A kardiológiai ellátást jelentősen átalakítják az új mesterséges intelligencia alkalmazások, amelyek a meglévő diagnosztikai eszközök pontosságát növelik. A mélytanulási modellek képesek a mellkasi röntgenfelvételeket szűrőeszközzé alakítani a mitralis stenosis kimutatására, míg a CT-képeket elemző gépi tanulás a szívinfarktus kockázatát jelző plakkokat azonosítja.
A kutatók a generatív AI alkalmazását is vizsgálták: a ChatGPT-4o modellt tapasztalt kardiológusokkal hasonlították össze a koszorúér-angiográfia értelmezésében. Emellett az XGBoost és CatBoost algoritmusok a műtétek utáni kockázatok becslésében segítenek, míg az úgynevezett magyarázható AI (XAI) keretrendszerek érthetővé teszik az orvosok számára az egyéni kockázati tényezőket.
A szakértők szerint a klinikai gyakorlatba történő sikeres átültetéshez elengedhetetlen, hogy a komplex AI-modellek illeszkedjenek a meglévő orvosi munkafolyamatokba, és a végső döntés mindig az orvos kezében maradjon.
Az eredeti szöveg (European Medical Journal)
This site is intended for healthcare professionals
ARTIFICIAL INTELLIGENCE (AI) is reshaping cardiovascular medicine by improving diagnostic precision, risk stratification, and clinical decision-making, according to a new editorial published in Open Heart. The editorial examines emerging applications of AI in cardiovascular care, highlighting research that could help bridge the gap between algorithm development and clinical practice.
AI-based approaches are expanding the capabilities of established diagnostic tools, enabling clinicians to identify abnormalities that may be difficult to detect through conventional assessment.
Research highlighted in the editorial demonstrated how deep learning could transform chest X-rays into screening tools for mitral stenosis, potentially improving access to cardiovascular diagnostics.
Machine learning applied to coronary CT imaging also enabled identification of high-risk plaque characteristics that traditional visual assessment may overlook, offering opportunities for more precise myocardial infarction risk prediction.
Another emerging application involved generative AI in coronary angiography interpretation. Researchers compared ChatGPT-4o with experienced interventional cardiologists, exploring the potential of large language models to support complex clinical assessments.
Additional studies investigated radiomics and pulse wave analysis for evaluating atrial fibrillation recurrence risk and central haemodynamics without invasive procedures.
Beyond diagnosis, machine learning is advancing cardiovascular risk stratification by moving beyond conventional population-based scoring systems.
The editorial highlighted gradient-boosting algorithms, including XGBoost and CatBoost, which were investigated for guiding coronary imaging decisions and estimating 30-day risks following non-cardiac surgery.
Importantly, explainable AI frameworks, including Shapley additive explanations, were used to clarify the factors influencing individual risk predictions, addressing concerns about the interpretability of complex algorithms.
Unsupervised clustering approaches were also explored for heart failure phenotyping, potentially enabling more individualised assessment beyond traditional ejection fraction classifications.
Systematic reviews and meta-analyses examining AI-based atrial fibrillation detection and cardiac CT demonstrated the growing evidence base supporting AI-assisted cardiovascular diagnostics.
However, the editorial emphasised that successful clinical implementation requires integration into existing workflows while maintaining clinician oversight.
The authors concluded that the next stage of cardiovascular innovation should focus on translating increasingly sophisticated AI models into practical, interpretable tools that support clinical decisions and more personalised patient care.
Tang Y et al. Navigating the intelligence revolution: transforming cardiovascular care through AI-driven innovation. Open Heart. 2026;13(1):e004160.
Each article is made available under the terms of the Creative Commons Attribution-Non Commercial 4.0 License.
Rate this content's potential impact on patient outcomes
No votes so far! Be the first to rate this content.
