Analysis files and meta-data to accompany the manuscript "Modeling shared and specific variances of irritability, inattention, and hyperactivity yields novel insights into white matter perturbations" published in the Journal of the American Academy of Child and Adolescent Psychiatry1.
This repository contains analysis scripts (MPlus, bash, Matlab, R) used for preprocessing and analysis of the data found in our OpenNeuro dataset. Although the code used to deface the anatomical scans (T1, T2) is not included in this repository, we de-identified these images using using v1.0.0 of the DSST Defacing Pipeline.
These files perform exploratory factor analysis (EFA), confirmatory factor analysis (CFA), bifactor model analysis, and cross-correlations in MPlus. Scripts should be run in order from Script1 to Script4.
Note: These MPlus input files assume that missing data are coded as 999, but the phenotype.tsv file as shared on OpenNeuro contains n/a to indicate missingness.
Performs EFA using items from the Conners (3rd edition) Comprehensive Behavioral Rating Scale (CBRS) for parents2.
Performs CFA using the top six highest loading CBRS items from the three-factor EFA (Script1) for the irritability, inattention, and hyperactivity factors. This is done separately for the Phenotyping and DTI subsamples.
Performs bifactor model analysis using the top six highest loading CBRS items from the three-factor EFA (Script1) for the irritability, inattention, and hyperactivity factors. This is done separately for the Phenotyping and DTI subsamples.
Performs cross-validation of the latent variables from the bifactor model (by applying the loadings from the Phenotyping sample to the DTI sample and vice versa) and determining the correlations between the latent factor scores.
The following scripts assume the following directory structure:
data
├── BIDS
│ ├── sub-[participant-id]
│ └── ses-1
│ ├── anat
│ └── dwi
├── code
├── lists
├── PROC
│ ├── sub-[participant-id]
│ └── ses-1
│ ├── anat
│ └── dwi
├── slurm
└── TBSS_Tortoise
│ ├── AD
│ ├── FA
│ ├── origdata
│ ├── PALM
│ ├── RD
│ └── stats
Bash scripts that perform basic DTI preprocessing using common neuroimaging tools: Analysis of Functional NeuroImages (AFNI)3, Tolerably Obsessive registration and Tensor Optimization Indolent Software Ensemble (TORTOISE)4 5, and Tract-Based Spatial Statistics (TBSS) in FSL6. Scripts should be run in order from 01.Preproc to 06.TBSS.Proc.nonFA.
Create directories in the PROC directory where all outputs will be saved. Perform T2 skull stripping (using FSL's bet function). Axialize T2 images (using AFNI's fat_proc_axialize_anat). Set origin for DWI images to 0,0,0.
Loop through participants with bad DWI volumes and produce a file called dwi_sel_AP with the volumes to keep (using the AFNI's fat_proc_select_vols).
Loop through participants again. Convert DWIs to create new matrices including bvals and bvecs (using AFNI's fat_proc_convert_dcm_dwis). Filter out bad volumes from vols and b-matrices (using AFNI's fat_proc_filter_dwis). Save gradients and bvalues as row matrixes (using AFNI's 1dDW_Grad_o_Mat++).
Loop through participants with no bad DWI volumes and produce a file called dwi_sel_AP.
Loop through participants with bad DWI volumes. Unzip T2w NifTI files. Add DIFFPREP to swarm file. Run swarm.
Loop through participants with no bad DWI volumes. Unzip T2w NifTI files. Add DIFFPREP to swarm file. Run swarm.
Loop through participants. Extract a single volume out of a 4D DWI image based on desired bvalue (using TORTOISE's ExtractBValVolume). Perform skull stripping on AP_DMC_bval_0_1.nii (using FSL's bet function). Unzip b0_brain_mask.nii.gz. Add EstimateTensorNLLSRESTORE and ComputeAllTensorMaps.bash from TORTOISE to swarm file to perform tensor fitting. Run swarm.
Create TBSS directory if it does not already exist; populate with fractional anisotropy (FA) maps. Run basic TBSS preprocessing scripts: tbss_1_preproc_NA, tbss_2_reg, tbss_3_postreg, and tbss_4_prestats. Create PALM directory and copy skeleton and tracts into it. Unzip necessary .nii.gz files.
Create create directories for the non-FA images of every participant: axial diffusivity (AD) and radial diffusivity (RD). Create RD maps using average of AP_DMC_R1_DT_EV0001.nii.gz and AP_DMC_R1_DT_EV0002.nii.gz. Copy and organize AD and RD maps. Run tbss_non_FA for AD and RD maps.
The file DTI_Directions.xlsx contains information about the directions removed from the DTI data, which was used during the DTI preprocessing to create the lists Bad.DWIs.txt and No.Bad.DWIs.txt. The columns in this spreadsheet are as follows:
-
participant_id: participant identifier -
B0 to Remove: index of B0 images to remove -
B300 to Remove: index of B300 images to remove -
B1000 to Remove: index of B1000 images to remove -
Extra B0: index of extra acquired B0 images -
Extra B300: index of extra acquired B300 images -
Extra B1000: index of extra acquired B1000 images -
N B0: total number of B0 images (10 -B0 to Remove+Extra B0) -
N B300: total number of B300 images (10 -B300 to Remove+Extra B300) -
N B1000: total number of B0 images (60 -B1000 to Remove+Extra B1000)
Analyses of the DTI data were conducted with Permutation Analysis of Linear Models (PALM)7, using preprocessed TBSS data as inputs. Prior to running these analyses, the mask of the FA skeleton (mean_FA_skeleton_mask.nii) must be visually inspected and modified, if neccessary, to remove any voxels containing skull/dura, gray matter, etc.
RUN_PALM.m is a Matlab script that uses PALM to run analyses on the preprocessed TBSS data. Commands are shown for analyses, all controlling for age and mean FA of the white matter skeleton:
- Latent factors from bifactor model with FA
- Shared factor clusters as masks with FA, with either natal sex or psychotropic medication load as additional covariates
- Shared factor clusters as masks with AD
- Shared factor clusters as masks with RD
- Latent factors from CFA with FA
- Questionnaire scores (Affective Reactivity Index (ARI)8, CBRS) with FA
- Diagnostic group differences in FA
The file EB.csv contains the exchangability blocks for PALM9, which were used in the analysis for diagnostic group differences in FA, although the column headers must be removed before using with PALM. The columns in this spreadsheet are as follows:
participant_id: participant identifier, which must be removed before running analyses in PALMconstant: column containing -1 for all participantshas_rel: 1 = no siblings in dataset; 2 = has sibling in datasetpair_id: 1 = no siblings in dataset; 2+ = unique identifier for each sibling pairsub_id: ascending numbers unique to each participant
There are 11 pairs of related participants:
- 7 HV-HV pairs (
pair_id= 2, 7, 8, 9, 10, 11, 12) - 1 DMDD-HV pair (
pair_id= 3) - 1 DMDD-ADHD pair (
pair_id= 4) - 1 DMDD+ADHD-ADHD pair (
pair_id= 5) - 1 ADHD-HV pair (
pair_id= 6)
GENERATE_FIGS.R is an R script that load relevant packages and uses data from phenotype.tsv and the results of the bifactor model for the Phenotyping subsample to create:
- Figure 2: validation of latent factors
- Figure 3: validation of latent factors with item-level data
- Figure S10: association of latent factors with impairment
- Figure S11: diagnostic group differences in latent factors
- Figure S12: association of latent factors with age
- Figure S13: sex differences in latent factors
Footnotes
-
McKay, C.C., Scheinberg, B., Xu, E.P., Kircanski, K., Pine, D.S., Brotman, M.A., Leibenluft, E., & Linke, J.O. (2024). Modeling shared and specific variances of irritability, inattention, and hyperactivity yields novel insights into white matter perturbations. Journal of the American Academy of Child and Adolescent Psychiatry. doi: 10.1016/j.jaac.2024.02.010 ↩
-
Conners, C. K., Pitkanen, J., & Rzepa, S. R. (2011). Conners 3rd Edition (Conners 3; Conners 2008). In J. S. Kreutzer, J. DeLuca, & B. Caplan (Eds.), Encyclopedia of Clinical Neuropsychology (pp. 675–678). Springer New York. ↩
-
Taylor, P. A., & Saad, Z. S. (2013). FATCAT: (an efficient) Functional and Tractographic Connectivity Analysis Toolbox. Brain connectivity, 3(5), 523–535. doi: 10.1089/brain.2013.0154 ↩
-
Pierpaoli, C., Walker, L., Irfanoglu, M. O., Barnett, A., Basser, P., Chang, L-C., Koay, C., Pajevic, S., Rohde, G., Sarlls, J., Wu, M. (2010). TORTOISE: an integrated software package for processing of diffusion MRI data. ISMRM 18th annual meeting, Stockholm, Sweden, abstract #1597 ↩
-
Irfanoglu, M. O., Nayak, A., Jenkins, J., Pierpaoli, C. TORTOISEv3:Improvements and New Features of the NIH Diffusion MRI Processing Pipeline. ISMRM 25th annual meeting, Honolulu, HI, abstract #3540 ↩
-
Smith, S. M., Jenkinson, M., Johansen-Berg, H., Rueckert, D., Nichols, T. E., Mackay, C. E., Watkins, K. E., Ciccarelli, O., Cader, M. Z., Matthews, P. M., & Behrens, T. E. (2006). Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data. NeuroImage, 31(4), 1487–1505. doi: 10.1016/j.neuroimage.2006.02.024 ↩
-
Winkler, A. M., Ridgway, G. R., Webster, M. A., Smith, S. M., & Nichols, T. E. (2014). Permutation inference for the general linear model. NeuroImage, 92(100), 381–397. doi: 10.1016/j.neuroimage.2014.01.060 ↩
-
Stringaris, A., Goodman, R., Ferdinando, S., Razdan, V., Muhrer, E., Leibenluft, E., & Brotman, M. A. (2012). The Affective Reactivity Index: a concise irritability scale for clinical and research settings. Journal of child psychology and psychiatry, and allied disciplines, 53(11), 1109–1117. doi: 10.1111/j.1469-7610.2012.02561.x ↩
-
Winkler, A. M., Webster, M. A., Vidaurre, D., Nichols, T. E., & Smith, S. M. (2015). Multi-level block permutation. NeuroImage, 123, 253–268. doi: 10.1016/j.neuroimage.2015.05.092 ↩