Cross-Setting Rehabilitation Data Integration: An Edge AI and Knowledge Graph Framework

Abstract

Rehabilitation services are delivered across home, school, and clinical settings, creating fundamental barriers to achieving personalized, continuous care. Data generated from wearable devices, electronic health records, and clinical assessments are heterogeneous and siloed, limiting the potential for AI-driven insights. This study proposes a four-layer framework combining edge AI and knowledge graphs to enable the integration of rehabilitation data across these diverse settings. Drawing on recent advances in rehabilitation knowledge discovery, data standardization, and knowledge-graph-enhanced personalization, the framework addresses three core challenges: the semantic unification of multi-source physiological and behavioral data, context-aware reasoning for personalized intervention design, and clinical decision support with data provenance. The framework encompasses processes such as wearable data acquisition, provenance-enabled knowledge graph integration, and role-based clinical review. This conceptual architecture lays the foundation for future empirical validation across various rehabilitation populations and care scenarios.