Progress Report

Integration of Asian humanities and brain informatics to enhance peace and compassion of the mind1. Data-driven modeling

Progress until FY2022

1. Outline of the project

We began to work on constructing a "high-low mix database" to construct a data-driven model for personality subtyping. Specifically, we selected the items of the questionnaire and of measuring daily behavior using smartphones that constitute the "low" portion and began preliminary research. The contents of the brain image database, which forms the "high" part, were formulated.

2. Outcome so far

Theme 1: (1) Based on successfully identifying subtypes of healthy subjects using various psychiatric symptoms in collaboration with Theme 3, nine clinical evaluation scales were selected for items in a large-scale Internet/smartphone survey. Questionnaires on comfort and vitality were developed in collaboration with Theme 2 and the social implementation team to start data collection in the next fiscal year. We confirmed that the data structure was suitable for building the subtyping model. In collaboration with Theme 2, we conducted a preliminary survey after receiving approval from an ethics committee.

Figure 1: High-low mix database
Figure 1: High-low mix database

(2) In collaboration with Theme 3, we adopted the protocol for multicenter disease research of the Human Connectome Project, an international standard for developing brain data protocols. We have completed the construction of an environment that enables us to conduct the preprocessing pipeline for neuroimaging data.

Table 1:Questionnaire measurements
Table 1:Questionnaire measurements

There 2: A preliminary survey was conducted in preparation for a large-scale survey to construct high-low mix data in the next fiscal year and beyond. We confirmed that the created database was suitable for developing the subtyping model. In consultation with other research projects, we designed the experimental design, including the implementation period, subject segments, questionnaire items, and types of daily behavioral data (experience sampling and behavioral data) to be obtained. A data collection application was created, and a preliminary survey was conducted after receiving approval from an ethics committee. In the first stage, a screening questionnaire was administered to 50,000 participants, and 2,000 participants were selected based on their responses. In addition, approximately 100 respondents were selected among those who responded to the questionnaire, and a behavior measurement survey was conducted.

Table 2:Behavioral measurements
Table 2:Behavioral measurements

There 3: The contents of the "high" neuroimaging database were examined. In neuroimaging data collection, the type of imaging data and scanning parameters are crucial. We examined these points regarding the quality, scanning time, and other factors of open data. In addition, various preprocessing methods have been proposed day by day. We collected information on these methods and selected appropriate ones. We also determined the database structure in terms of data storage method, preprocessing method, operation method, and model construction. The type of neuroimaging data and the scanning parameters were determined to be compatible with the plan for the subtyping model constructed in Theme 1. We confirm that the dataset will contain sufficient data to build the subtyping model using neuroimaging data by the method for stratifying diseases (Hierarchical supervised/unsupervised learning).

3. Future plans

Begin collecting data for the "high" and "low" categories. The "low" data, questionnaires about personal characteristics, will be conducted online. Collect smartphone logs, daily behavior data, and mood sampling through smartphones and wearable devices over several months. Neuroimaging data, the "high" data, will also be collected to build a high-low mix database, and these data will also be used for subtyping.

(TANAKA Saori:ATR, NAKAMURA Hajime:KDDI Research, SAKAIYuki:XNef))