Signal Processing & Biomedical Analysis

Part of Statistical Methods — MORIE’s statistical-methods reference.

MORIE provides 25 biomedical signal processing functions via morie.signal and individual morie.fn.* modules. All functions are dataset-agnostic: numpy arrays in, result objects out.

Digital Filters

Butterworth zero-phase filters (via scipy.signal):

from morie.signal import buttlp, butthp, buttbp, buttbs, sgolay

result = buttlp(ecg_signal, fs=500, cutoff=40, order=4)
filtered = result.filtered

  • buttlp – Lowpass

  • butthp – Highpass

  • buttbp – Bandpass

  • buttbs – Bandstop (notch, default 59-61 Hz for mains hum)

  • sgolay – Savitzky-Golay polynomial smoothing

Spectral Analysis

from morie.signal import welch, pburg

psd = welch(signal, fs=256)        # Welch PSD
ar_psd = pburg(signal, order=16)   # Burg AR PSD (parametric)

Fractal Complexity

Pure-numpy implementations for nonlinear time-series characterization:

from morie.signal import hfd, kfd, pfd, dfa, sampen, hurst

result = hfd(signal, kmax=10)    # Higuchi fractal dimension
alpha = dfa(signal).value        # DFA scaling exponent

  • hfd – Higuchi fractal dimension (Higuchi, 1988)

  • kfd – Katz fractal dimension

  • pfd – Petrosian fractal dimension

  • dfa – Detrended fluctuation analysis

  • sampen – Sample entropy

  • hurst – Hurst exponent (R/S analysis)

Typical values: Brownian motion HFD ~1.5, white noise DFA alpha ~0.5.

ECG and Heart Rate Variability

from morie.signal import ecgdet, rrint, hrvtd, hrvfd, hrvnl

peaks = ecgdet(ecg, fs=360)              # Pan-Tompkins QRS detection
rr = rrint(peaks.extra["qrs_indices"], fs=360)
td = hrvtd(rr)                            # SDNN, RMSSD, pNN50
fd = hrvfd(rr)                            # VLF/LF/HF power
nl = hrvnl(rr)                            # Poincare SD1/SD2

Pan-Tompkins detector (Pan & Tompkins, 1985) with adaptive thresholding. HRV metrics follow Task Force (1996) standards.

Phonocardiogram (PCG) Analysis

For cardiotoxicity studies in addiction/substance use research:

from morie.signal import pcgflt, pcgenv, pcgseg, pcgmur

filtered = pcgflt(pcg, fs=2000)           # 25-400 Hz bandpass
envelope = pcgenv(pcg, fs=2000)           # Shannon energy envelope
segments = pcgseg(envelope.filtered, fs=2000)  # S1/S2 segmentation
score = pcgmur(pcg, fs=2000)              # Murmur detection score

The murmur detection score combines Higuchi fractal dimension, high-frequency energy ratio, and spectral entropy. Scores are uncalibrated (0-1 range); calibration requires labeled clinical data.

References

  • Higuchi, T. (1988). Approach to an irregular time series on the basis of the fractal theory. Physica D, 31(2), 277-283.

  • Pan, J. & Tompkins, W.J. (1985). A real-time QRS detection algorithm. IEEE Trans. Biomed. Eng., 32(3), 230-236.

  • Task Force of ESC/NASPE (1996). Heart rate variability: Standards of measurement. Circulation, 93(5), 1043-1065.

  • Rangayyan, R.M. (2015). Biomedical Signal Analysis. IEEE Press.