Artificial intelligence and multimodal diagnostic approaches in cardiovascular disease

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Cardiovascular diseases pose a significant global health burden.
• Conventional diagnostic methods face limitations in accuracy and efficiency.
• Emerging artificial intelligence (AI) models offer potential advancements in cardiovascular diagnostics.

Purpose of the Study:

• To evaluate the clinical impact and applicability of AI models in cardiovascular diagnosis.
• To compare the diagnostic accuracy, efficiency, and reliability of AI with conventional methods.
• To identify barriers and facilitators for AI integration into clinical practice.

Main Methods:

• A critical literature review of retrospective studies, multicenter trials, and external validations.
• Analysis of AI applications in cardiac imaging, electrocardiographic/phonocardiographic signals, and biomarkers.
• Inclusion of machine learning and deep learning algorithms.

Main Results:

• AI models in cardiac imaging exceeded 90% accuracy for segmentation and dysfunction detection.
• Deep learning models achieved an area under the ROC curve of ~0.99 for predicting atrial fibrillation and ischemic heart disease.
• Ensemble models with integrated data reached >95% diagnostic accuracy, but external validation showed decreased performance and generalizability issues.

Conclusions:

• AI demonstrates transformative potential in cardiovascular diagnostics, improving accuracy and accessibility.
• Barriers include reduced performance in external validations, limited generalizability, and clinician reluctance due to explainability.
• Robust validation, workflow integration, and enhanced explainability are crucial for AI adoption in routine care and precision medicine.