Przegląd metod sztucznej inteligencji jako narzędzia diagnostycznego w analizie obrazów 2D z echokardiografii przezklatkowej, w tym ultrasonografii point-of-care u dorosłych pacjentów
Folia Cardiologica · 2025
DOI: 10.5603/fc.102765 ↗Kurz gefasst
Fachartikel in der Zeitschrift Folia Cardiologica, 2025. Verfasst von Gabriela Bernadeta Orzeł-Łomozik und 9 weiteren Autor:innen.
Abstract (Originalsprache)
OBJECTIVE: The aim of this systematic review was to evaluate whether artificial intelligence (AI) tools, including machine learning (ML) and deep learning (DL) methods, could effectively assist in the interpretation of two-dimensional transthoracic echocardiography (2D TTE) images in acute cardiac conditions among adult patients. INCLUSION AND EXCLUSION CRITERIA: Original studies were included if they applied AI methods to the analysis of 2D TTE images in adult patients requiring urgent cardiological assessment. Studies had to compare AI-derived results with expert assessments or other reference methods. Excluded studies involved pediatric populations, animal studies, exclusively transesophageal echocardiography (TEE), three-dimensional (3D) imaging, or lacking quantitative data on AI effectiveness. INFORMATION SOURCES: The PubMed and Web of Science databases were searched (last search: December 2023) using logical operators to combine keywords (“AI” or “machine learning” and “echocardiography” and “shock”). After removing duplicates, titles and abstracts were screened, followed by full-text analysis. DATA ANALYSIS METHODS: Given the diversity of AI algorithms, patient populations, and outcomes, qualitative synthesis was performed. Where at least two studies reported consistent measures (LVEF, left ventricular ejection fraction), a descriptive comparison of diagnostic values (sensitivity, specificity, correlation coefficients) was conducted. MAIN RESULTS: Across 15 included studies, AI significantly enhanced emergency echocardiographic diagnostics, achieving accuracy and reproducibility comparable to expert assessments. Algorithms effectively supported the assessment of LVEF (correlation approximately 0.80–0.90 with experts), pericardial effusion detection (sensitivity up to 89%, specificity up to 92%), and automatic echocardiographic projection classification (accuracy up to 96%). Limitations included poorer image quality in emergency conditions and limited, often single-center datasets. CONCLUSIONS: This review confirms the high potential of AI in point-of-care echocardiography (POCUS), being particularly beneficial for less experienced operators under time constraints. Further multicenter studies with diverse populations are necessary to fully evaluate the generalizability and clinical impact of such solutions (time to treatment initiation, mortality).
Zitieren
APA
Orzeł-Łomozik, G. B., Łomozik, Ł., Kampka, Z., Zygmunt, N., Gondko, D., Roman, J., Drabczyk, M., Gruchlik, B., jr, M. P., & Mizia-Stec, K. (2025). Przegląd metod sztucznej inteligencji jako narzędzia diagnostycznego w analizie obrazów 2D z echokardiografii przezklatkowej, w tym ultrasonografii point-of-care u dorosłych pacjentów. Folia Cardiologica. https://doi.org/10.5603/fc.102765
Vancouver
Orzeł-Łomozik GB, Łomozik Ł, Kampka Z, Zygmunt N, Gondko D, Roman J, et al. Przegląd metod sztucznej inteligencji jako narzędzia diagnostycznego w analizie obrazów 2D z echokardiografii przezklatkowej, w tym ultrasonografii point-of-care u dorosłych pacjentów. Folia Cardiologica. 2025. doi:10.5603/fc.102765
BibTeX
@article{Orzeomozik2025,
author = {Orzeł-Łomozik, Gabriela Bernadeta and Łomozik, Łukasz and Kampka, Zofia and Zygmunt, Natalia and Gondko, Daniel and Roman, Jakub and Drabczyk, Mateusz and Gruchlik, Bartosz and jr, Maciej Pruski and Mizia-Stec, Katarzyna},
title = {Przegląd metod sztucznej inteligencji jako narzędzia diagnostycznego w analizie obrazów 2D z echokardiografii przezklatkowej, w tym ultrasonografii point-of-care u dorosłych pacjentów},
journal = {Folia Cardiologica},
year = {2025},
doi = {10.5603/fc.102765}
}
EUVIMED