Comprehensive PPG Analysis

In this tutorial we will learn how to extract biomarkers from a photoplethysmogram (PPG) signal.

Our objectives are to:

Detect the standard fiducial points on PPG pulse waves

Calculate pulse wave biomarkers from the fiducial points

Saving data in different data format

You can use the sample PPG data by downloading it from the following link: Sample PPG data.

Setup

Import Python packages:

Install the pyPPG toolbox for PPG analysis

pip install pyPPG==1.0.41

Import required components from pyPPG

from pyPPG import PPG, Fiducials, Biomarkers
from pyPPG.datahandling import load_data, plot_fiducials, save_data
import pyPPG.preproc as PP
import pyPPG.fiducials as FP
import pyPPG.biomarkers as BM
import pyPPG.ppg_sqi as SQI

Import other packages

import numpy as np
import sys
import json
import pandas as pd

Setup input parameters:

The following input parameters are inputs to the pyPPG.example module (see the documentation for further details).

data_path = "Sample_PPG_MAT_125Hz.mat" # the path of the file containing the PPG signal to be analysed
start_sig = 0 # the first sample of the signal to be analysed
end_sig = -1 # the last sample of the signal to be analysed (here a value of '-1' indicates the last sample)
savingfolder = 'temp_dir'
savingformat = 'csv'

Loading a raw PPG signal:

# Load the raw PPG signal
signal = load_data(data_path=data_path, start_sig=start_sig, end_sig=end_sig, use_tk=False)
signal.v = signal.v [0:20*signal.fs] # 20 second long signal to be analysed

Plot the raw PPG signal:

# import plotting package
from matplotlib import pyplot as plt

# setup figure
fig, ax = plt.subplots()

# create time vector
t = np.arange(0, len(signal.v))/signal.fs

# plot raw PPG signal
ax.plot(t, signal.v, color = 'blue')
ax.set(xlabel = 'Time (s)', ylabel = 'raw PPG')

# show plot
plt.show()

PPG signal processing

Prepare the PPG data:

Filter the PPG signal and obtain first, second and third derivatives (vpg, apg, and jpg respectively).

signal.filtering = True # whether or not to filter the PPG signal
signal.fL=0.5000001 # Lower cutoff frequency (Hz)
signal.fH=12 # Upper cutoff frequency (Hz)
signal.order=4 # Filter order
signal.sm_wins={'ppg':50,'vpg':10,'apg':10,'jpg':10} # smoothing windows in millisecond for the PPG, PPG', PPG", and PPG'"

prep = PP.Preprocess(fL=signal.fL, fH=signal.fH, order=signal.order, sm_wins=signal.sm_wins)
signal.ppg, signal.vpg, signal.apg, signal.jpg = prep.get_signals(s=signal)

Plot the derived signals

# setup figure
fig, (ax1,ax2,ax3,ax4) = plt.subplots(4, 1, sharex = True, sharey = False)

# create time vector
t = np.arange(0, len(signal.ppg))/signal.fs

# plot filtered PPG signal
ax1.plot(t, signal.ppg)
ax1.set(xlabel = '', ylabel = 'PPG')

# plot first derivative
ax2.plot(t, signal.vpg)
ax2.set(xlabel = '', ylabel = 'PPG\'')

# plot second derivative
ax3.plot(t, signal.apg)
ax3.set(xlabel = '', ylabel = 'PPG\'\'')

# plot third derivative
ax4.plot(t, signal.jpg)
ax4.set(xlabel = 'Time (s)', ylabel = 'PPG\'\'\'')

# show plot
plt.show()

Store the derived signals in a class

# Initialise the correction for fiducial points
corr_on = ['on', 'dn', 'dp', 'v', 'w', 'f']
correction=pd.DataFrame()
correction.loc[0, corr_on] = True
signal.correction=correction

# Create a PPG class
s = PPG(signal)

Identify fiducial points:

Initialise the fiducials package

fpex = FP.FpCollection(s=s)

Extract fiducial points

fiducials = fpex.get_fiducials(s=s)

Display the results

print("Fiducial points:\n",fiducials + s.start_sig) # here the starting sample is added so that the results are relative to the start of the original signal (rather than the start of the analysed segment)

Plot fiducial points:

# Create a fiducials class
fp = Fiducials(fp=fiducials)

# Plot fiducial points
plot_fiducials(s, fp, savingfolder, legend_fontsize=12)

PPG fiducial points

Calculate PPG SQI:

# Get PPG SQI
ppgSQI = round(np.mean(SQI.get_ppgSQI(ppg=s.ppg, fs=s.fs, annotation=fp.sp)) * 100, 2)
print('Mean PPG SQI: ', ppgSQI, '%')

Calculate PPG biomarkers:

# Init the biomarkers package
bmex = BM.BmCollection(s=s, fp=fp)

# Extract biomarkers
bm_defs, bm_vals, bm_stats = bmex.get_biomarkers()
tmp_keys=bm_stats.keys()
print('Statistics of the biomarkers:')
for i in tmp_keys: print(i,'\n',bm_stats[i])

# Create a biomarkers class
bm = Biomarkers(bm_defs=bm_defs, bm_vals=bm_vals, bm_stats=bm_stats)

Save PPG data:

# Save PPG struct, fiducial points, biomarkers
fp_new = Fiducials(fp.get_fp() + s.start_sig) # here the starting sample is added so that the results are relative to the start of the original signal (rather than the start of the analysed segment)
save_data(s=s, fp=fp_new, bm=bm, savingformat=savingformat, savingfolder=savingfolder)

Extracted fiducial points
Extracted biomarkers
Biomarkers statistics
Biomarkers definitions