Example analysis of a PPG/ECG waveform using the vital_sqi package

The following notebook shows an example of PPG/ECG waveform processing using the vital_sqi package. The aim of the package is to automate signal quality classification for PPG/ECG waveforms. It is achieved by computing various signal quality indices for each signal segment and using them to form a decision.

The pipeline can be briefly summarized as follows:

1. Load dataset under analysis

2. Preprocess and segment the dataset

3. Compute SQI for each dataset segment

4. Make decision for each segment

Global Imports

import numpy as np
import warnings
import os
import pandas as pd
import matplotlib.pyplot as plt
import vital_sqi
from vital_sqi.data.signal_io import ECG_reader,PPG_reader

Start by importing the signal via the PPG_reader function

The function expects a .csv or similar data format with named columns. The column names are used to separate between data column, timestamp columns and any additional information columns. This returns a SignalSQI class that is compatible with other vital_sqi package functions, the main class members of interest are: * signals: an ndarray of shape (m, n) where m is the number of rows and n is the number of channels of the signal * sqis: an ndarray of shape (m, n) where m is the number of signal segments, n is the number of SQIs. * sampling_rate: sampling rate in hertz (Hz)

from vital_sqi.pipeline.pipeline_highlevel import get_ppg_sqis

file_in = os.path.abspath('test_data/ppg_smartcare.csv') #FIle input location
sqi_dict = os.path.abspath('test_data/sqi_dict.json') #input dictionary -> which sqi features
segments, signal_obj = get_ppg_sqis(file_in,
                                    signal_idx=['PLETH'],
                                    timestamp_idx=['TIMESTAMP_MS'],
                                    sqi_dict_filename = sqi_dict)

100%|██████████| 21/21 [00:02<00:00,  7.20it/s]

signal_obj.sqis[0].describe()

	perfusion	kurtosis_1	skewness_mean_sqi	skewness_median_sqi	skewness_std_sqi	entropy	kurtosis_2	skewness_1	entropy_2	signal_to_noise	...	hfnu_sqi	total_power_sqi	vlf_sqi	triangular_index_sqi	tinn_sqi	csi_sqi	cvi_sqi	Modified_csi_sqi	start_idx	end_idx
count	21.000000	21.000000	21.000000	21.000000	21.000000	21.000000	21.000000	21.000000	21.000000	21.0	...	21.000000	21.000000	21.000000	21.000000	0.0	21.000000	21.000000	21.000000	21.000000	21.000000
mean	187.404653	-1.360325	0.073129	0.080781	0.120721	4.343845	-0.966196	0.086825	7.777617	-1.0	...	53.024653	2457.690263	587.473086	5.004001	NaN	1.596050	4.751727	765.778807	30000.000000	33000.000000
std	20.712381	0.032043	0.058008	0.035630	0.115958	0.023989	0.395934	0.113371	0.159168	0.0	...	23.529422	3642.831489	998.300743	1.670246	NaN	0.695445	0.878519	742.250417	18614.510469	18614.510469
min	112.313216	-1.396156	-0.086416	0.009774	0.019471	4.313126	-1.349569	-0.318782	7.103965	-1.0	...	17.870115	109.769282	8.139479	2.857143	NaN	0.822317	3.590477	59.345391	0.000000	3000.000000
25%	183.541473	-1.383222	0.058113	0.061718	0.032456	4.327253	-1.277692	0.063396	7.783738	-1.0	...	35.720578	222.530802	57.790560	3.555556	NaN	1.161429	3.868977	119.530778	15000.000000	18000.000000
50%	186.970824	-1.367053	0.070987	0.077784	0.077805	4.334973	-1.103788	0.081037	7.809623	-1.0	...	54.651397	811.417421	252.630297	4.250000	NaN	1.418324	5.074484	562.534403	30000.000000	33000.000000
75%	195.615534	-1.354891	0.114273	0.108330	0.202787	4.367132	-0.822027	0.132159	7.826186	-1.0	...	70.309215	2499.129760	493.237543	6.500000	NaN	1.885878	5.541223	1183.176936	45000.000000	48000.000000
max	215.095923	-1.287866	0.147944	0.146486	0.455517	4.387753	0.067313	0.269913	7.893871	-1.0	...	89.929842	11283.942003	4052.382611	8.600000	NaN	3.300646	6.005650	2769.520662	60000.000000	63000.000000

8 rows × 79 columns

Get decision

from vital_sqi.pipeline.pipeline_highlevel import get_qualified_ppg
rule_dict_filename = os.path.abspath('test_data/rule_dict_test.json') #  location dictionary contain threshold
ruleset_order = {3: 'skewness_1',
                  2: 'entropy',
                  1: 'perfusion'}
file_in = os.path.abspath('test_data/ppg_smartcare.csv') #FIle input location
sqi_dict = os.path.abspath('test_data/sqi_dict.json') #input dictionary -> which sqi features
signal_obj = get_qualified_ppg(file_in,
                                    signal_idx=['PLETH'],
                                    timestamp_idx=['TIMESTAMP_MS'],
                                    sqi_dict_filename = sqi_dict,
                                    rule_dict_filename = rule_dict_filename,
                                    ruleset_order=ruleset_order,
                                    output_dir=None)

100%|██████████| 21/21 [00:02<00:00,  9.76it/s]
0it [00:00, ?it/s]
100%|██████████| 21/21 [00:00<00:00, 57.97it/s]