Predicting complications and mortality in myocardial infarction patients using a graph neural network model

Area of Science:

• Cardiology
• Artificial Intelligence
• Data Science

Background:

• Myocardial infarction (MI) complications necessitate accurate risk stratification.
• Existing models often predict single composite endpoints and neglect patient similarities and temporal data from electronic health records.
• There is a need for advanced models that can simultaneously predict multiple distinct outcomes.

Purpose of the Study:

• To develop and evaluate a novel graph neural network (GNN) framework for simultaneous prediction of 12 post-MI complications and in-hospital mortality.
• To leverage patient similarities and temporal dynamics in electronic health records for improved risk assessment.
• To provide an interpretable model for early, individualized risk stratification in acute cardiac care.

Main Methods:

• Development of a graph neural network framework integrating three key innovations:
• 1. A density-adaptive K-nearest neighbor graph for capturing patient similarities.
• 2. Dual-branch short- and long-term temporal encoders with dynamic gating.
• 3. Cross-modal attention for fusion of multi-scale temporal features.
• Model evaluation on a dataset of 1700 patients with MI complications.

Main Results:

• The GNN model achieved an average AUC of 0.7330 across 12 distinct complications.
• Mortality prediction performance was notably high, with an AUC of 0.8828.
• SHAP analysis and attention weights identified age, serum sodium, and dynamic laboratory trends as significant predictors, aligning with clinical knowledge.

Conclusions:

• The developed interpretable GNN framework offers a significant advancement in simultaneously predicting diverse post-MI complications and mortality.
• This approach enhances individualized risk assessment for patients with acute cardiac conditions.
• The model's ability to integrate patient similarities and temporal data holds promise for improving clinical decision-making in cardiology.