Progress Report
Last updated:
Integration of Asian humanities and brain informatics to enhance peace and compassion of the mind[1] Data-driven modeling
Progress until FY2025
１． Outline of the project
We expanded the High-Low Mixed Database to construct a data-driven model of personality subtypes related to peace of mind and vitality. We collected and preprocessed questionnaire, smartphone/EMA, MRI, cognitive, and decision-making data, and developed an algorithm for applying a pretrained clustering model to new samples.
２． Outcome so far
Item 1: Data-driven modeling
- (1) Construction and curation of the High-Low Mixed Database
- With Research Task 2 and the Social Implementation Team, the second-phase survey collected 17,413 cumulative online questionnaire samples and behavioral monitoring data from 781 participants. After curation, 10,834 valid samples and 520 intervention-study samples were integrated for external validation.
- (2) Data-driven construction of a model for classifying personality subtypes related to peace of mind and vitality
- For "low" data, we implemented a variational-posterior scoring algorithm for a pretrained clustering model with Gaussian and categorical features. Applying it to 10,834 new samples and 520 Social Implementation samples reproduced the View 1-3 subtype distributions, supporting generalizability.
Subtype interpretation identified an integrated positive subtype with high interdependent happiness and compassion, self-critical/socially anxious or depressive subtypes with lower self-compassion, and a View 1 subtype with lower FFMQ acting with awareness, suggesting weaker attention to ongoing actions under stronger intrusive thoughts and impulsivity (Figures 1-2).
For "high" MRI data, structural, diffusion-weighted, resting-state fMRI, and field-map data were collected from 141 cumulative participants using the HCP-CRHD protocol and stored in XNAT. Preprocessing used an HCP-style pipeline with 36p + spike regression denoising.
For preprocessed MRI data, hierarchical supervised/unsupervised learning using 160 functional connectivity features identified a stable, non-extreme optimal model. Frontoparietal and somatosensory networks contributed strongly to subtype identification.
Item 2: Large-scale survey using the Internet and smartphones
We continued a large-scale survey combining online questionnaires, smartphone logs, actual behavior, and ecological momentary assessment. The dataset was designed to evaluate peace of mind, vitality, interdependent happiness, outward compassion, self-compassion, psychiatric symptoms, and daily behavior in an integrated manner (Figure 1).
３． Future plans
We will complete curation of the full second-phase survey dataset and identify stable personality subtypes by integrating low-burden and high-burden data. For FY2025 MRI data, clustering will proceed as preprocessing is completed, and relationships with peace of mind, vitality, and compassion will be examined. We ultimately aim to identify five or more stable personality types for intervention and social implementation.