A weighted patient network-based framework for predicting chronic diseases using graph neural networks.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Computational Biology

Background:

• Chronic disease prediction is vital for healthcare management.
• Existing machine learning methods face challenges with high-dimensional data and bias.
• Developing robust predictive models is crucial for early intervention.

Purpose of the Study:

• To propose a novel framework for chronic disease prediction using Graph Neural Networks (GNNs).
• To address limitations of existing methods, including data bias and high dimensionality.
• To enhance the accuracy and robustness of chronic disease prediction models.

Main Methods:

• Projecting a patient-disease bipartite graph to create a weighted patient network (WPN).
• Utilizing GNN-based techniques to build prediction models on extracted WPN features.
• Comparing GNN model performance against traditional machine learning methods for cardiovascular and chronic pulmonary diseases.

Main Results:

• The GNN framework significantly enhances chronic disease prediction accuracy.
• The model with attention mechanisms achieved 93.49% accuracy for cardiovascular disease and 89.15% for chronic pulmonary disease.
• Visualizations confirmed the discriminative strength of the GNN framework across patient cohorts.

Conclusions:

• The proposed GNN framework offers a robust and accurate approach to chronic disease prediction.
• This method effectively mitigates issues of high dimensionality and bias present in traditional models.
• The framework has the potential to improve health management systems for at-risk populations.