MEDICAL ELECTRONICS & BIODEVICES
Detecting atrial fibrillation via ECG AFE and PPG
By ASteve Koh, Chris Hopwood and Andrew Burt trial fibrillation (AF)
affects three to six
million people in
the U.S., and its monitoring
is essential for correct
Fig. 1: Signal flow in Pan-Tompkins algorithm.
diagnosis and treatment.
Because AF is paroxysmal in nature, it is often difficult to detect
at a doctor’s office. Wearable devices present an option for AF
detection via an electrocardiogram analog front-end (ECG AFE)
or an optical photoplethysmography (PPG). Each respective
signal has its own advantages and disadvantages in accuracy,
sensor location, and R-R detection.
AF is an abnormal heart rhythm caused by deterioration of
the electrical impulses in the upper cardiac chambers known as
atria, resulting in irregular, rapid heart rate. Possible symptoms
of AF are heart palpitations, weakness, fainting, dizziness,
chest pain, and shortness of breath. AF prevents the cardiac
ventricle from completing full contraction. As a result, blood is
not pumped efficiently
to the rest of the body,
causing an unusually
fast heart rate, quivering,
or thumping sensations
in the heart. Blood
is highly viscous and
a higher pumping rate
prevents it from flowing
effectively.
AF also results in ineffectual
atrial contractions.
These decrease
blood volume within
ventricles, lowering ventricular
filling pressure
and mean atrial pressure
and resulting in
poor cardiac output and
vulnerability to blood
clot formation that can
result in stroke events
that impact quality of
life. Those who have AF
are five to seven times
more likely to form
blood clots and suffer
an ischemic stroke, and
AF is responsible for 15
to 20% of all strokes.
Fortunately, AF may
be treated with medication,
by sending electric
shocks, or by disconnecting
the re-entrant
cardiac electrical nerve via a catheter ablation procedure. A clinically
accepted method to detect AF is to observe ECG using
skin-mounted electrodes. With the help of simultaneous ECG
vectors, clinicians focus on the part of the heart to be investigated.
While clinically accepted AF detection considers more
than the increased beat-to-beat heart-rate variability (HRV), an
abrupt increase of HRV is a good indicator of a possible AF
rhythm. Once AF is suspected based on HRV, confirmation of
AF requires complete ECG signal analysis to uncover the presence
of P-wave in the ECG complex.
Recent trends in wearable watches provide an opportunity
to detect AF with an optical sensor as well as ECG, making it
possible to detect AF anywhere by
observing irregularity of the heart
rhythm. With an ECG watch, an
algorithm can detect beat-to-beat
irregularity and abnormality in the
PQRST complex. From PPG, an
algorithm can determine the surrogate
index for an R-R interval,
cardiac output, and flow rate to
assess for possible AF.
How to detect R-peak
from chest ECG
AF has traditionally been detected
using an ECG system, such as a
Holter Monitor which calls for a
patient to wear a ECG recording
device about 24 hours to 2 weeks
and to track daily activity for the
physician. Nowadays, a costly but
convenient method is to implant
an insertable cardiac monitoring
device, such as Medtronic Reveal,
into the upper chest. The ECG
is wirelessly transmitted to the
monitoring station, where a trained
clinician coupled with AI algorithms
interpret the ECG waveform
for a report sent to the responsible
cardiologist. Once the ECG data
stream is captured, an R-peak detection
algorithm such as the Pan-
Tompkins algorithm can detect
R-peak after rejecting respiration
or motion-related noise.
Fig. 2: ECG is transformed to averaged waveform to facilitate R-peak
detection.
1. Raw ECG signal and bandpass filtered signal. Bandpass filter
removes respiratory signal and filtered signal has high-frequency,
component-related QRS complex.
2. Differentiated signal of bandpass filter output will emphasize
high-frequency signal.
3. Differentiated signal is squared to preserve the power of the
signal.
Moving average window is applied and adaptive threshold is
calculated based on the amplitude of previous beats. Circles indicate
the detected R peaks.
There are several R-peak detection methods. However, many
of them need data after R-peak to confirm R-peak detection,
which prevents real-time signal processing. Out of these R-peak
detection algorithms, the Pan-Tompkins algorithm is unique,
enabling adaptive threshold and adaptive gain control features
necessary for a real-time on-chip algorithm.
Steve Koh is Principal Member of the Technical Staff at Maxim
Integrated - www.maximintegrated.com
Chris Hopwood is Program Manager and Andrew Burt
is Executive Business Manager, Industrial and Healthcare
Business Unit at Maxim Integrated.
www.eenewseurope.com eeNews Europe December 2019 News 31
/
/www.maximintegrated.com
/www.eenewseurope.com