Page 126 - AIH-2-3

Artificial Intelligence in Health Opportunities for AI-based arrhythmia screening

and temporal pathological conditions. This provides include but are not limited to, heart attack, arrhythmias, heart
additional information beyond standard cardiac rhythm failure, cardiomyopathy, and heart valve disease. Additional
identifiers, such as heart rate and various specific interval diagnoses of other pathological conditions, including several
measurements. rare or atypical cardiac depolarization patterns, are not
To isolate the true depolarization sequence, noise must listed due to their infrequent occurrence and the current
be eliminated using techniques such as cutoff filtering, limitations of the ongoing research and development
frequency filtering, wavelet filtering, and Wiener filtering. program. Furthermore, many arrhythmias have yet to be
These methods can be applied to eliminate motion artifacts, investigated for AI-based discovery. For quantifiable time-
impedance mismatches, and tissue conductivity variation. specific conditions, refer to Figure 1.
The data acquisition device will also incorporate 5. Conclusion
electronic mechanisms for signal (pre-) processing, Artificial intelligence-supported diagnostics offer a
including gain adjustment, offset correction, frequency powerful tool for rapid patient screening and identification
filtering (primarily low-pass), and hardware-based of cardiac abnormalities. However, a thorough investigation
automated noise reduction. 65,66 In addition, frequent and
sequential repetition of deviations in the ECG pattern into the root cause of these deviations is essential to
65,66
can be analyzed using Fourier analysis, wavelet filtering, develop a comprehensive, patient-specific treatment.
and matched filter analysis. The wavelet, matched filter, The growing recognition of AI’s accuracy and reliability
and Wiener filtering analysis rely on an accurate base in diagnostics is gaining widespread acceptance. In
template for analysis, which must be customized to match particular, AI-driven screening and risk-stratification,
based on comparison with specific signal patterns of
the individual’s specific history. The choice of template a broad range of pathological cardiac conditions, has
used for screening and diagnosis is typically based on
the boundary conditions of the patient group being demonstrated significant value. These tools not only
investigated, such as athlete status, activity level, age, improve diagnostic precision but also save time in clinical
gender, genetics, weight, habits, and disabilities. Women settings. Time-domain and frequency-domain filtering
are often misdiagnosed due to the use of male-centered and analyses have long been used with excellent results.
cardiac rhythm templates. Therefore, great care must However, the introduction of more advanced techniques,
be taken to apply the appropriate screening boundary such as wavelet analysis and matched filtering, has enabled
conditions and construct a respective diagnostic template the identification of complex disease patterns with higher
based on available prior monitoring data for the individual accuracy. AI-based diagnosis can statistically determine
or a similar group. By utilizing existing data streams under the prevalence of certain arrhythmogenic conditions by
healthy conditions, neural network-generated templates matching ECG data with predefined templates or groups
can be used to detect deviations and reconstruct a template of arrhythmias. Each classification may encompass various
that matches the suspected pathological condition through cardiac rhythm morphologies. In addition, AI can offer
machine learning analysis. details about the duration of pathological events and the
frequency of specific phenomena over time, which may
The use of AI and deep learning techniques can indicate the presence of one or more arrhythmias. The
be enhanced through convolutional neural network feasibility study described here provides preliminary
processing to provide scheduled updates, as well as to insights into the use of AI for cardiac health screening.
verify and validate the sampling templates used in each Nonetheless, further analysis by a physician, incorporating
analytical procedure for the individual patient and their the patient’s history and likely additional tests, will lead to a
corresponding patient group selected as the baseline. 46-59 patient-specific diagnosis. Alternatively, the physician may
The automated ECG analysis routines currently extract treat the arrhythmia spectrum identified by AI as a broad-
the following values from the data: corrected QT interval, based issue, addressing it with a single, multipurpose
heart rate, P height, PQR-interval, QRS width, QT interval, medication. 67,68 Several drugs currently available on the
R height, RR interval, and ST interval. 61 market can effectively manage a wide range of cardiac
Several verified diagnoses, as well as several desired rhythm problems. However, certain conditions, such as
requirements for cardiac health monitoring, are listed in VF, cannot be controlled pharmacologically and require
Table 2. Some of these diagnostic modalities are applicable to an implantable cardioverter defibrillator. Similarly,
3-lead, 5-lead, and 12-lead ECG. 64-66 In addition, numerous other arrhythmias that cannot be well-controlled by
diagnostic techniques are currently being developed and are pharmaceutical means require the use of a pacemaker,
in various stages of preparation for commercial release under as determined based on follow-up examination by the
regulatory constraints. The associated pathological conditions physician. 67-74

Volume 2 Issue 3 (2025) 120 doi: 10.36922/aih.8468