dvars

Repo for calculating DVARS in data stored on JHPCE

Structure

Files are stored as arrow datasets.

$ tree -L 5 dvars | head
dvars
├── dataset=HCPAgingRec
│   ├── sub=6002236
│   │   └── ses=1
│   │       ├── src=rfMRI_REST1_AP_Atlas_MSMAll_hp0_clean
│   │       │   └── dvars.arrow
│   │       ├── src=rfMRI_REST1_PA_Atlas_MSMAll_hp0_clean
│   │       │   └── dvars.arrow
│   │       ├── src=rfMRI_REST2_AP_Atlas_MSMAll_hp0_clean
│   │       │   └── dvars.arrow

Each file is a table

t	A	D	S	DPD	ZD	DVARS	dataset	sub	ses	src
1	3.784818	1.985132	0.2807895	-6.1827153	-4.6384849	2.817895	ukb	5902812	2	20227-filtered_func_data_clean
2	4.727538	2.182186	0.5185044	-2.7343983	2.0862912	2.954445	ukb	5902812	2	20227-filtered_func_data_clean
3	4.715485	2.297970	0.6058910	-0.7082396	0.5208611	3.031811	ukb	5902812	2	20227-filtered_func_data_clean
4	4.787917	2.285580	0.6165337	-0.9250505	0.6858148	3.023627	ukb	5902812	2	20227-filtered_func_data_clean
5	5.249457	2.357410	0.6653134	0.3319351	-0.2623765	3.070772	ukb	5902812	2	20227-filtered_func_data_clean
6	5.497924	2.396127	0.7443832	1.0094449	-0.7654791	3.095885	ukb	5902812	2	20227-filtered_func_data_clean

• t: volume
• A: total ("all") variability
• D: Fast ("difference") variability
• S: Slow ("average") variability
• DPD: "delta percent D", excess variance in the signal change as a percentage of the mean (total) signal change
• ZD: z-score for signal change
• DVARS: raw DVARS (likely not useful)

For interpretation, see Afouni & Nichols (2018) and Pham et al. (2023).

Calculations for various dvars measures based on fMRIscrub.

Import

In python, try polars.scan_ips

import polars as pl
pl.scan_ipc("derivatives/dvars").filter(pl.col("dataset")=="ukb").collect()

In R, try arrow::open_dataset

arrow::open_dataset("/Users/psadil/data/dvars/derivatives/dvars", format = "ipc") |> 
  filter(t > 0, dataset == "ukb") |>
  dplyr::collect()

Datasets

HCPAgingRec

Script

hcpaging

Notes

• Files with src containing "hp#" were provided by the HCP after (at least some) high-pass filtering, and no additional filtering was done before calculating DVARS
• Files with src containing "clean" were provided by the HCP after ICA FIX-ing, and no additional cleaning was done before calculating DVARS

Dataset Documentation

https://humanconnectome.org/study/hcp-lifespan-aging

HCPDevelopmentRec

Script

hcpadev

Notes

• Files with src containing "hp#" were provided by the HCP after (at least some) high-pass filtering, and no additional filtering was done before calculating DVARS
• Files with src containing "clean" were provided by the HCP after ICA FIX-ing, and no additional cleaning was done before calculating DVARS

Dataset Documentation

https://humanconnectome.org/study/hcp-lifespan-development

ukb

Script

ukb

Notes

• The UKB provides data that have already undergone high-pass filtering and so no additional cleaning was done before calculating DVARS
• The resting state vs task scans can be identified with the src entity
  • src=20227-filtered_func_data_clean: resting state (after ICA FIX'ing)
  • 'src=20249-filtered_func_data: task scans

Dataset Documentation

https://biobank.ndph.ox.ac.uk/showcase/download.cgi