Multiview Deep Learning-Based Efficient Medical Data Management for Survival Time Forecasting.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Computational Biology

Background:

• Remote medical management increasingly relies on data-driven approaches for tasks like survival time forecasting.
• Current methods often lack multimedia information, limiting the depth of analysis in healthcare data.
• Intelligent algorithms can enhance healthcare management by monitoring patient physical characteristics.

Purpose of the Study:

• To propose an efficient medical data management framework (MDL-MDM) using multiview deep learning for survival time forecasting.
• To enhance feature representation and knowledge discovery in remote healthcare by integrating diverse data perspectives.
• To address the challenge of limited multimedia information in purely data-driven medical scenarios.

Main Methods:

• Encoding basic monitoring data of patient body indexes as the foundation for forecasting.
• Developing a hybrid computing framework by combining Convolutional Neural Network (CNN), Graph Attention Network (GAT), and Graph Convolutional Network (GCN).
• Implementing a multiview feature learning framework through the ensemble of these neural network models.

Main Results:

• Experiments were conducted on a realistic US cancer patient dataset.
• The proposed MDL-MDM framework demonstrated improved survival time forecasting.
• The system achieved a 1% to 2% reduction in prediction error compared to existing methods.

Conclusions:

• The multiview deep learning approach effectively enhances feature representation for medical data management.
• MDL-MDM provides an efficient solution for survival time forecasting in remote healthcare settings.
• The framework shows significant potential for improving knowledge discovery and prediction accuracy in clinical applications.