A risk factor attention-based model for cardiovascular disease prediction.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Computational Medicine

Background:

• Cardiovascular disease (CVD) is a leading cause of mortality, necessitating advanced diagnostic tools.
• Electronic Medical Records (EMRs) contain valuable data for predicting CVD, but current NLP methods have limitations.
• Automated CVD prediction from EMRs is crucial for intelligent diagnosis and treatment.

Purpose of the Study:

• To develop an advanced model for predicting cardiovascular disease (CVD) using electronic medical records (EMRs).
• To overcome the limitations of existing natural language processing (NLP) methods in CVD prediction.
• To leverage both general EMR text and specific CVD risk factors for improved prediction accuracy.

Main Methods:

• Proposed a Risk Factor Attention-based Model (RFAB) integrating deep neural network attention mechanisms.
• Fused character sequences from EMR text with identified CVD risk factors.
• Utilized BiLSTM-CRF model for risk factor identification, labeling category and time attributes.

Main Results:

• The RFAB model demonstrated significant improvements in CVD prediction performance.
• Achieved a high F-score of 0.9586, outperforming existing related methods.
• Effectively utilized fine-grained information within EMRs for prediction.

Conclusions:

• RFAB effectively utilizes 12 key CVD risk factors and their associated EMR information.
• The model fuses risk factor details with character sequence information for accurate CVD prediction.
• RFAB offers a reliable approach for CVD prediction by leveraging detailed EMR data.