Clinical Research Directory

Applying Comfort Theory to an AI-Assisted Group Music Intervention: Psychosocial and Physiological Outcomes and Feasibility in Long-Term Care Residents.

The research design adopts a sequential two-phase framework. The study will be conducted in three residential LTC facilities located in Taichung City or Changhua County, with a total planned enrollment of 60 residents. In Phase 1, 20 residents from one facility will participate in a six-week group music intervention, with one session per week. This phase will focus on multimodal image data construction and AI model development. Facial expression data will be manually annotated to establish reference standards for emotion recognition, while upper-limb movements will be analyzed using pose estimation and skeletal recognition techniques to extract dynamic features. These data will be used to develop and preliminarily validate an AI-based emotion and activity response monitoring model. In Phase 2, 40 residents who did not participate in Phase 1 will be recruited and divided into four groups to participate in a six-week group music intervention, with one session per week. The AI model developed in Phase 1 will be implemented in the LTC setting to transform residents' facial emotional expressions and upper-limb activity responses into low-interference visual information for the music therapist as an adjunct to activity facilitation and professional judgment. Evaluation indicators include subjective psychosocial measures of loneliness, anxiety, and depression, with heart rate variability (HRV) assessed in Phase 2 as an objective physiological measure. This study expects to establish a multimodal dataset and an AI-based emotion and activity response monitoring model for group music activities in LTC settings and to preliminarily evaluate the feasibility of a human-AI collaborative approach characterized by "AI-assisted monitoring and human professional decision-making.