Skip to main content
. 2025 Aug 11;16:1596408. doi: 10.3389/fneur.2025.1596408

Figure 2.

Diagram of the Dynamic Medical Graph Framework (DMGF) showing the process flow. It includes attention-guided message passing, multi-modal data fusion, and graph-based temporal learning. The stages involve patch embedding, global filter layers, global average pooling, frequency domain features, learnable global filters, linear transformations, MLP, and Layer Norm. Images represent input data processing stages, highlighting graph-based and frequency analysis techniques.

Overview of the Dynamic Medical Graph Framework (DMGF). This framework integrates attention-guided message passing, graph-based temporal learning, and multi-modal data fusion to model complex interactions in healthcare data. Key components include a graph-based temporal learning module for capturing structural and temporal dependencies, an attention-guided mechanism to enhance interpretability, and a multi-modal fusion layer for integrating heterogeneous medical data. The framework is optimized with temporal consistency and contrastive learning to ensure robust and clinically meaningful predictions.