Page 118 - AIH-2-3

Artificial Intelligence in Health Opportunities for AI-based arrhythmia screening

the time interval between the T wave and the subsequent amplitude variability within specific time frames, durations
P wave. This measure is important because it reflects the and intervals between sequences of events, and correlation
separation between two key cardiac events, that is, the analysis between signals or between certain variables.
excitation pulse from the sinus node, represented by the The most critical temporal feature in ECG analysis is the
P wave, and the repolarization of the ventricles, which heart depolarization cycle period, which is determined by
generates the T wave. measuring the duration from one R wave to the next. Other
To successfully capture and analyze various cellular important features include the duration of individual waves
depolarization frequency patterns, multiple signal and the time interval between them (e.g., the TP interval:
processing approaches can be employed, including time interval between the T and P waves). The TP interval
analog processing, digital signal processing, or a mixed is important because it reflects the separation between two
signal processing format. Other techniques include peak important events, that is, the pulse rate of the sinus node,
detection, frequency determination, the time interval which is represented by the P wave, and the ventricular
between specific peaks analysis (either same function or repolarization, which generates the T wave. A broad range
different function [Figure 2]: e.g., QT separation, QQ of signal processing techniques, applicable in the time and
11,29-34
separation, and P-wave repetition rate [which can indicate frequency domains, are summarized in Table 1. While
atrial fibrillation]), pulse widths of various components in it is beyond the scope of this article to discuss each of these
the PQRST complex analysis, wavelet transform (however techniques, they can be utilized for different purposes, such
limited due to the temporal nature of pacing events), as signal pre-processing, noise reduction, and identifying
discrete wavelet transform, Butterworth filtering, discrete the pathological origins of signal deviations. For instance,
path transform, Fourier transform, short-term Fourier unique, idiosyncratic, or infrequent deviations from the
transform, Laplace signal transform, and compressed or normal ECG pattern may be detected using specialized
full matched filter analysis. approaches like modified matched filter analysis. The
spatial domain is particularly significant in 12-lead ECG
Time constants, such as heart rate and various interval and multidimensional decomposition techniques, such as
period durations (e.g., PR interval, QT interval, and the matched filter approach.
P-wave repetition rate), can be efficiently determined using
peak detection or Fourier transform techniques. These Various standard ECG characteristics can be resolved
computational methods are applicable to any ECG data using straightforward mathematical techniques. For
acquisition system. Notably, peak detection and heart rate instance, peak detection can quickly determine heart rate,
variability tests are decade-old techniques that have been while wavelet transform and matched filter approaches can
used in standard monitoring devices. define specific intervals and segments in the time domain
(Figure 1). However, due to continuous advancements
For preliminary analyses, various data banks containing in signal processing and innovative approaches, it is
healthy and pathological ECG recordings are available, impossible to cover all mechanisms within this article.
such as those hosted on PhysioNet (https://physionet.org/). Peak detection results related to the time domain can also
Collaborations with cardiology groups have also yielded be used for the recognition of certain arrhythmias. Besides,
highly specific data streams. All analytical signals in this atrial and VF require a more complex mathematical
study were acquired using routine clinical monitoring approach for early and proactive detection.
equipment. Some analyses were conducted as single-
blinded studies, where the analytical team was provided An important time-domain feature in ECG is the
with a well-established arrhythmogenic signal without duration of the QRS complex. In general, the QRS complex
knowing that it was abnormal. The physicians supplying is identified by its characteristic shape and relatively stable
the signals were fully aware of the deviations, which were time constant within the repetitive ECG pattern. In terms
often clearly defined. However, due to the early stages of of frequency content, the ECG waveform, including the
development, not all hidden characteristics and atypical QRS complex, is primarily confined to the high-frequency
ECG patterns have been investigated for diagnostic region, while the P and T waves represent the lower-
purposes at this time. frequency components. The ST segment, on the other
hand, is time-restricted and characterized by its low-
3.3. Signal analysis in amplitude, temporal frequency content. 8
framework, spatial framework, and frequency The frequency content of a normal ECG often differs
spectrum significantly from that of a pathological ECG. For instance,
Various approaches are employed to evaluate the temporal a normal heart rate ranges between 60 and 100 beats/min,
aspect of ECG signals, such as peak-to-peak analysis, whereas arrhythmias or fibrillation can result in heart rates

Volume 2 Issue 3 (2025) 112 doi: 10.36922/aih.8468