Presentation Information
[2F4-OS-33-03]Analysis of Onset Patterns of BPSD in Patients with Dementia Using Non-Negative Tensor Factorization
〇Ayaka Yamanaka1, Mana Sasagawa1, Masahiro Kohjima1, Hyuta Onuma2, Kota Kurihara2, Shunichi Seko1, Yasuhiro Minami2 (1. NTT Human Informatics Laboratories, 2. The Univ. of Electro-Communications)
[[online]]
Keywords:
Dementia,BPSD,Time variation
Behavioral and Psychological Symptoms of Dementia (BPSD) comprise a broad spectrum of symptoms that significantly impair the quality of life of individuals with dementia and impose a substantial burden on caregivers. Identifying patterns of BPSD onset may facilitate proactive interventions and the development of appropriate care strategies, thereby contributing to the reduction of caregiver burden.
In this study, we analyze BPSD onset frequency data using non-negative tensor factorization. The data are represented as a three-way tensor with dimensions corresponding to symptoms, individuals, and time periods. A Tucker-1 decomposition model is employed to extract time-independent clusters of BPSD symptoms while simultaneously capturing temporal changes in individuals’ cluster affiliations.
Using onset frequency data collected from 34 individuals over 2 years, we identified five distinct BPSD onset patterns. Furthermore, the results revealed the coexistence of individuals whose cluster memberships remained stable over time and those whose memberships transitioned or fluctuated across periods.
In this study, we analyze BPSD onset frequency data using non-negative tensor factorization. The data are represented as a three-way tensor with dimensions corresponding to symptoms, individuals, and time periods. A Tucker-1 decomposition model is employed to extract time-independent clusters of BPSD symptoms while simultaneously capturing temporal changes in individuals’ cluster affiliations.
Using onset frequency data collected from 34 individuals over 2 years, we identified five distinct BPSD onset patterns. Furthermore, the results revealed the coexistence of individuals whose cluster memberships remained stable over time and those whose memberships transitioned or fluctuated across periods.