ECG Pipeline: QRS Detection & HRV Domains

Prince E. Adjei

Kwame Nkrumah University of Science and Technology

Topic: ECG Pipeline Module 3: Biomedical Signal Processing

Biosignal Processes And Analysis (BME 366)

Topics

(1). ECG Feature Detection and Signal Basics

(2). QRS Detection and RR Interval Extraction

(3). Introduction to Heart Rate Variability (HRV)

(4). Time and Frequency Domain HRV Analysis

(5). HRV Visualization and Clinical Interpretation

ECG Pipeline: QRS Detection & HRV Domains

Learning Objectives

•Apply QRS detection techniques and extract clean RR intervals

for further analysis.

•Explain the physiological basis and relevance of Heart Rate

Variability (HRV).

•Perform HRV analysis using both time-domain and frequency-

domain methods.

•Interpret HRV metrics using visualizations and relate findings to

clinical conditions.

Review

•An electrocardiogram (ECG)

reflects the electrical activity

of the heart.

•The pacemaker cells, known as

the sinoatrial (SA) node,

control the heart rhythm.

•The PQRST represents one

complete ECG cycle.

•The ECG is the most commonly

known, recognized, and used

biomedical signal.

REVIEW

Wave Components:

•P wave –Atrial depolarization

•QRS complex –Ventricular

depolarization

•T wave –Ventricular repolarization

Key Intervals:

•PR: Atrial to ventricular delay

(~0.12–0.20s)

•QT: Ventricular activity (rate-

dependent)

•ST: Should be isoelectric (flat)

REVIEW

Sampling:

•Clinical ECG: 250–500 Hz

•Follows Nyquist (≥2× highest freq)

•Higher rates improve R-peak

accuracy

12-Lead ECG:

•Limb (frontal): I (LA–RA), II (LL–RA),

III (LL–LA), aVR (RA), aVL (LA), aVF

(LL)

•Chest (horizontal): V1–V6 from 4th

ICS (sternum) to midaxillary (V6)

Pan-Tompkins algorithm

•A classic and efficient real-time method for R-peak detection in

ECG signals.

•It uses a series of signal processing steps to accurately isolate QRS

complexes, even in noisy environments.

Signal Processing Pipeline

1.Band-pass Filtering

⚬Removes baseline wander and high-frequency noise

⚬Combines low-pass (~11 Hz) and high-pass (~5 Hz) filters

2.Differentiation

⚬Highlights rapid changes in slope

⚬Emphasizes steep QRS complex upstrokes

Pan-Tompkins algorithm

3. Squaring Function

⚬Converts the signal to positive values

⚬Amplifies large differences and suppresses small ones

4. Moving Window Integration

⚬Smooths the signal using a sliding window (~150 ms)

⚬Represents energy and the approximate width of the QRS

5. Adaptive Thresholding

⚬Continuously adjusts the threshold based on noise and signal

peaks

⚬Detects true R-peaks while rejecting false positives

Pan-Tompkins demo

•The RR interval is the time between two successive R-peaks on an

ECG signal.

•It reflects the duration of one cardiac cycle (i.e., one heartbeat).

•If tn and tn+1 are the time positions of two consecutive R-peaks (in

seconds), then:

•Measured in milliseconds (ms) or seconds (s)

•Used to calculate heart rate:

RR interval extraction

Ectopic Skip Handling

•Ectopic beats (e.g., PVCs) cause irregular RR intervals

•These lead to spurious short or long RR intervals

•Must be detected and flagged to avoid distortion in HRV or rhythm

analysis.

Error Cleaning

•Remove or correct intervals caused by:false/extra peaks, missed

detections, and motion artifacts

•Use signal quality index (SQI), moving average filters, or

interpolation

RR interval extraction

Heart Rate Variability (HRV) is the variation in time between successive

heartbeats, usually derived from the RR interval series.

Time-Domain Measures

•Based directly on the RR intervals

•Simple and widely used in clinical settings

•Reflects overall variability, short-term parasympathetic activity.

Frequency-Domain Measures

•Uses spectral analysis (e.g., FFT, Welch)

•Analyzes the distribution of power across frequency bands

•Reflects rhythmic oscillations in autonomic control

HRV domains:Time vs Frequency

•Time-domain HRV metrics quantify variability in the RR interval

time series directly, without requiring signal transformation.

Key Metrics:

1. SDNN (Standard Deviation of NN intervals)

⚬Reflects total HRV over the recording period

⚬Sensitive to both short- and long-term changes

⚬Common in 24-hour recordings

Time-Domain HRV

2. RMSSD (Root Mean Square of Successive Differences)

•Captures short-term variability

•Strongly associated with parasympathetic (vagal) activity

•Suitable for short recordings (e.g., 1–5 min)

3. pNN50 (% of successive RR intervals differing by >50 ms)

•Also reflects short-term parasympathetic modulation

•Correlates with RMSSD, but is more sensitive to outliers

Time-Domain HRV

Time-Domain: Recommended Time Windows

Metric

Minimum Duration

Typical Use

SDNN

≥5 min (best: 24h)

Overall variability

RMSSD

≥1 min

Short

-term HRV

pNN50

≥2

–5 min

Short

-term HRV

Frequency Domain HRV

•Analyzes how power (variability) is distributed across frequencies in the RR

interval series. Reflects autonomic nervous system rhythms.

Key Metrics:

1.LF Power (Low Frequency)

⚬Related to both sympathetic and parasympathetic influences

⚬Usually derived from 0.04–0.15 Hz band

2.HF Power (High Frequency)

⚬Reflects parasympathetic (vagal) activity

⚬Often associated with respiratory sinus arrhythmia (RSA)

3.LF/HF Ratio

⚬Proposed indicator of sympathovagal balance

⚬Interpretation is debated; it may oversimplify autonomic control

Questions

1) What is the QRS complex, and why is

it important in ECG analysis?

2) What is the purpose of the squaring

step in the Pan-Tompkins algorithm?

3) What is an RR interval, and how is it

used in HRV analysis?

1.Tachogram

•RR intervals vs.time (or beat

number)

•Shows beat-to-beat fluctuation

and long-term trends

•Useful for spotting sudden

shifts (e.g. ectopics, artifacts)

or gradual changes (e.g., stress,

recovery)

HRV Visualization Techniques

2. Poincaré Plot

•Scatter plot of RRₙ₊₁vs.RRₙ

•Elliptical shape →healthy

variability

•Flattened or clustered shape →

reduced HRV

•Quantified with:

•SD1 (short-term)

•SD2 (long-term)

3. Frequency Band Overlay

•Plot of Power Spectral Density

(PSD) with:

•LF and HF regions are shaded or

labeled

Makes it easy to see:

•Dominant rhythms

•Shifts in autonomic balance

•Often based on Welch’s method

HRV Visualization Techniques

•There is no fixed “normal” HRV value, but general ranges are known for

healthy adults:

HRV Standards

METRIC

RANGE

SDNN

–50 ms

RMSSD

–40 ms

pNN50

–15%

LF/HF

–2

Stress

•HRV decreases

•Reduced parasympathetic tone (vagal withdrawal)

•LF/HF ratio may increase

•Marker of mental/emotional/physical load

Sleep

•HRV increases, especially during deep (NREM) sleep

•Dominated by parasympathetic activity

•RMSSD and HF power rise

•Useful for tracking recovery or sleep quality

HRV in Different States

Stroke

•HRV markedly decreases

•SDNN and RMSSD are reduced in acute stroke

•Loss of autonomic control —a predictor of poor outcome

•Bilateral or brainstem lesions show the strongest effect

Exercise

•HRV decreases during exercise (especially RMSSD, HF)

•Reflects sympathetic activation

•HRV rebounds post-exercise during recovery

•Used to monitor training load and overtraining

HRV in Different States

The lab demonstrates how to:

•Load a real ECG recording from PhysioNet

•Use neurokit2.ecg_process() to detect R-peaks and extract

features

•Visualize the ECG signal with detected landmarks

•Optionally, compute HRV metrics and RR intervals

Lab Preview: Analyzing PhysioNet ECG with neurokit2

•The Pan-Tompkins algorithm detects R-peaks through a sequence of

filtering, differentiation, squaring, and adaptive thresholding.

•RR intervals form the basis of HRV analysis, requiring careful

preprocessing to handle ectopic beats and noise.

•HRV is assessed in time and frequency domains using metrics like

SDNN, RMSSD, pNN50,and LF/HF ratio, often computed via Welch PSD.

•HRV can be visualized using tachograms, Poincaré plots, and frequency

band overlays to illustrate variability and autonomic balance.

Summary