Logo


Project Information and Reproducibility Guide

View the Project on GitHub PennLINC/spatiotemp_dev_plasticity



Spontaneous Activity Development Unfolds Hierarchically Along the Sensorimotor-Association Axis

Spontaneous activity critically refines cortical circuits in the developing brain. In the rodent brain, spontaneous activity in sensory regions evolves from strong and synchronized to sparse and decorrelated as the cortex transitions from plastic to mature. Synchronized bursts in intrinsic activity are therefore considered a functional hallmark of plastic cortices. Here, we leverage this functional hallmark to characterize the developmental unfolding of activity-indexed plasticity in the human brain during youth. We hypothesized that age-related changes in spontaneous activity would be spatially and temporally governed by the sensorimotor-association (S-A) axis of cortical organization and provide evidence for hierarchical neurodevelopment in childhood and adolescence. Elucidating spatially-localized windows of enhanced and diminished cortical plasticity will help to identify windows wherein youth educational and psychiatric interventions may be maximally beneficial.

Project Lead

Valerie J. Sydnor

Faculty Lead

Theodore D. Satterthwaite

Analytic Replicator

Bart Larsen

Collaborators

Bart Larsen, Azeez Adebimpe, Maxwell A. Bertolero, Matthew Cieslak, Sydney Covitz, Yong Fan, Raquel E. Gur, Ruben C. Gur, David R. Roalf, Russell T. Shinohara, Dani S. Bassett, Theodore D. Satterthwaite

Project Start Date

June 2021

Current Project Status

Manuscript in preparation

Datasets

RBC PNC-Health Exclude (primary) and LTN (sensitivity)

Github Repository

https://github.com/PennLINC/spatiotemp_dev_plasticity

Conference Presentations

Cubic Project Directory

/cbica/projects/spatiotemp_dev_plasticity

CBF/: parcel-wise cerebral blood flow maps for each participant, generated with ASLPrep  
code/: directory with the spatiotemp_dev_plasticity github repo  
FluctuationAmplitude/PNC/: vertex-wise and parcel-wise fluctuation amplitude maps for each participant, generated with xcp-d and connectome workbench  
FluctuationAmplitude/GAMRESULTS/: gam model outputs (effect sizes, p-values, fitted values, smooth estimates, smooth characteristics, derivatives)  
Maps/: surface parcellation files and SNR masks (Maps/parcellations/) and S-A axis github repo (Maps/S-A_ArchetypalAxis/)
Myelin/: myelin development maps including the age effect size map (r2), the age of maximal growth map (age of max slope), and the annualized rate of change map (annualized roc) from Baum et al., 2021  
sample_info/: sample demographics, factor scores, rbcid-bblid key, and final project participant list (PNC_FinalSample_N1033.csv)
software/: project software   
Structural/: freesurfer output for each participant  
Timeseries/: vertex-wise and parcel-wise fully processed resting fMRI timeseries data for each participant, generated with fmriprep and xcp-d  



CODE DOCUMENTATION

The entire analytic workflow implemented in this project is described in the following sections and links to the corresponding github code are provided. This workflow includes quantification of regional fluctuation amplitude, PNC sample selection, fitting of generalized additive models (GAMs), and characterization of relationships between fluctuation amplitude, age, environmental variability, and the sensorimotor-association axis. Scripts were implemented in the order outlined below.

Fluctuation Amplitude Quantification

Resting state functional MRI data were processed with fmriprep 20.2.3 and xcp-d 0.0.4 to quantify fluctuation amplitude at each vertex on the fslr 32k cortical surface.

fmriprep 20.2.3 was run with the following parameters:

$ singularity run pennlinc-containers/.datalad/environments/fmriprep-20-2-3/image inputs/data prep participant --output-spaces MNI152NLin6Asym:res-2 --participant-label "$subid" --force-bbr --cifti-output 91k -v -v

xcp-d 0.0.4 was run with the following parameters:

$ singularity run pennlinc-containers/.datalad/environments/xcp-abcd-0-0-4/image inputs/data/fmriprep xcp participant --despike --lower-bpf 0.01 --upper-bpf 0.08 --participant_label $subid -p 36P -f 10 –cifti

Vertex-wise fluctuation amplitude maps were then parcellated with /fluctuation_amplitude/parcellate.Rmd to quantify mean fluctuation amplitude in each cortical region. Regions were defined with the HCP multimodal atlas (i.e. Glasser 360, primary analyses) and with the Schaefer 400 atlas (sensitivity analysis).

Fluctuation amplitude analyses were only conducted in brain regions that reliably exhibited high signal to noise ratio (SNR) in PNC functional MRI data. The vertex level SNR map generated in Cui et al., 2020, Neuron was parcellated with Glasser 360 and Schaefer 400 atlases with the script /fluctuation_amplitude/SNR_mask.Rmd for use in study analyses. Regions wherein >= 25% of vertices had attenuated signal were excluded from analyses.

Sample Construction

The final study sample was constructed with /sample_construction/finalsample.Rmd. The final sample was generated from the 1374 PNC participants with dominant group ses-PNC1_task-rest_acq-singleband scans (non-variant CuBIDS acquisitions). The following exclusions were applied to generate the final sample of N = 1033 participants:

Health exclude: 120 participants with medical problems that could impact brain function or incidentally-encountered brain structure abnormalities were excluded from the sample
T1 QA exclude: 23 participants with T1-weighted scans that failed visual QC were excluded from the sample
fMRI motion exclude: 179 participants with a mean relative RMS motion value > 0.2 during the resting state fMRI scan were excluded from the sample
Fluctuation amplitude outlier exclude: from the remaining 1052 participants, 19 individuals that had outlier (+- 4 SD from the mean) fluctuation amplitude data in more than 5% of Glasser 360 parcels were excluded from the sample

Model Fitting

GAM Functions
To characterize age-dependent changes in spontaneous activity fluctuations across the developing cortex as well as associations between fluctuation amplitude and environmental factors, generalized additive models were fit in each cortical region. GAM models and associated statistics, fitted values, smooths, and derivatives were quantified with the set of functions included in /gam_models/GAM_functions.R. This script includes the following functions:

Model Fitting: Age-Dependent Changes in Regional Fluctuation Amplitude
Developmental models were fit with gam_models/fitGAMs_fluctuationamplitude_age.R, which calls the functions described above. Age-focused GAMs were implemented for all regions included in the Glasser 360 and Schaefer 400 atlases, using the final study sample of N = 1033 generated during the sample construction process.

Model Fitting: Developmental Environment-Dependent Variation in Regional Fluctuation Amplitude
Models examining associations between fluctuation amplitude and environmental and family characteristics were run via gam_models/fitGAMs_fluctuationamplitude_environment.R on the N = 1033 study sample.

Data Interpretation and Visualization

Model results were examined and studied within our hierarchical neurodevelopmental plasticity framework in the markdown file /developmental_effects/hierarchical_development.Rmd. This markdown executes the following:

A rendered html of hierarchical_development.Rmd can be viewed here!

Sensitivity Analyses

The robustness of our developmental findings was confirmed in a series of sensitivity analyses. Sensitiviy analysis GAMs were fit with sensitivity_analyses/fitGAMs_sensitivityanalyses.R and results were examined in sensitivity_analyses/sensitivityresults.Rmd. The following sensitivity analyses were performed: