CardioNet: a manually curated database for artificial intelligence-based research on cardiovascular diseases

Area of Science:

• Biomedical Informatics
• Cardiology
• Artificial Intelligence

Background:

• Cardiovascular diseases (CVDs) pose diagnostic challenges, with easily overlooked risk factors.
• Artificial intelligence (AI) offers potential for early CVD prediction and personalized treatment.
• Specialized, curated databases from electronic health records (EHRs) are crucial for applying AI to CVD data.

Purpose of the Study:

• To develop CardioNet, a specialized database for cardiovascular diseases (CVDs).
• To enable AI-driven insights for improved CVD management and patient care.
• To facilitate multi-center research through standardized, interoperable data.

Main Methods:

• Collected anonymized EHR data from 748,474 patients with CVDs.
• Integrated structured EHR data with unstructured medical examination readings.
• Applied natural language processing to extract and structure key CVD variables.
• Standardized data using a common data model for multi-center research interoperability.

Main Results:

• Established CardioNet, a comprehensive database for AI model training in CVDs.
• The database includes 27 tables, a code-master, and a descriptive table for data utilization.
• CardioNet integrates EHR data and digital test readings for clinical management support.

Conclusions:

• The CardioNet database is established and specialized for cardiovascular diseases (CVDs).
• It supports ongoing multi-center research initiatives in CVD.
• CardioNet serves as a foundational resource for future CVD research and AI applications.