Examples¶

Signal Preprocessing¶

PhysioView includes built-in functions and methods for preprocessing ECG, PPG, EDA, and accelerometer signals.

Electrocardiography (ECG)¶

When you import the physioview package, the ECG.Filters and ECG.BeatDetectors classes are automatically available as ECGFilters and ECGBeatDetectors, respectively. These provide convenient access to ECG filtering and beat detection methods.

import physioview as pv
import pandas as pd

# Load the ECG data sampled at 1024 Hz
ecg = pd.read_csv('sample_ecg_1024hz.csv')
fs = 1024

# Filter the ECG signal
ecg['Filtered'] = pv.ECGFilters(fs).filter_signal(ecg['mV'])

# Detect beat locations in the filtered ECG signal
beats_ix = pv.ECGBeatDetectors(fs).manikandan(ecg['Filtered'])

Photoplethysmography (PPG)¶

Similarly, the PPG.Filters and PPG.BeatDetectors classes are automatically available as PPGFilters and PPGBeatDetectors when physioview is imported.

import physioview as pv
import pandas as pd

# Load PPG data sampled at 64 Hz
ppg = pd.read_csv('sample_ppg_64hz.csv')
fs = 64

# Filter the PPG signal
ppg['Filtered'] = pv.PPGFilters(fs).filter_signal(ppg['BVP'])

# Detect beat locations in the filtered PPG signal
beats_ix = pv.PPGBeatDetectors(fs).adaptive_threshold(ppg['Filtered'])

Electrodermal Activity (EDA)¶

Access EDAFilters as EDAFilters when importing physioview.

import physioview as pv
import pandas as pd

# Load EDA data sampled at 4 Hz
eda = pd.read_csv('sample_eda_4hz.csv')
fs = 4

# Filter the EDA signal
eda['Filtered'] = pv.EDAFilters(fs).filter_signal(eda['EDA'])

Feature Extraction¶

PhysioView also provides feature extraction utilities that convert raw physiological signals into interpretable measures, such as cardiac interbeat intervals (IBIs), heart rate variability (HRV) metrics, and EDA components.

Interbeat Intervals & Heart Rate Variability¶

Extract cardiac features such as IBIs and HRV metrics from an array of beat locations using compute_ibis and compute_hrv.

Note: PhysioView runs flirt under the hood. See FLIRT’s documentation for more information about HRV metrics.

import physioview as pv

# Compute IBI
ibi = pv.compute_ibis(ecg, fs, beats_ix, ts_col = 'Timestamp')

# Compute HRV metrics across 60-sec sliding windows at 15-sec intervals
hrv = pv.compute_hrv(ecg, fs, beats_ix, window_size = 60,
                     step_size = 15, ts_col = 'Timestamp')

Outputs:

The resulting ibi DataFrame has the same number of rows as the input data and provides IBI values aligned with detected beat locations in the signal.

In [1]: ibi
Out[2]:
                         Timestamp  IBI
0       2016-10-14 10:10:51.000000  NaN
1       2016-10-14 10:10:51.000977  NaN
2       2016-10-14 10:10:51.001953  NaN
3       2016-10-14 10:10:51.002930  463.867188
4       2016-10-14 10:10:51.003906  NaN
...                            ...  ...
381948  2016-10-14 10:17:03.996094  NaN
381949  2016-10-14 10:17:03.997070  458.984375
381950  2016-10-14 10:17:03.998047  NaN
381951  2016-10-14 10:17:03.999023  NaN
381952  2016-10-14 10:17:04.000000  NaN
[381953 rows x 2 columns]

The resulting hrv DataFrame contains HRV features computed over sliding windows, with each row corresponding to a window and each column to a specific HRV metric.

In [3]: hrv.head()
Out[4]:
                         num_ibis  hrv_mean_nni  ...  hrv_perm_entropy  hrv_svd_entropy
Timestamp                                    ...
2016-10-14 10:11:52       129    467.629603  ...          0.995452         0.413629
2016-10-14 10:12:07       128    468.673706  ...          0.998846         0.385510
2016-10-14 10:12:22       128    469.207764  ...          0.996257         0.304091
2016-10-14 10:12:37       128    474.205017  ...          0.996257         0.243242
2016-10-14 10:12:52       128    469.207764  ...          0.999584         0.120002
[5 rows x 52 columns]

Tonic Skin Conductance Level¶

Compute the tonic skin conductance level (SCL) from an EDA signal across segments or for the entire signal.

import physioview as pv

# Compute tonic SCL across 3-minute windows
tonic_scl_segment = pv.EDA.compute_tonic_scl(eda['Filtered'], fs, seg_size = 180)

# Compute tonic SCL for the entire EDA signal
tonic_scl_entire = pv.EDA.compute_tonic_scl(eda['Filtered'], fs)

EDA Decomposition¶

Extract the phasic and tonic components of an EDA signal with the convex optimization approach [1].

import physioview as pv

eda_fs = 4  # sampling rate
phasic, tonic = pv.EDA.decompose_signal(eda['EDA'], eda_fs)

Outputs:

In [1]: phasic
Out[2]:
array([0.        , 0.        , 0.1267762 , ..., 0.15150715, 0.14166146,
       0.13092194])

In [3]: tonic
Out[4]:
array([-3.4407805 , -3.50473019, -3.5572274 , ...,  5.36595805,
        5.36271308,  5.35938202])

Statistical EDA Features¶

Compute statistical EDA features based on the extracted phasic and tonic components.

Note: PhysioView runs flirt under the hood. See FLIRT’s documentation for more information about the outputted EDA features.