Screening for Left Ventricular Hypertrophy Using Artificial Intelligence Algorithms Based on 12 Leads of the...

Area of Science:

• Cardiology
• Medical Artificial Intelligence
• Diagnostic Accuracy

Background:

• Artificial intelligence (AI) offers potential solutions to global physician shortages.
• Challenges in AI implementation include usability, privacy, inequality, and misdiagnosis.
• This review specifically examines AI's role in diagnosing left ventricular hypertrophy (LVH) using 12-lead electrocardiograms (ECGs).

Purpose of the Study:

• To review the diagnostic accuracy of AI algorithms in detecting LVH from ECGs.
• To compare AI performance against traditional ECG criteria.
• To identify factors influencing AI efficacy in cardiovascular diagnostics.

Main Methods:

• A systematic literature review following PRISMA 2020 guidelines.
• Searches conducted across major databases (PubMed, CENTRAL, Google Scholar, Web of Science, Cochrane Library).
• Included randomized controlled trials, observational studies, and case-control studies; registered in PROSPERO (531468).

Main Results:

• Seven studies were included in the meta-analysis.
• Co-CNN models demonstrated the highest weighted average sensitivity (0.84).
• AI models showed variable but comparable accuracy to traditional ECG criteria (SLV, CV), with demographic factors influencing performance.

Conclusions:

• AI algorithms can analyze ECG data rapidly with high sensitivity for LVH detection.
• The diagnostic performance of AI models is influenced by clinical data and training populations.
• Future research must prioritize diverse ECG data to enhance AI generalizability in cardiology.