SPACE-fmri-change Analysis Pipeline

Analysis code for the SPACE (Supportive Parenting for Anxious Childhood Emotions) fMRI change study: neuroimaging and clinical-outcome analyses of an emotion face processing task collected pre- and post-treatment, under parent-accompanied and alone scan conditions, across two randomized-controlled-trial phases (Phase I: SPACE vs. PE; Phase II: SPACE vs. CBT).This repo contains the complete analysis pipeline underlying the reported neuroimaging and clinical results.

Pipeline Stages

Script	Stage	Produces
01_aggregate_data.sh	Data aggregation (HPC)	Smoothed, condition-contrasted first-level activation volumes
02_extract_roi_activation.py	ROI extraction (HPC)	Per-subject Shen 368-parcel median activation (long format)
03_create_clinical_brain_volumes.py	Covariate volumes (HPC)	Per-subject symptom-score brain volumes for downstream masking
04_build_final_datasets.py	Dataset construction	5 FINAL analysis datasets (Phase I/II x ITT/PP, cross-phase)
05_neuroimaging_lme.R	Primary neuroimaging analysis	Table 3, Figure 2 — hierarchical TIME x TX x CONDITION LMEs per ROI
06_clinical_outcomes_lme.R	Primary clinical analysis	Tables 1-2, Figure 1 — TIME x TX LMEs for PARS/FASA-P/FASA-C
07_clinical_neuro_subsamples.R	Supplementary clinical analysis	Clinical LMEs restricted to the neuroimaging subsamples
08_scanner_sensitivity.R	Sensitivity analysis	Leave-two-out refit excluding subjects scanned on a different scanner
09_phase_sensitivity.R	Sensitivity analysis	Permutation test distinguishing power loss from Phase I-specific effects
10_qc_selection_analysis.py	Selection-bias analysis	QC-included vs. QC-excluded participant comparison
11_plot_amygdala_3way.R	Figure generation	Phase II amygdala TIME x TX x CONDITION interaction plot

Shared logic used by more than one numbered script is factored into four helper files, each sourced by the scripts that need it:

Helper	Used by
data_loading_helpers.R	05, 08, 09, 11
neuro_lme_helpers.R	05, 08
neuro_plot_helpers.R	05, 11
clinical_lme_helpers.R	06, 07

Dependencies

R (developed against R 4.3.1): lme4, lmerTest, ggplot2, dplyr, ggExtra, lavaan, grid, patchwork, stringr, ggpubr, tidyr, foreach, doParallel.

Python: numpy, pandas, scipy, joblib, nibabel, pyreadstat.

External tools: AFNI (3dmaskave, 3dmerge, 3dcalc) for stages 1-2; an HPC environment with an environment-module system (module load AFNI) for stages 1-3, which are designed to run on a compute cluster rather than locally.

Usage

Run the numbered scripts in order; each stage's output feeds the next. Stages 1-3 require raw imaging data and AFNI/HPC access and are not reproducible outside that environment; stages 4 onward operate on the tabular outputs of stages 1-3 and can be run on any machine with the Python/R dependencies above installed.

bash 01_aggregate_data.sh              # HPC only
python 02_extract_roi_activation.py    # HPC only
python 03_create_clinical_brain_volumes.py   # HPC only
python 04_build_final_datasets.py
Rscript 05_neuroimaging_lme.R
Rscript 06_clinical_outcomes_lme.R
Rscript 07_clinical_neuro_subsamples.R
Rscript 08_scanner_sensitivity.R
Rscript 09_phase_sensitivity.R
python 10_qc_selection_analysis.py
Rscript 11_plot_amygdala_3way.R

Scripts 05-11 use control flags (near the top of each file) to toggle optional branches (e.g. exploratory post-hoc reporting, SEM, residualization); all flags default to the setting used for the reported manuscript results.

Note on file paths: scripts in this directory read their base directories from environment variables (e.g. SPACE_BASE, SPACE_DATA_DIR, SPACE_OUT_DIR) with placeholder defaults. Set these variables to point to the location of the corresponding input files before running the scripts in your environment.