Poster 1009: Artificial Intelligence in Ejection Fraction Measurement: A Systematic Review of Current Applications, Validation Studies, and Future Directions

Background: Left ventricular ejection fraction (LVEF) measurement is crucial for cardiovascular disease management. While Modified Simpson's Biplane method remains the standard, it is time-consuming and operator-dependent. This review evaluates artificial intelligence's potential to automate and standardize EF measurements, improving efficiency and reproducibility in clinical practice.

Methods: The methodology involves a systematic literature review of studies published between 2015-2024, focusing on artificial intelligence applications in ejection fraction measurement. Primary databases searched include PubMed and Google Scholar using keywords: "artificial intelligence AND ejection fraction," "deep learning AND LVEF measurement," and "automated AND ejection fraction assessment."

Inclusion criteria:
- Original research articles
- Validation studies
- Comparative analyses between AI and manual methods
- English language publications
- Peer-reviewed journals

Exclusion criteria:
- Case reports
- Conference abstracts
- Opinion pieces
- Studies without quantitative results

Data extraction will focus on AI methodology, accuracy metrics, clinical validation, and comparison with conventional MBS method.

Outcome: The implementation of artificial intelligence in measuring ejection fraction has demonstrated significant advantages over traditional manual methods. Multiple validation studies show high feasibility rates (76-91.5%) across diverse patient populations, with strong correlation coefficients (0.81-0.92) when compared to expert measurements. A notable benefit is the 77% reduction in acquisition and processing time, saving approximately 5.3 minutes per examination while maintaining accuracy comparable to conventional methods.

AI-based tools have shown superior reproducibility compared to inter-observer measurements, effectively reducing operator dependency and measurement subjectivity. The technology provides real-time feedback during scanning, enabling immediate assessment and potential adjustments. This standardization of EF measurements particularly benefits resource-limited settings and non-specialist facilities.

However, certain limitations exist, including variable success rates in different clinical scenarios, accuracy challenges with larger left ventricular volumes, and dependence on image quality. The studies also noted limited validation in specific patient populations, such as those with severe obesity or arrhythmia.

Despite these constraints, the evidence suggests that AI-based EF measurement tools significantly improve efficiency, reproducibility, and standardization in cardiac assessment workflows, indicating a promising future in healthcare delivery.

Conclusion: AI-based EF measurement tools demonstrate significant promise in healthcare delivery through improved efficiency (77%-time reduction), enhanced reproducibility (correlation coefficients 0.81-0.92), and standardized measurements (76-91.5% feasibility rate). Despite some limitations, these advances suggest potential for widespread implementation across various clinical settings, ultimately improving patient care.