Analysis of Current Applications of Artificial Intelligence in Cardiology Diagnosis with Focus on Explainable AI, Holter Monitoring and Blood Biomarkers

Abstract

This systematic review examines current applications of artificial intelligence in cardiovascular diagnosis, with particular emphasis on explainable AI (XAI) methodologies, blood biomarker analysis, and Holter monitoring applications. A comprehensive literature search was conducted using PubMed database for articles published between 2020–2025, focusing on studies that combine AI with cardiovascular diagnostic data. From an initial dataset of 195 articles, 29 studies met the inclusion criteria after systematic screening. The analysis reveals that only 48.3% of cardiovascular AI studies incorporate explainable AI methods, with SHapley Additive exPlanations (SHAP) being the most prevalent approach (64.3% of XAI studies). Risk assessment applications (27.6% of studies), emphasize preventive rather than diagnostic approaches. Tree-based algorithms, particularly XGBoost, Random Forest, and CatBoost, are the preferred methodologies (58.6% of studies). Blood biomarker integration shows significant promise across multiple applications, while Holter monitoring remains underutilized. The findings highlight a critical gap between AI model development and clinical implementation requirements, particularly regarding the EU AI Act's mandates for transparency in healthcare applications. This review aims to provide insights for future development of multimodal, explainable AI systems that integrate diverse cardiovascular data sources while meeting both clinical needs and regulatory requirements.

Ähnliche Arbeiten