Lumina: A Multimodal DT Framework for Context-Aware Wellbeing Monitoring and Personalized Intervention

Abstract

The increasing availability of personal health and behavioral data has created new opportunities for digital wellbeing monitoring. However, existing systems remain fragmented, often focusing on isolated dimensions and lacking a unified and dynamic representation of the individual. In this paper, we present "Learning, Unified Modeling for Integrative Next-generation Analytics" (LUMINA), a multimodal and context-aware Digital Twin (DT) framework for personalized wellbeing modeling. The proposed approach represents the individual as a continuously evolving multidimensional system, integrating structured behavioral data with unstructured textual inputs within a unified computational model. LUMINA combines an omic-based representation of wellbeing with multimodal data fusion and adaptive interaction mechanisms, enabling the joint modeling of objective and subjective aspects of human experience. The framework incorporates AI-driven components for sleep-related prediction and emotional analysis as auxiliary signals for DT updating. A preliminary deployment involving 27 participants provides initial evidence of feasibility, showing variability across wellbeing profiles, including 26% of users in the critical PsychoOmics range. Sleep-prediction performance was evaluated separately on an external behavioral sleep dataset, and should not be interpreted as in-context validation on the deployment cohort. These findings support LUMINA as an exploratory foundation for personalized and context-aware wellbeing monitoring.