Fusion of Graph and Tabular Deep Learning Models for Predicting Chronic Kidney Disease

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Nephrology

Background:

• Chronic Kidney Disease (CKD) poses a significant global health burden.
• Accurate prediction of CKD progression is crucial for timely intervention and improved patient outcomes.

Purpose of the Study:

• To develop and evaluate an innovative fusion deep learning model for predicting CKD progression.
• To leverage both graph-structured and tabular data for enhanced predictive capabilities.

Main Methods:

• A fusion deep learning model integrating a Graph Neural Network (GNN) and a tabular data model was developed.
• The fusion model's performance was compared against standalone GNN, tabular models, and a baseline using accuracy, precision, recall, and F1-score.
• Hyperparameter optimization was performed using grid search for robust model comparison.

Main Results:

• The fusion model demonstrated superior performance across all evaluation metrics compared to individual models and the baseline.
• The GNN component contributed significantly to the enhanced predictive power, highlighting the value of graph-structured data.
• The fusion model showed consistent performance across various data splits, indicating robustness to dataset variations.

Conclusions:

• The proposed fusion deep learning model offers a more precise and reliable method for predicting CKD progression.
• This approach has the potential to advance precision medicine and improve the early detection and management of CKD.
• Integrating graph and tabular data enhances predictive accuracy for complex diseases like CKD.