BME354 - ECG I, Biopotential Amplfier

Advaitha Anne TA: Demi Shen

Abstract – The goal of the lab was to successfully build a amplifier circuit, the Common Mode Rejection Ratio (CMRR)
biopotential amplifier, and after verifying its use, gather ECG is calculated. This is done by assessing the differential and
data from a patient simulator, and from a human subject. With common mode gains (1) and can be represented also in decibel
the subject, the effects of jogging for various times were analyzed form (2). A larger CMRR signifies a better biopotential
to see changes in the ECG waveform and heartrate (HR). After amplifier circuit, removing the noise.
experimentation it was seen that male and females have different
resting heart rates, and that adding various exercise levels
impacted the HR and PR interval for the subjects. Overall, this
experiment proved the validity of the circuit to capture ECG
signals, though improvements could be made to reduce the noise
in data collection, and the identification of live ECG waveforms.

I. INTRODUCTION

To understand the electrical activity of the heart, an Fig. 1. Biopotential Amplifier Circuit
ECG is a common signal technique that provides insight into a
patient’s heart function. This electrical signal reflects the
action of the cardiac muscle as it depolarizes and re-polarizes 
during the cardiac cycle. The heart uses its the sinoatrial node
(SA) cells to initiate the action potentials that cause
contraction and the action potential spreads through the atria, (2)
and after a delay, to the ventricles after a delay. This action
potential occurs in a much smaller timeframe than the overall
III. RESULTS
heart contraction cycle. When analyzing the ECG of one
heartbeat, the waveform is seen to show the sequence of A. Data
depolarization and re-polarization of the heart, along with the
diastole and systole. This waveform can be seen in Figure 2. In building the biopotential amplifier, the circuit was tested to
show a normal sinus rhythm at 60 bpm, and also the effects of
of the lab manual. To correctly amplify and identify these
signals, this lab includes the constructions of a biopotential cell phone interference and jiggling the wires. These can be
seen in Figures 2-4 respectively. These outputs were analyzed
amplified, and subsequent testing with a TechPatient Cardio
ECG Simulator, and with a student’s ECG. The biopotential to ensure that the circuit was working properly and the output
amplifier circuit uses a commercial instrumentation amplifiers
INA126 and LF353 and works to successfully represent the
respective signal being inputted.

II. METHODS

A. Lab 6 Manual – BME 354

In this laboratory, exercises 4.1 to 4.4 from
laboratory manual[1] were followed. The only deviation was
Fig. 2. Biopotenial Amplifier Circuit with 60 bpm sinus rhythm
made was the lack of ISO-SWITCH device attached to the
leads with TechPatient Cardio Simulator. In constructing the
biopotential amplifier, the designed high-pass filter used right
before the output had values of R1 = 4.7kΩ, R2 = 150Ω, C =
0.1 μF, and the power supply was 12V for all the amplifiers.
The Rg used was 48.6 kΩ. This can be seen in Figure 1.
B. Equations
To be able to properly visual the ECG signal, the
biopotential amplifier circuit uses both instrumentation and
differential amplifiers. To quantify the performance of the
Fig. 3. Cell Phone Interferance B. ECG Figures and Data
Fig. 7. ECG Resting Human Data
has minimal noise and disturbance. To calcualte the CMRR,
the differential and common mode gains were determined.
Fig. 8. ECG Data – Zoomed in Resting
Using an input of a sinsoidal function with an amplitiude of
1mV, the differential gain was seen as the ampltiude of the
Fig. 9. ECG Data – 5 Seconds of Exercise
output (Figure 5), 480 mV. The common mode gain was seen
as the amplitude of the output when the circuit was modified
as seen in Figure 4. of the lab manual [1]. This was
determined to be approxmately 440 mV (Figure 6). The
CMRR was then calculated to be 1.09.
When testing the ECG on a resting human
subject, a signal recorded over 10 seconds, and displayed
multiple heartbeats. This can be seen in Figure 7. To analyze
the waveform, 1-3 heartbeats were selected from the resting
and each exercise period data (Figures 8-13). From these
results, the components were identified and the PR interval,
PR segment and ST segment were calculated (Table 1).

Fig. 4.
“Jiggling”
Wire Effect

Fig. 5.
Differential
Mode Gain
(100 mV
scale)

Fig. 6.
Common
Mode Gain
(200 mV

scale)
Fig. 10. ECG Data – 15 Seconds of Exercise

Fig. 13. ECG Data – 120 Seconds of Exercise

Fig. 14. Frequency Analysis of Resting ECG

TABLE I. PR INTERVAL, PR SEGMENTS, AND ST SEGMENTS

a. Exercise ECG Interval Data (seconds)
C
Time PR Interval PR Segments ST Segments
Resting 0.15 0.06 0.05

5 Seconds 0.25 0.04 0.04

15 Seconds 0.15 0.06 0.05

30 Seconds 0.12 0.05 0.05

60 Seconds 0.17 0.05 0.03

120 Seconds 0.13 0.03 0.02

Fig. 11. ECG Data – 30 Seconds of Exercise

Fig. 12. ECG Data – 60 Seconds of Exercise

alculated from Figure 8 - 13

 A frequency domain analysis of the resting ECG data from To help this, a deep notching of the noise at direct current can
Fig. 14 showed main peak frequencies at around 0-5 Hz, 60 be done, but this can lead to nonlinear phase response
Hz, and 120 Hz, with smaller peaks at 158 Hz, 180 Hz, 227 distortion, so designs such as those seen in [2] have been
Hz, 240 Hz, 300 Hz. 320 Hz, 360 Hz, and 410 Hz. An error created to help combat this noise.
bar plot was created with the means of each group and their If four groups employed the bandpass filter used in this
respective standard deviations, providing a cleaner visual. circuit to capture EEG signals and varied in sampling rates,
then they will also vary in sampling quality of the EEG signals.
Nyquist sampling requires the rate to be at least twice the
bandwidth of the maximum frequency in the signal, so in this
case, 2*100 Hz, or 200 Hz. Any sampling rates below this
would be in violation of Nyquist and would not properly
sample the signal. Groups with sampling rates of 5 kHz and
300 Hz would be able to adequately acquire the 85.5 Hz
waveform (> 2*85.5 Hz), and the sampling would be adequate
for reconstruction, but more specificity can be completed based
on the specifics of the signal variation and data noise.
To analyze the effect of the different exercise groups on the
heart rate and PR intervals, an analysis was done to show the
error for the individual group data, all male data, and all
female data. The plots in Figures 15 and 16 show a plot of the
average heart rate and PR interval , along with the associated
error bars for each group. The error bars in each plot show the
variability with the individual, or overall gender measuring.
This was not differentiated by each individual when doing
gender analysis, as those error bars were messy graphically and
did not provide much insight. If the data came from
independent trials, then the error would be reduced, as the
Fig. 15. Errorbar Plot of HR Data continuous sampling makes the sampling not truly
independent. To see if male students and female students has
Fig. 16. Errorbar Plot of PR Interval Data
the same resting heart rate, a Welch’s t-test was completed, as
unequal variances were assumed, as we had no substantial
proof otherwise. With a P-value of 0.000778, the null
IV. DISCUSSION hypothesis was rejected, supporting that males and females
have different resting heart rates.
During a frequency domain analysis of the ECG signal
peak frequencies can be seen between 0-5 Hz, with a Examining the interval data for the subject from our group,
significant peak at 60 Hz. The smaller peaks at higher there seems to be no real significant patterns signifying a
frequencies are most likely due to other noise in the signal, due relationship between interval lengths and the period of
to the various factors effecting the precision of the data exercise. But considering how close the numbers are, and how
collection and instrumentation. The high noise at 60 Hz is the data collected was relatively noisy, making it difficult to
difficult to filter out with normal RC filters, even though it is specify the bounds of the waveform components, this does not
common mode noise that is diminished with the differential necessarily mean that the length of exercise does not affect the
amplifier. This due to the fact that electromagnetic field from PR interval, PR segment, and ST segment lengths. To further
external electricals, or powerlines, can cause this interference, explore this, data from the entire class was utilized in statistical
and these are necessary for the ECG collection [2]. The RC analysis. By subtracting the resting PR interval from each and
filters do not have the high attenuation needed to combat this. all corresponding exercise time data, giving a null hypothesis
that there was no difference between each of the groups, a
one-way ANOVA was performed using the AnalysisToolPak
on Microsoft Excel. The null hypothesis This resulted in a P-
value of 0.00013 and as F > F crit (5.80 >2.38). This was done
under the assumptions that the groups were independent, the
dependent variables were approximately normally distributed,
a homogeneity of variances, and there were no significant
outliers. Another statistical test using a repeated measures
design would be necessary to support this conclusion, as the
use of one patient for each of the independent groups for each
samples affects the assumption of group and sample
independence. This study also does not have a varied
population to ensure that approximate normality of the
populations exists, so better sampling would also need to be QRS waveform, specifically QRs [3], as the Q-wave was more
included to validate this conclusion. negative generally. This pattern was seen through all the
varying intervals it was collected, showing it was likely a
To see if a relationship exists between jogging and the heart characteristic of the subject. The results for the individual
rate, a one-way ANOVA was also completed for the heart rate group and the overall male and female data showed that the
data collected for the entire class. Using a similar process of genders had different resting heart rates through a t-test, and
subtracting the corresponding resting heart rate for each one-way ANOVA’s showed that exercise affected the PR
individual from each exercise time category, the null Intervals and HR. This analysis applied to the students of BME
hypothesis was that all the individual exercise data groups 354 not this semester and needs to be tested across larger data
were the same and had no significant variation. With a P-value sets to extend applicability across other larger groups of
of 4.8319 * 10-54 and F > Fcrit ( 73.27 > 2.384), the null people.
hypothesis was very safely rejected. While this has the same
issues in the limitations of the study and the same assumptions
as the test done on the PR intervals, the significantly higher F- REFERENCES
value shows that is less likely that more specified statistical
testing will reverse the rejection of this null hypothesis. [1] Duke BME 354 Lab 1 Protocol: Lab Equipment Overview, Spring,
2019.
[2] Asgari S, Mehrnia A (2017) A novel low-complexity digital filter design
V. CONCLUSIONS for wearable ECG devices. PLoS ONE 12(4): e0175139.
The lab showed the presence of ECG signals from the bio- https://doi.org/10.1371/journal.pone.0175139
amplifier circuit. Initially, the circuit was verified with a sinus [3] Rawshani, A. (2019, May 31). The QRS complex: ECG features of the
rhythm, and other noise factors were visualized. Then, after Q-wave, R-wave, S-wave & duration. Retrieved March 4, 2020, from
https://ecgwaves.com/ecg-qrs-complex-q-r-s-wave-duration-interval/
successfully using a patient simulator, human data was
collected within the lab group. The individual results showed a