Enabling Consumer Centric Digital Twins in Healthcare 5.0: A Medical Image Fusion Framework With Edge Guidance

Abstract

In the era of healthcare 5.0, consumer electronics like wearables and remote patient monitoring devices generate vast streams of health data, yet this data lacks the deep anatomical context needed for a truly high-fidelity digital twin. Medical image fusion (MIF) plays a pivotal role in creating this foundational anatomical blueprint to bridge the context gap. However, prevailing methods face challenges in retaining fine anatomical details, precisely delineating functional boundaries, and effectively integrating information from diverse imaging modalities. These difficulties arise due to deep networks often diminishing critical edge details and insufficient synergy between spatial and frequency domains. To address these issues, we propose DECFusion, a Dual-Branch Edge Guidance Cross-domain Fusion Network. Our method introduces a dual-branch architecture that enables hierarchical interaction between deep semantic and edge features, ensuring consistent reinforcement of structural details. We designed the global perceive injection module to guide the transmission of shallow structural features along anatomical edges, mitigating the progressive attenuation of high-frequency diagnostic details. Furthermore, the cross-domain fusion module combines a two-round four-directional spatial awareness with a frequency feature fusion component, jointly modeling global spatial dependencies and multi-frequency representations. Extensive experiments on public medical datasets demonstrate that DECFusion achieves state-of-the-art performance. This work establishes DECFusion as a critical enabling technology, providing the essential high-fidelity data needed to unlock the full potential of consumer health electronics.