Projects ECG (Electrocardiogram) Monitoring

Development technology

Summary
Development of an analytical method that accurately measures various indexes (PQRST cycle and size) necessary for ECG (Electrocardiogram) diagnosis and processes them to quantified values.

Classify data in commonly used methods

Noise elimination with filtering technique such as Highpass Filtering

Actual ECG signals contain a lot of noise, which interfere with the accurate detection of Q R S peaks, etc.

Noise elimination by designing and applying appropriate filters step by step.

- Example of Filter Design

Calculate ECG diagnosis elements into digitized data by converting ECG signals into Hilbert to obtain Nyquist diagram and applying Circle Fitting.

Many studies have been done to develop algorithms for ECG peak detection, but show limitations on unstable signals.

Identify the ECG diagnosis elements accurately using the Hilbert transformation method, and solve the problem by deriving it using digitized data methods.(related part is applied for PCT patent)

- Examples of Traditional algorithm research

Detection Algorithm for ECG using Double Difference and RR Interval Processing

d1(i) = e(i+1) - e(i), i = 1,2... n-1
d2(j) = d1(j+1) - d1(j), j = 1,2... n-2
d(j) = [d2(j)]²

- Test results for some diseased patients

- R Peak Detection by existing method (example of a normal signal)
-> For normal signals, R peak detection shows excellent results.
-> Here, the peak is detected for values greater than 0.2.

-> It shows inappropriate results for unstable signals. Applying a 0.2 (threshold) value, as shown above, will result in missing peak and lowering the threshold hold value will result in too many peaks being detected.
■   ECG Graph
■   Complete R peak detection results using improved filters and Hilbert transformations

two-time differential for peak detection is generally employed, and the study above suggests that detection sensitivity is 99.8%, but only for normal signals. The two-time differential method has limitations in detecting the correct peak for unstable signals.

- Hilbert Conversion of ECG Signals

-> If you convert an ECG signal to Hilbert, you can obtain an imaginary part of the signal. The pink graph is the imaginary part of the ECG.

- Nyquist Diagram


-> If Imaginary part data obtained by ECG data and Hilbert transformation is drawn in X-Y Graph format, it is as above. This is called the Nyquist diagram.

- Circle Fitting and Calculating Numerical Data


-> Circle fitting the P R T peak from the Nyquist diagram and obtain the diameter of each part.
-> The diameter of the circle corresponds to the amplitude of P R T.

- Hilbert Transformation, Nyquist Diagram and 3D ECG

■   (Rear) : An imaginary graph obtained by Hilbert conversion
■   (floor) : raw ECG graph
■   (left-hand side) : Nyquist Black: 3D ECG Graph

How to get ECG Axis Deviation using Hilbert Transform

*** The data used here utilized the ECG database of 'Physikalisch-Technische Bundesanstalt(PTB), the National Metrology Institute of Germany'.

When a certain real signal is x(t) and the Hilbert-transformed signal is xˆ(t), the signal x_p(t) = x(t) + jxˆ(t) can be considered by extending it to the complex number.



The signum function of f is defined as follows:


The Fourier transform of the Hilbert transform of x(t) is defined as follows

F{(xˆ)}=−j sgn f⋅F{x(t)}

Now, it is possible to have Nyquist plot if the real part of xp(t) is the y-axis and the imaginary part of x_p(t) is the x-axis.

Typical ECG Signal is shown above.


The blue line was obtained using Hilbert Transform. Let's do this as an Imagine part. The pink line is drawn by obtaining the envelope using (Figure 1) Real part and (Figure 2) Imaginary part.

Figure 3. is a graph drawn to help you understand. This is often referred to as the Nyquist Plot, with ECG Signal's Real part on the Y-axis and Imaginary part on the X-axis.

There are several methods to obtain the axis deviation, but here, as in the figure above, the sum of vectors is obtained and expressed as the slope of the vector sum. The angle took the positive direction of the Y-axis to 0 degrees.

First, the Nyquist diagram for the 12 lead ECG is as follows.

By the method of Figure 4,5, eight data of healthy controls, myocardial infarction, and bundle branch block, respectively, are shown in Figures 6, 7, and 8 as follows.







-> Table 1 and 2 show the summary of above three cases.








In the case of Bundle branch blocks, left and right asymmetries occur frequently. This is shown in Table 3 compared to Healthy controls.



-> Figures 11 to 13 show V1 to V6 in a summary.