def correct_peaks(peaks, extra_correction=True, missed_correction=True,
short_correction=True, long_correction=True,
ectopic_correction=True):
"""Correct long, short, extra, missed and ectopic beats in peaks vector.
Parameters
----------
peaks : 1d array-like
Boolean vector of peaks.
Returns
-------
correction : dictionnary
The corrected RR time series and the number of artefacts corrected:
* clean_peaks: 1d array-like
The corrected boolean time-serie.
* ectopic: int
The number of ectopic beats corrected.
* short: int
The number of short beats corrected.
* long: int
The number of long beats corrcted.
* extra: int
The number of extra beats corrected.
* missed: int
The number of missed beats corrected.
"""
if isinstance(peaks, list):
peaks = np.asarray(peaks, dtype=bool)
clean_peaks = peaks.copy()
nEctopic, nShort, nLong, nExtra, nMissed = 0, 0, 0, 0, 0
artefacts = rr_artefacts(np.diff(np.where(clean_peaks)[0]))
# Correct missed beats
if missed_correction:
if np.any(artefacts['missed']):
for this_id in np.where(artefacts['missed'])[0]:
this_id += nMissed
clean_peaks = correct_missed_peaks(clean_peaks, this_id)
nMissed += 1
artefacts = rr_artefacts(np.diff(np.where(clean_peaks)[0]))
# Correct extra beats
if extra_correction:
if np.any(artefacts['extra']):
for this_id in np.where(artefacts['extra'])[0]:
this_id -= nExtra
clean_peaks = correct_extra_peaks(clean_peaks, this_id)
nExtra += 1
artefacts = rr_artefacts(np.diff(np.where(clean_peaks)[0]))
return {'clean_peaks': clean_peaks, 'ectopic': nEctopic, 'short': nShort,
'long': nLong, 'extra': nExtra, 'missed': nMissed}
plot1 = plot_timedomain(rr) plotly.io.show(plot1) #%% # Frequency domain # ---------------- plot2 = plot_frequency(rr) plotly.io.show(plot2) #%% # Nonlinear domain # ---------------- plot3 = plot_nonlinear(rr) plotly.io.show(plot3) #%% # Artefact detection # ------------------ artefacts = rr_artefacts(rr) #%% # Subspaces visualization # ----------------------- # You can visualize the two main subspaces and spot outliers. The left pamel # plot subspaces that are more sensitive to ectopic beats detection. The right # panel plot subspaces that will be more sensitive to long or short beats, # comprizing the extra and missed beats. plot_subspaces(rr)
def test_rr_artefacts(self):
rr = simulate_rr() # Import PPG recording
artefacts = rr_artefacts(rr)
artefacts = rr_artefacts(list(rr))
assert all(350 == x
for x in [len(artefacts[k]) for k in artefacts.keys()])
def plot_subspaces(x, c1=.17, c2=.13, xlim=10, ylim=5, ax=None):
"""Plot hrv subspace as described by Lipponen & Tarvainen (2019).
Parameters
----------
x : 1d array-like
Array of RR intervals or subspace1. If subspace1 is provided, subspace2
and 3 must also be provided.
c1 : float
Fixed variable controling the slope of the threshold lines. Default is
0.13.
c2 : float
Fixed variable controling the intersect of the threshold lines. Default
is 0.17.
xlim : int
Absolute range of the x axis. Default is 10.
ylim : int
Absolute range of the y axis. Default is 5.
ax : `Matplotlib.Axes` or None
Where to draw the plot. Default is *None* (create a new figure).
Returns
-------
ax : `Matplotlib.Axes`
The figure.
References
----------
[1] Lipponen, J. A., & Tarvainen, M. P. (2019). A robust algorithm for
heart rate variability time series artefact correction using novel beat
classification. Journal of Medical Engineering & Technology, 43(3),
173–181. https://doi.org/10.1080/03091902.2019.1640306
"""
if not isinstance(x, (np.ndarray, np.generic)):
x = np.asarray(x)
artefacts = rr_artefacts(x)
# Rescale to show outlier in scatterplot
if xlim is not None:
artefacts['subspace1'][artefacts['subspace1'] < -xlim] = -xlim
artefacts['subspace1'][artefacts['subspace1'] > xlim] = xlim
if ylim is not None:
artefacts['subspace2'][artefacts['subspace2'] < -ylim] = -ylim
artefacts['subspace2'][artefacts['subspace2'] > ylim] = ylim
artefacts['subspace3'][artefacts['subspace3'] < -ylim*2] = -ylim*2
artefacts['subspace3'][artefacts['subspace3'] > ylim*2] = ylim*2
# Filter for normal beats
normalBeats = ((~artefacts['ectopic']) & (~artefacts['short']) &
(~artefacts['long']) & (~artefacts['missed']) &
(~artefacts['extra']))
#############
# First panel
#############
if ax is None:
fig, ax = plt.subplots(1, 2, figsize=(10, 5))
# Plot normal beats
ax[0].scatter(artefacts['subspace1'][normalBeats],
artefacts['subspace2'][normalBeats],
color='gray', edgecolors='k', s=15,
alpha=0.2, zorder=10, label='Normal')
# Plot outliers
ax[0].scatter(artefacts['subspace1'][artefacts['ectopic']],
artefacts['subspace2'][artefacts['ectopic']],
color='r', edgecolors='k', zorder=10, label='Ectopic')
ax[0].scatter(artefacts['subspace1'][artefacts['short']],
artefacts['subspace2'][artefacts['short']],
color='b', edgecolors='k', zorder=10,
marker='s', label='Short')
ax[0].scatter(artefacts['subspace1'][artefacts['long']],
artefacts['subspace2'][artefacts['long']],
color='g', edgecolors='k', zorder=10,
marker='s', label='Long')
ax[0].scatter(artefacts['subspace1'][artefacts['missed']],
artefacts['subspace2'][artefacts['missed']],
color='g', edgecolors='k', zorder=10, label='Missed')
ax[0].scatter(artefacts['subspace1'][artefacts['extra']],
artefacts['subspace2'][artefacts['extra']],
color='b', edgecolors='k', zorder=10, label='Extra')
# Upper area
def f1(x): return -c1*x + c2
ax[0].plot([-1, -10], [f1(-1), f1(-10)], 'k', linewidth=1, linestyle='--')
ax[0].plot([-1, -1], [f1(-1), 10], 'k', linewidth=1, linestyle='--')
x = [-10, -10, -1, -1]
y = [f1(-10), 10, 10, f1(-1)]
ax[0].fill(x, y, color='gray', alpha=0.3)
# Lower area
def f2(x): return -c1*x - c2
ax[0].plot([1, 10], [f2(1), f2(10)], 'k', linewidth=1, linestyle='--')
ax[0].plot([1, 1], [f2(1), -10], 'k', linewidth=1, linestyle='--')
x = [1, 1, 10, 10]
y = [f2(1), -10, -10, f2(10)]
ax[0].fill(x, y, color='gray', alpha=0.3)
ax[0].set_xlabel('Subspace $S_{11}$')
ax[0].set_ylabel('Subspace $S_{12}$')
ax[0].set_ylim(-ylim, ylim)
ax[0].set_xlim(-xlim, xlim)
ax[0].set_title('Subspace 1 \n (ectopic beats detection)')
ax[0].legend()
##############
# Second panel
##############
# Plot normal beats
ax[1].scatter(artefacts['subspace1'][normalBeats],
artefacts['subspace3'][normalBeats],
color='gray', edgecolors='k', alpha=0.2,
zorder=10, s=15, label='Normal')
# Plot outliers
ax[1].scatter(artefacts['subspace1'][artefacts['ectopic']],
artefacts['subspace3'][artefacts['ectopic']],
color='r', edgecolors='k', zorder=10, label='Ectopic')
ax[1].scatter(artefacts['subspace1'][artefacts['short']],
artefacts['subspace3'][artefacts['short']],
color='b', edgecolors='k', zorder=10,
marker='s', label='Short')
ax[1].scatter(artefacts['subspace1'][artefacts['long']],
artefacts['subspace3'][artefacts['long']],
color='g', edgecolors='k', zorder=10,
marker='s', label='Long')
ax[1].scatter(artefacts['subspace1'][artefacts['missed']],
artefacts['subspace3'][artefacts['missed']],
color='g', edgecolors='k', zorder=10, label='Missed')
ax[1].scatter(artefacts['subspace1'][artefacts['extra']],
artefacts['subspace3'][artefacts['extra']],
color='b', edgecolors='k', zorder=10, label='Extra')
# Upper area
ax[1].plot([-1, -10], [1, 1], 'k', linewidth=1, linestyle='--')
ax[1].plot([-1, -1], [1, 10], 'k', linewidth=1, linestyle='--')
x = [-10, -10, -1, -1]
y = [1, 10, 10, 1]
ax[1].fill(x, y, color='gray', alpha=0.3)
# Lower area
ax[1].plot([1, 10], [-1, -1], 'k', linewidth=1, linestyle='--')
ax[1].plot([1, 1], [-1, -10], 'k', linewidth=1, linestyle='--')
x = [1, 1, 10, 10]
y = [-1, -10, -10, -1]
ax[1].fill(x, y, color='gray', alpha=0.3)
ax[1].set_xlabel('Subspace $S_{21}$')
ax[1].set_ylabel('Subspace $S_{22}$')
ax[1].set_ylim(-ylim*2, ylim*2)
ax[1].set_xlim(-xlim, xlim)
ax[1].set_title('Subspace 2 \n (long and short beats detection)')
ax[1].legend()
plt.tight_layout()
return ax
# * The category in which the artefact belongs will have an influence on the # correction procedure (see Artefact correction). #%% # Simulate RR time series # ----------------------- # This function will simulate RR time series containing ectopic, extra, missed, # long and short artefacts. rr = simulate_rr() #%% # Artefact detection # ------------------ outliers = rr_artefacts(rr) #%% # Subspaces visualization # ----------------------- # You can visualize the two main subspaces and spot outliers. The left pamel # plot subspaces that are more sensitive to ectopic beats detection. The right # panel plot subspaces that will be more sensitive to long or short beats, # comprizing the extra and missed beats. plot_subspaces(rr) #%% # References # ---------- # .. [#] Lipponen, J. A., & Tarvainen, M. P. (2019). A robust algorithm for
def correct_rr(rr, extra_correction=True, missed_correction=True,
short_correction=True, long_correction=True,
ectopic_correction=True):
"""Correct long and short beats using interpolation.
Parameters
----------
rr : 1d array-like
RR intervals (ms).
correct_extra : boolean
If True, correct extra beats in the RR time series.
correct_missed : boolean
If True, correct missed beats in the RR time series.
correct_short : boolean
If True, correct short beats in the RR time series.
correct_long : boolean
If True, correct long beats in the RR time series.
correct_ectopic : boolean
If True, correct ectopic beats in the RR time series.
Returns
-------
correction : dictionnary
The corrected RR time series and the number of artefacts corrected:
* clean_rr: 1d array-like
The corrected RR time-serie.
* ectopic: int
The number of ectopic beats corrected.
* short: int
The number of short beats corrected.
* long: int
The number of long beats corrcted.
* extra: int
The number of extra beats corrected.
* missed: int
The number of missed beats corrected.
"""
if isinstance(rr, list):
rr = np.asarray(rr)
clean_rr = rr.copy()
nEctopic, nShort, nLong, nExtra, nMissed = 0, 0, 0, 0, 0
artefacts = rr_artefacts(clean_rr)
# Correct missed beats
if missed_correction:
if np.any(artefacts['missed']):
for this_id in np.where(artefacts['missed'])[0]:
this_id += nMissed
clean_rr = correct_missed(clean_rr, this_id)
nMissed += 1
artefacts = rr_artefacts(clean_rr)
# Correct extra beats
if extra_correction:
if np.any(artefacts['extra']):
for this_id in np.where(artefacts['extra'])[0]:
this_id -= nExtra
clean_rr = correct_missed(clean_rr, this_id)
nExtra += 1
artefacts = rr_artefacts(clean_rr)
# Correct ectopic beats
if ectopic_correction:
if np.any(artefacts['ectopic']):
# Also correct the beat before
for i in np.where(artefacts['ectopic'])[0]:
if (i > 0) & (i < len(artefacts['ectopic'])):
artefacts['ectopic'][i-1] = True
this_id = np.where(artefacts['ectopic'])[0]
clean_rr = interpolate_bads(clean_rr, [this_id])
nEctopic = np.sum(artefacts['ectopic'])
# Correct short beats
if short_correction:
if np.any(artefacts['short']):
this_id = np.where(artefacts['short'])[0]
clean_rr = interpolate_bads(clean_rr, this_id)
nShort = len(this_id)
# Correct long beats
if long_correction:
if np.any(artefacts['long']):
this_id = np.where(artefacts['long'])[0]
clean_rr = interpolate_bads(clean_rr, this_id)
nLong = len(this_id)
return {'clean_rr': clean_rr, 'ectopic': nEctopic, 'short': nShort,
'long': nLong, 'extra': nExtra, 'missed': nMissed}
def test_rr_artefacts(self):
rr = simulate_rr() # Simulate RR time series
artefacts = rr_artefacts(rr)
artefacts = rr_artefacts(list(rr))
assert all(350 == x
for x in [len(artefacts[k]) for k in artefacts.keys()])