
for a=1:length(h)
[~,ix]=findpeaks(r2,'MinPeakHeight',h(a),'MaxPeakWidth',round(0.1*fs));
if ~isempty(ix),dd=60./(diff(ix)/fs);thr2(a)=nansum(dd>55&dd<120)./std(dd);end
end
[~,ithr2]=max(thr2);ecg.proc.thresh2=h(ithr2);
%% find peaks based on second threshold
[pks,i]=findpeaks(r2,'MinPeakHeight',ecg.proc.thresh2,'MaxPeakWidth',round(0.1*fs));
disp('...final ECG artefact detection done...')
%% save info
ecg.stats.intervals = 60./(diff(i)./fs);
ecg.hr=60/(nanmedian(diff(i))/fs);
ecg.nandata = data;
ecg.cleandata=data;
cbins = zeros(size(data));
if numel(pks)
for a = 1:length(pks)
tss=size(ecg.proc.template2,2);
ic=i(a):i(a)+tss-1;
% keyboard
mirrorrange=[i(a):-1:(i(a)-round(tss/2)+2) (i(a)+tss+round(tss/2))-1:-1:(i(a)+tss)];
try
ecg.cleandata(ic)=data(mirrorrange);
catch
ecg.cleandata(ic)=0;
end
ecg.stats.n = numel(pks);
cbins(ic)=1;
end
else
ecg.stats.n = 0;
end
ecg.nandata(find(cbins))=nan;
ecg.ecgbins = cbins;
ecg.stats.pctartefact = nansum(cbins)/ns*100;
ecg.stats.msartefact = nansum(cbins)/fs;
ecg.stats.ecglength = length(ecg.proc.template2)/fs;
%% decide on detection
detstring = {'Unreliable or no ECG','Consistent ECG'};
if (ecg.hr<55 || ecg.hr>120) || ecg.stats.n <= 0.5*ns/fs || (ii2-ii1)/fs > 0.075 || pdif < 20
disp('No reliable ECG detection possible.')
ecg.detected = 0;
else
ecg.detected = 1;
end
disp([detstring{ecg.detected+1} ' detected.'])
%% plot
if plotit
t = linspace(0,ns/fs,ns);
[~,f,rpow]=perceive_fft(data(find(~isnan(data))),fs,fs*2);
[~,f,rnpow]=perceive_fft(ecg.cleandata(find(~isnan(ecg.cleandata))),fs,fs*2);
nt2 = linspace(-.05,.1,size(ecg.proc.template2,2));
% perceive_figure('Ecg');
figure
subplot(2,2,1);
plot(nt2,ndata2','linewidth',0.1,'color',[.9 .9 .9]);
hold on
plot(nt2,ecg.proc.template2,'color','k');
xlabel('Time [s]');ylabel('Amplitude');xlim([-.02 .1]);
title([detstring{ecg.detected+1} ' detected.'])
subplot(2,2,2)
plot(f,rpow,f,rnpow,'linewidth',2);
xlim([4 30]);legend('original','cleaned'); xlabel('Frequency [Hz]')
ylabel('Relative spectral power [%]');title([' HR: ' num2str(ecg.hr,3) '/min N = ' num2str(ecg.stats.n,3)]);
subplot(2,2,3:4);
plot(t,data,'color','r'); hold on
plot(t,ecg.cleandata,'color','k');
legend('original','cleaned');ylabel('Amplitude');xlabel('Time [s]')
end
else
warning('Insufficient data length for ECG correction.')
ecg=[];
ecg.cleandata =nan(size(data));
end