Find the amplitude with sptool

I posted this for days ago. Then a guy called star strider came up with a answer. It was with i could use a filter or something. But by accident i clicked accept answar to anather guys answar( i dont know why) and then the ansawr from star strider disappered(sorry star strider). So i hope its okay i post again.

Im writing a code for calculating BP. My problem is i don't know how to solve this. I think it will be a lot easier if i knew how to "rotate" my graph(there is a picture attached). My data is number 1 in the picture. I would like to have it as no. 2. I also attached my data.

I hope one of your guys can help me.

In advance thanks :)

Answer 1

[Star Strider]

I delete my Answers that aren’t Accepted or Voted. Others do as well.

I will help you design a filter for your data, but I need a sample of it first. I would use a bandpass filter, so I could simultaneously filter out the baseline drift and high-frequency noise.

If it is not stated or available in your signal record (a .mat file is preferable), I will need your uniform sampling interval ‘Ts’ or sampling frequency ‘Fs’. I would of course want the units, but I would anticipate the time variable in seconds or fractions of seconds, and the BP in Torr or mm Hg. I will do my best do explain my filter design choices and code.

[saxocat]

Sorry i thought i attached it. The sample rate is 200 Hz. The time is in seconds and pressure is in mmHg. Thank you so much for helping me :)

[Star Strider]

Sorry for the delay. Life intrudes.

I’m happy with this filtered signal. I did my best to pass all the interesting information and as little of the noise as possible. I couldn’t always see the dichrotic notch in the original signal, and did my best to preserve those I was able to identify.

The code:

fidi = fopen('BP.txt','r');
D  = textscan(fidi, '%f %f', 'HeaderLines',1);
T  = D{1};
BP = D{2};
T  = T-T(1);                        % Start Time At Zero

figure(1)
plot(T, BP)                         % Original Signal
grid
xlabel('Time (s)')
ylabel('BP (Torr)')

Ts = mean(diff(T));                 % Sampling Time & Frequency
Fs = 1/Ts;
Fn = Fs/2;

FTBP = fft(BP)/length(BP);          % Examine Frequency Content
Fv = linspace(0, 1, fix(length(BP)/2)+1)*Fn;
Iv = 1:length(Fv);

figure(2)
plot(Fv, abs(FTBP(Iv)))
grid
axis([0  15    ylim])
xlabel('Frequency (Hz)')
ylabel('Amplitude')

Wp = [0.3 15]/Fn;                   % Design Bandpass Filter
Ws = [0.5 1/0.5].*Wp;
Rp =  1;
Rs = 10;
[n,Wn] = buttord(Wp, Ws, Rp, Rs);
[b, a] = butter(n, Wn);
[sos, g] = tf2sos(b, a);

figure(3)
freqz(sos, 2048, Fs)                % Filter Characteristics

BPF = filtfilt(sos, g, BP);         % Filter ‘BP’ Signal

figure(4)
plot(T, BPF)                        % Plot Filtered Signal
grid
axis([0  10    ylim])

The axis call in figure(4) allowed me to see the details to be sure the filtered signal looked correct. Expand its x-limits, or comment it out to see the entire signal.

[Star Strider]

saxocat’s ‘Answer’ moved here...

wauw. its actually works :) It finds the MAP +- 2mmHG compared to loggerPro. Now I will try work on algortinm for finding the diastolic and systolic from the maxamplitude. Thank you again:)

[Star Strider]

My pleasure!

To find the systolic and diastolic peaks from the filtered signal, use the findpeaks function. Use it on the original ‘BPF’ signal to find the locations of the systolic peaks, then to find the locations of the diastolic troughs, invert it as ‘-BPF’. Use those two sets of indices on the original ‘BPF’ signal to find the systolic and diastolic pressures. The findpeaks function allows you to determine these values in a few lines of code.

If my answer solved your problem, please Accept it.

[saxocat]

I thought the easiest way was to find the locating point where slope becomes most steep. Another way was to fit a line to the oscillatory peak data. But i dont know how to do that..

How do i use the method you descripe? The picture shows the peaks,

[Star Strider]

My comment related to finding the individual peaks (systolic) and troughs (diastolic) pressures.

Append this code after ‘figure(4)’. It detects and plots the systolic and diastolic peak and trough pressures:

[SBPF, IxS] = findpeaks(BPF, 'MinPeakWidth',25);    % , 'MinPeakDistance',160
[DBPF, IxD] = findpeaks(-BPF, 'MinPeakWidth',25);

figure(5)
plot(T, BPF)
hold on
plot(T(IxS), BPF(IxS), '^r')
plot(T(IxD), BPF(IxD), 'vg')
hold off
grid
% axis([0  10    ylim])

You may want to experiment with the various ‘Name-Value Pair Arguments’ to increase the robustness of the detection.

I notice a correlation between the original blood pressure trace and the systolic-diastolic differences. I don’t know what your experiment is, but since the heart rate does not appear to change, this suggests preload reduction to me, and that you likely volume-overloaded your test animals, or less likely bled them.

[saxocat]

Sorry for my late reply :)

My experiment is determine the effect caffeine and sugar has on heart rate, blood pressure and concentration...on human and the data is from my BP. Preload reduction...what is that..is it something serious ?

[saxocat]

I can i ask you to explain the code below:

FTBP = fft(BP)/length(BP);          % Examine Frequency Content
Fv = linspace(0, 1, fix(length(BP)/2)+1)*Fn;
Iv = 1:length(Fv);

[Star Strider]

No worries!

In this context, ‘serious’ depends on whether the heart is healthy or decompensated. In a healthy heart (and in the absence of exogenous vascular extremes such as haemorrhagic shock or volume expansion), there is a wide range of adaptive ability in the cardiovascular system generally. ‘Preload’ essentially refers to the end-diastolic volume, closely linked to venous return, a function of venous muscular tone and intravascular volume. ‘Afterload’ is the load that opposes shortening of the cardiac muscle in systole, most notably systemic vascular resistance.

Since caffeine is a potent inhibitor of cyclic adenosine monophosphate phosphodiesterase (C-AMP), as are all the methylxanthines, you would expect it to have a positive inotropic and chronotropic effect, simulating an alpha-adrenergic effect. This would affect both preload and afterload, but would not exceed healthy physiological compensation.

However ‘sugar’ is different. It might well depend on the type of sugar (glucose, sucrose, etc.), since sucrose has a fructose monomer and fructose is handled differently than glucose. (I will leave the biochemistry of fructose metabolism to you.) Apparently, sucrose (through its fructose moiety) has an acute cardiovascular effect, at least in rats, that may have effects synergistic to those of caffeine (Ref. wrong white crystals: not salt but sugar as aetiological in hypertension and cardiometabolic disease Open Heart. 2014; 1(1): e000167). I am not aware of an acute cardiovascular effect of glucose, other than osmotic diuresis and dehydration with plasma glucose concentrations in excess of about 270 mg/dL, 15 mM.

Yours appears to me to be an important study. In addition to the Signal Processing Toolbox, you are going to quickly become expert with the Statistics Toolbox (and perhaps the Global Optimization Toolbox) as well.

I would very much appreciate a PDF of your paper when it has been accepted for publication.

[Star Strider]

I missed your later Comment while I was responding to your earlier one.

That code takes the discrete-time Fourier transform of the blood pressure signal. The ‘Fv’ assignment is the frequency vector corresponding to the fft results, and the ‘Iv’ assignment is the index vector limiting the plotted data to the frequencies corresponding to the fft results. It makes the plot call easier to write.