Hi, I am a university student at the University of Manchester that working on my disse

Hi <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url="

Questions about the default method of ecg_peaks() in neurokit2 about biopeaks HOT 1 CLOSED

skywalkerzhang commented on September 27, 2024

Questions about the default method of ecg_peaks() in neurokit2

from biopeaks.

Comments (1)

JanCBrammer commented on September 27, 2024

Hi @skywalkerzhang, thanks for your question. Have a look at https://joss.theoj.org/papers/10.21105/joss.02621. In the paper there's a validation of the R-peak detector that was originally implemented for biopeaks and is now also used as default method for R-peak detection in NeuroKit2. Note that the paper doesn't include details on the internals of the algorithm, only a validation, but maybe it can still be useful as a reference.

You can have a look at the implementation here:

biopeaks/biopeaks/heart.py

Lines 14 to 104 in 9d11af4

 def ecg_peaks(signal, sfreq, smoothwindow=.1, avgwindow=.75, 

 gradthreshweight=1.5, minlenweight=.4, mindelay=.3, 

 enable_plot=False): 

 """Detect R-peaks in an electrocardiogram (ECG). 

  QRS complexes are detected based on the steepness of the absolute gradient 

  of the ECG signal. Subsequently, R-peaks are detected as local maxima in 

  the QRS complexes. 

  Parameters 

  ---------- 

  signal : ndarray 

  The ECG signal. 

  sfreq : int 

  The sampling frequency of `signal`. 

  smoothwindow : float, optional 

  Size of the kernel used for smoothing the absolute gradient of `signal`. 

  In seconds. Default is .1. 

  avgwindow : float, optional 

  Size of the kernel used for computing the local average of the smoothed 

  absolute gradient of `signal`. In seconds. Default is .75. 

  gradthreshweight : float, optional 

  Factor used to offset the averaged absolute gradient of `signal`. The 

  resulting time series is then used as a threshold for QRS detection. 

  Default is 1.5. 

  minlenweight : float, optional 

  The average QRS duration is multiplied by `minlenweight` in order to 

  obtain a threshold for the minimal duration of QRS complexes (QRS 

  complexes shorted than the minimal duration will be discared). Default 

  is .4. 

  mindelay : float, optional 

  Minimal delay between R-peaks. R-peaks that follow other R-peaks by less 

  than `mindelay` will be discarded. In seconds. Default is .3. 

  enable_plot : bool, optional 

  Visualize `signal` along with the detection thresholds, as well as the 

  detected QRS complexes and R-peaks. Default is False. 

  Returns 

  ------- 

  peaks : ndarray 

  The samples within `signal` that mark the occurrences of R-peaks. 

  """ 

 if enable_plot: 

 plt.figure() 

 ax1 = plt.subplot(211) 

 ax2 = plt.subplot(212, sharex=ax1) 

 filt = butter_highpass_filter(signal, .5, sfreq) 

 filt = powerline_filter(filt, sfreq) 

 grad = np.gradient(filt) 

 absgrad = np.abs(grad) 

 smoothgrad = moving_average(absgrad, int(np.rint(smoothwindow * sfreq))) 

 avggrad = moving_average(smoothgrad, int(np.rint(avgwindow * sfreq))) 

 gradthreshold = gradthreshweight * avggrad 

 mindelay = int(np.rint(sfreq * mindelay)) 

 if enable_plot: 

 ax1.plot(filt) 

 ax2.plot(smoothgrad) 

 ax2.plot(gradthreshold) 

 qrs = smoothgrad > gradthreshold 

 beg_qrs, end_qrs, durations_qrs = find_segments(qrs) 

 # Identify R-peaks within QRS (ignore QRS that are too short). 

 min_len = np.mean(durations_qrs) * minlenweight 

 peaks = [0] 

 for beg, end, duration in zip(beg_qrs, end_qrs, durations_qrs): 

 if duration < min_len: 

 continue 

 if enable_plot: 

 ax2.axvspan(beg, end, facecolor="m", alpha=0.5) # visualize QRS 

 data = signal[beg:end] 

 locmax, props = find_peaks(data, prominence=(None, None)) # find local maxima and their prominence within QRS 

 if locmax.size > 0: 

 peak = beg + locmax[np.argmax(props["prominences"])] # identify most prominent local maximum 

 if peak - peaks[-1] > mindelay: # enforce minimum delay between R-peaks 

 peaks.append(peak) 

 peaks.pop(0) 

 if enable_plot: 

 ax1.scatter(peaks, filt[peaks], c="r") 

 return np.asarray(peaks).astype(int)

Maybe visualizing the R-peak detection by calling the function with enable_plot=True can give you some more insight in how the detection works.

from biopeaks.

Questions about the default method of ecg_peaks() in neurokit2 about biopeaks HOT 1 CLOSED

Comments (1)

Related Issues (20)

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent

	def ecg_peaks(signal, sfreq, smoothwindow=.1, avgwindow=.75,
	gradthreshweight=1.5, minlenweight=.4, mindelay=.3,
	enable_plot=False):
	"""Detect R-peaks in an electrocardiogram (ECG).

	QRS complexes are detected based on the steepness of the absolute gradient
	of the ECG signal. Subsequently, R-peaks are detected as local maxima in
	the QRS complexes.

	Parameters
	----------
	signal : ndarray
	The ECG signal.
	sfreq : int
	The sampling frequency of `signal`.
	smoothwindow : float, optional
	Size of the kernel used for smoothing the absolute gradient of `signal`.
	In seconds. Default is .1.
	avgwindow : float, optional
	Size of the kernel used for computing the local average of the smoothed
	absolute gradient of `signal`. In seconds. Default is .75.
	gradthreshweight : float, optional
	Factor used to offset the averaged absolute gradient of `signal`. The
	resulting time series is then used as a threshold for QRS detection.
	Default is 1.5.
	minlenweight : float, optional
	The average QRS duration is multiplied by `minlenweight` in order to
	obtain a threshold for the minimal duration of QRS complexes (QRS
	complexes shorted than the minimal duration will be discared). Default
	is .4.
	mindelay : float, optional
	Minimal delay between R-peaks. R-peaks that follow other R-peaks by less
	than `mindelay` will be discarded. In seconds. Default is .3.
	enable_plot : bool, optional
	Visualize `signal` along with the detection thresholds, as well as the
	detected QRS complexes and R-peaks. Default is False.

	Returns
	-------
	peaks : ndarray
	The samples within `signal` that mark the occurrences of R-peaks.
	"""
	if enable_plot:
	plt.figure()
	ax1 = plt.subplot(211)
	ax2 = plt.subplot(212, sharex=ax1)

	filt = butter_highpass_filter(signal, .5, sfreq)
	filt = powerline_filter(filt, sfreq)

	grad = np.gradient(filt)
	absgrad = np.abs(grad)
	smoothgrad = moving_average(absgrad, int(np.rint(smoothwindow * sfreq)))
	avggrad = moving_average(smoothgrad, int(np.rint(avgwindow * sfreq)))
	gradthreshold = gradthreshweight * avggrad
	mindelay = int(np.rint(sfreq * mindelay))

	if enable_plot:
	ax1.plot(filt)
	ax2.plot(smoothgrad)
	ax2.plot(gradthreshold)

	qrs = smoothgrad > gradthreshold
	beg_qrs, end_qrs, durations_qrs = find_segments(qrs)

	# Identify R-peaks within QRS (ignore QRS that are too short).
	min_len = np.mean(durations_qrs) * minlenweight
	peaks = [0]

	for beg, end, duration in zip(beg_qrs, end_qrs, durations_qrs):

	if duration < min_len:
	continue

	if enable_plot:
	ax2.axvspan(beg, end, facecolor="m", alpha=0.5) # visualize QRS

	data = signal[beg:end]
	locmax, props = find_peaks(data, prominence=(None, None)) # find local maxima and their prominence within QRS

	if locmax.size > 0:
	peak = beg + locmax[np.argmax(props["prominences"])] # identify most prominent local maximum
	if peak - peaks[-1] > mindelay: # enforce minimum delay between R-peaks
	peaks.append(peak)

	peaks.pop(0)

	if enable_plot:
	ax1.scatter(peaks, filt[peaks], c="r")

	return np.asarray(peaks).astype(int)