9  Review Questions

Open “Answer” below each question to see the answer and explanation. Think it through yourself first before opening it.

9.1 Basics

Question 1

Which electrical activity of the heart does each of the P wave, PR interval, QRS complex and T wave of the ECG correspond to?

  • P wave: atrial depolarization
  • PR interval: from the start of atrial excitation to the start of ventricular excitation (mainly the conduction delay in the atrioventricular node)
  • QRS complex: ventricular depolarization
  • T wave: ventricular repolarization

(Chapter 1)

Question 2

Using the potentials of the individual electrodes, derive the relationship that holds among standard limb leads I, II and III.

Since I = Φ_LA − Φ_RA, II = Φ_LL − Φ_RA and III = Φ_LL − Φ_LA,

I + III = (Φ_LA − Φ_RA) + (Φ_LL − Φ_LA) = Φ_LL − Φ_RA = II

(Chapter 1)

Question 3

A common-mode voltage (power-line noise) of 1 V is applied to an instrumentation amplifier with a CMRR of 80 dB. How many mV of spurious differential signal does this produce, referred to the input?

80 dB = a factor of 10⁴, so 1 V / 10⁴ = 0.1 mV. If the R wave is 1 mV, this corresponds to 10% of it. (Chapter 2)

Question 4

A signal containing 60 Hz power-line noise is sampled at 100 Hz (without an anti-aliasing filter). At what frequency is the power-line noise observed?

The Nyquist frequency is 50 Hz. 60 Hz is folded back and observed as |60 − 100| = 40 Hz. After sampling, it cannot be distinguished from a genuine 40 Hz component. (Chapter 4)

Question 5

Explain why a pulse oximeter uses two wavelengths, red and infrared, and takes the ratio of their AC/DC ratios.

  • Oxygenated and deoxygenated haemoglobin have opposite relative absorption at red and infrared, so their proportions (oxygen saturation) can be estimated from the ratio of absorption at the two wavelengths
  • Using the AC component extracts information about arterial blood only, which pulsates
  • Dividing by DC cancels differences between wavelengths due to the device and the individual, such as LED brightness, skin and tissue thickness and photodetector sensitivity

(Chapter 3)

9.2 Applied

Question 6

Why is it recommended that the cutoff frequency of the high-pass filter in a diagnostic ECG be set as low as 0.05 Hz?

A non-linear-phase high-pass filter shifts the phase of low-frequency components substantially when its cutoff frequency is raised, distorting the ST segment and T wave (spurious ST depression or elevation). Because changes in the ST segment are important for diagnosing myocardial ischaemia, the cutoff frequency is set very low so that this distortion does not occur. With a linear-phase (zero-phase) filter, up to 0.67 Hz is acceptable. (Chapter 5)

Question 7

HRV frequency-domain analysis was performed on 1 minute of RR-interval data. Roughly what is the frequency resolution? Also describe the problem with evaluating the LF band (0.04–0.15 Hz) at this resolution.

Resolution ≈ 1 / 60 s ≈ 0.017 Hz. The 0.11 Hz-wide LF band contains only a few frequency points, and a component near the lower limit of 0.04 Hz (period 25 s) completes only 2–3 cycles in 1 minute, so the estimate is coarse and unstable. Splitting into segments with Welch’s method lowers the resolution even further. (Chapter 6)

Question 8

In the Pan-Tompkins algorithm, explain the purpose of the derivative and squaring applied after the band-pass filter.

The QRS complex has very steep slopes compared with the P and T waves. Differentiation extracts the slope, and squaring makes it positive while further emphasizing large slopes, greatly widening the ratio between the QRS and the T wave, P wave and residual baseline wander. Squaring also makes detection possible regardless of the QRS polarity (upward or downward). (Chapter 7)

Question 9

When the heart rate of a wristwatch-type device was compared with a chest ECG, the correlation coefficient was r = 0.98. Can you conclude that “this device is very accurate”?

No. The correlation coefficient only expresses the strength of a linear relationship, not agreement. r is high even with systematic error (always offset by a constant) or proportional error, and a wide measurement range tends to make r large. You need to compute the bias and limits of agreement with Bland-Altman analysis and compare them with an acceptable range decided in advance according to the intended use. Evaluation for each condition (rest, exercise) is also required. (Chapter 8)

9.3 Advanced

Question 10

Give two reasons why power-line noise tends to be large in wearable ECG devices that use dry electrodes, and describe a countermeasure for each.

  1. The electrode–skin contact impedance is high → the signal is easily attenuated by the voltage divider formed with the amplifier’s input impedance, so noise has a relatively larger effect. Countermeasure: make the input impedance very high (buffered electrodes, active electrodes)
  2. The impedances of the two electrodes are unbalanced and fluctuate → the common-mode voltage is converted into a differential signal that the amplifier’s CMRR cannot remove. Countermeasures: lower the common-mode voltage itself, e.g. with a driven right leg; stabilize electrode contact (pressure, material); raise the common-mode input impedance

A digital notch filter can also be used as a supplement, but noise large enough to saturate the amplifier cannot be removed afterwards. (Chapters 2 and 5)