Page 341 - Invited Paper Session (IPS) - Volume 2

IPS273 Tomoki Tokuda et al.

rest while fMRI images are scanned), clinical questionnaires, and gene
expression data.
Data: The number of subjects is 134, comprising of 67 MDD and 67 healthy
subjects. Our data comprise several modalities, such as FC data in resting state
fMRI, clinical questionnaires, and biological data, which are mixed up with
2800 numerical, 100 categorical, and 2 integer features. For simultaneous
analysis of all datasets, we concatenated them into a single data matrix
(subjects are common to the sub-datasets). For FC data, functional
connectivity between regions of interest (ROIs) in the brain were evaluated,
which is defined as correlation coefficient of average time series of Blood
Oxygen Level Dependent (BOLD) fMRI signals between ROIs. Since ROIs
consist of 78 brain regions, this resulted in 2701 features (78 × 77/2) for FC.
Clinical questionnaire data comprise several types of scores that measure
depression severity (e.g., Beck Depression Inventory, BDI), personality (e.g.,
NEO personality Inventory), and life experiences such as Child Abuse Trauma
Scale (CATS). In particular, BDI and Hamilton Rating Scale of Depression
(HRSD) are important indicators for measuring the severity of depression
focusing on its several symptoms. Hence, we also included question items
about these indicators as a categorical feature. The same clinical
questionnaires were administered to both depressive and control subjects.
Importantly, for depressive subjects, these were administered before the onset
of Selective Serotonin-Reuptake Inhibitor (SSRI) treatment. Moreover, some
indices, such as BDI and HRSD, were again administered six weeks and six
months after the treatment (these features are denoted with endings of 6w
and 6m). We expected that inclusion of repeated measurements in these
scores would provide useful information on treatment effect. Beside functional
connectivity factors, we included several biological elements: sex, age, Brain
Derived Neurotrophic Factor (BDNF), Cortisol, Single Nucleotide
Polymorphisms (SNPs), and DNA Methylation.

328 | I S I W S C 2 0 1 9