TAY_multimodal

fMRI Date Processed: 06-01-25 to 06-11-25

NODDI Date Processed: 06-09-25 to 07-12-25

Dataset:

• StudyInfo: combining/comparing DWI NODDI and rsfMRI metrics from TAY data, as well as DWI NODDI and age.
• Dataset: TAY CAMH Cohort Study, 54 participants that completed the [18F]SynVesT-1 PET substudy (aged 16-24)
  • Outputs were derived from TIGRLab/SCanD_Project v2.0.0
    • NODDI outputs used QSIPREP for preprocessing and AMICO NODDI model to derive NDI and ODI values
    • fMRI outputs used fMRIprep fit for preprocessing and used xcp_d outputs

Contact:

• Principal Investigator: Erin Dickie
• Primary: Anisha Miah - [email protected] / [email protected]
• Other: Christin Schifani

System:

• Dependencies:

    library(ggcorplot)
    library(ggseg)
    library(ggplot2)
    library(ggridges)
    library(ggsegGlasser)
    library(ppcor)
    library(tidyverse)
    library(dplyr)

• RanFrom: Kimel

Commands/Scripts:

• Scripts located in 01dwi_wrangling.Rmd
• Atlases used: HCP MMP (Glasser) and HCP subcortical (aseg)
• Paths:
  • fMRI outputs: path_xcpd = "/scratch/amiah/SCanD_project/data/local/derivatives/xcp_d/0.7.3/"
  • DWI NODDI outputs: path_noddi = "/projects/jwong/qsiprep/output/noddi_extract/ciftify_parcellations/"
  • Age at time of MRI: age_df = "/scratch/jwong/tay_pet_age.csv"
• Functions made:

  wrangle_data(path, pattern) #concatenates subject files into one dataframe
  calculate_GC(dataframe, string) #calculates postitive, absolute and mean global signal correlation and can be customized for individual or group dataframe based on string input
  matrix_QC(dataframe, run_num, title_string) #creates a matrix between subject to subject for global signal correlation between atlases
  

• First section (lines 39-220) involves concatenating all age and fMRI metrics. Also includes brainmaps of ten random subjects global signal correlation maps, alonside group brain maps
• lines 222-284: looks at quality controlling global signal correlation by looking at subject to subject matrices, as well as graphing timeseries of each ROI of some random number of patients
• lines 285-315: concatenating all NODDI outputs between subjects
• lines 318-344: creating full dataframes of all required metrics (rsfMRI metrics, NODDI metrics, age) of each atlas
• lines 364-561: group brainmaps of each metric. Some brainmaps use the matlab colours, while others use viridis_c to visualize
• lines 565-829: looks at age-NODDI correlations and maps it onto a brainmap as well as density plots
• lines 833-1320: looks at NODDI-rsfMRI correlations and maps it onto a brainmap as well as density plots
• lines 1322-1902: looks at NODDI-rsfMRI correlations with age regressed out and brainmaps it

Quality Control:

• Primary: Anisha Miah
  • Visual NODDI QC file with comments can be found in: NODDI_QC.csv
    • sub-CMH00000125 removed from NODDI outputs due to misalignment of NDI/ODDI map with parcellations
  • Visual fMRI QC completed on: http://srv-dashboard.camhres.ca/study/TAY/pipeline/anisha_2025/niviz-rater

Data Requests/Publications: N/A

GOALS DURING PROJECT

For each metric, we need to:

1. Read literature and discuss to understand the metric
2. build/clean up scripts ot extract the metric
3. help design QA process and QC the metric related pipelines.

Another concept is parcallations

• read lit to understand what they are [x]
• understand how they are represented in BIDS derivatives outputs [x]

Diffusion Metrics (NODDI - NDI, ODI, free water)

• reading list from Christin [x]
• extracted pipeline outputs - /archive/data/TAY/pipelines/versioned_outputs/pet_subsample/
• need to understand these outputs and concanate group outputs by atlas [x]
• qc files exists but QC process not set
• qc should go through qsiprep both ways:
• https://imaging-genetics.camh.ca/documentation/#/resources/Qsiprep-QC-guide
• add also inspect the .html pages in the qsiprep folder that describe the workflow

rsfMRI metrics

• Literature: History of rsfMRI - good overview to start [x]
• XCP: https://pmc.ncbi.nlm.nih.gov/articles/PMC10690221/ [x]
• understand fMRIPREP QC and rsfMRI QC metrics [x]
• example completed pipeline outputs /scratch/galinejad/ScanD/SHARED_FOLDER_25/

fALFF/ALFF + reHO (will discuss with Colin)

• literature: (https://pmc.ncbi.nlm.nih.gov/articles/PMC11245004/#S2) [x]
• extraction - all done by XCP-D (which is stage-3.sh of TIGR_BIDS)
• learn how to understand output to find fALFF outputs [x]
• learn how to concatenate across subjects from the output [x]
• also grab coverage files to incorporate into QC [x]
• understand/do visual qc of XCP and fmriprep to grab motion related QC metrics [x]
• Decision with Colin: are we going to work from the GSR XCP-outputs - and take the reHo and alff values [x] (We decided on XCP_D outputs)

Global Corr

• literture for it (prob from Anticivic lab) - https://elifesciences.org/articles/66968 [x]
• additonal script ontop for XCP needed to calcualte it
• to calculate - grad correlation matrix [x] (calculate_GC function does this)
• Total GC = mean of row
• Positive GC = mean of all positive values in the row
• Absolute GC = mean of absolute values in the row

Principle Gradient

• literature []

  • Margiles paper, https://pmc.ncbi.nlm.nih.gov/articles/PMC5098630/
  • Ju-Chi most recent paper: https://pubmed.ncbi.nlm.nih.gov/39260567/
  • More updated review: https://www.sciencedirect.com/science/article/pii/S1053811922001161?via%3Dihub

• additional calculation in python (using BrainSpace package) to calulate it - Ju-Chi has scripts []

• To get new templates (on a lab computer - done by erin):

• module load connectome_workbench; wb_command -cifti-parcellate /scratch/edickie/parcellate_gradients/hcp.embed.dscalar.nii /scratch/amiah/SCanD_project/data/local/derivatives/xcp_d/0.7.3/atlases/atlas-Glasser/atlas-Glasser_space-fsLR_den-91k_dseg.dlabel.nii COLUMN /scratch/edickie/parcellate_gradients/atlas-Glasser_space-fsLR_den-91k_desc-Margulies2016_gradients.pscalar.nii"

• getting a jupyter env on scinet (in the scinet terminal)

module load python/
cd $SCRATCH
virtualenv --system-site-packages brainspace_2025
source brainspace_2025/bin/activate
pip install brainspace
venv2jup   

(then open up jupyter on demand and you should see "brainspace_2025" as an option

notes on gradients from meeting with Ju-Chi

• Ju-Chi's gradient calculating code from paper: /scratch/jcyu/spin_gradients/code/3.3_spins_gradiant.py

  • inside this script - the calc_aligned_gradient() function is the most important

• need to re-parcellate from margiles file to get templates: /scratch/jcyu/spin_gradients/hcp.embed.dscalar.nii

  July goals

  • Compare the NODDI to the rsfMRI metrics using stats [x]
  • pulling all the metrics into one dataframe?[x]
  • correlate across the brain within subject [x]
  • correlate across subject wihtin parcels [x]
  • make pretty ggseg pictures [x]