Effects of anhedonia from multiple linear regressions controlling for sociodemographic factors and clinical comorbidities across the ABCD study samples.

Reproduction and replication t-test results comparing controls and those with anhedonia across ABCD study samples.

The partial regression coefficients (Estimate), standard errors (Std.Err), t-values, p-values, significance, model R2 (R^2), Bonferroni-Hochberg Adjusted p-values (BH adjustment), Bonferroni adjustd p-values, Durbin-Watson statistic (DW_statistic), Breusch-Pagan Chi^2 (BP Chi^2), and Breusch-Pagan (BP) p-values are presented. The ICC is the proportion of variance in rsfMRI connectivity explained by the family structure random effect. (XLSX)

Effect size (effect_CIs) with 95% confidence intervals represents the percentage of the total variance (proportion of variance * 100%) in a rsfMRI connectivity measure accounted for by each predictor (psychiatric symptom or diagnosis). Total model R^2 (total_R2), partial regression coefficients (Estimate), standard errors (Std.Err), t-values, p-values, and Benjamini-Hochberg corrected p-values (BH_adjustment) values are reported. (XLSX)

Single-echo/multi-echo comparison pilot

This dataset contains ABCD-protocol single-echo BOLD scans, along with complex-valued, multi-echo BOLD scans for comparison. The multi-echo BOLD protocol uses the CMRR MB-EPI sequence, and comes from collaborators at UMinn. These scans include five echoes with both magnitude and phase reconstruction.

The primary goal of this dataset was to evaluate the usability of the multi-echo fMRI protocol in a larger study, via direct comparison to the ABCD fMRI protocol, as well as via test-retest reliability analyses. However, these data may be useful to others (e.g., for testing complex-valued models, applying phase regression to multi-echo data, testing multi-echo denoising methods).

Dataset contents

This dataset includes 8 participants, each with between 1 and 3 sessions. MR data were acquired using a 3-Tesla Siemens Prisma MRI scanner.

The imaging data were converted to NIfTI-1 format with dcm2niix v1.0.20220505, using heudiconv 0.13.1.

In each session, the following scans were acquired:

Structural data

A T1-weighted anatomical scan (256 slices; repetition time, TR=1900 ms; echo time, TE=2.93 ms; flip angle, FA=9 degrees; field of view, FOV=176x262.144 mm, matrix size=176x256; voxel size=1x0.977x0.977 mm).

Functional data

One run of Penn fractal n-back task five-echo fMRI data (72 slices; repetition time, TR=1761 ms; echo times, TE=14.2, 38.93, 63.66, 88.39, 113.12 ms; flip angle, FA=68 degrees; field of view, FOV=220x220 mm, matrix size=110x110; voxel size=2x2x2 mm; multiband acceleration factor=6). Both magnitude and phase data were reconstructed for this run. The run was 7:03 minutes in length, including the three no-radiofrequency-excitation volumes at the end. After the _noRF volumes were split into separate files, each run was 6:58 minutes long.

Two runs of open-eye resting-state five-echo fMRI data (72 slices; repetition time, TR=1761 ms; echo times, TE=14.2, 38.93, 63.66, 88.39, 113.12 ms; flip angle, FA=68 degrees; field of view, FOV=220x220 mm, matrix size=110x110; voxel size=2x2x2 mm; multiband acceleration factor=6). Both magnitude and phase data were reconstructed for these runs. Each run was 5:59 minutes in length, including the three no-radiofrequency-excitation volumes at the end. After the _noRF volumes were split into separate files, each run was 5:54 minutes long.

Two runs of open-eye resting-state single-echo fMRI data acquired according to the ABCD protocol (60 slices; repetition time, TR=800 ms; echo time, TE=30 ms; flip angle, FA=52 degrees; field of view, FOV=216x216 mm, matrix size=90x90; voxel size=2.4x2.4x2.4 mm; multiband acceleration factor=6). Only magnitude data were reconstructed for these runs. Each run was 6:00 minutes in length.

Field maps

Two sets of field maps were acquired for the multi-echo fMRI scans.

One set was a multiband, multi-echo gradient echo PEpolar-type field map (acq-MEGE), acquired with the same parameters as the multi-echo fMRI scans (except without magnitude+phase reconstruction). For each acquisition, we have created a copy of the single-band reference image from the first echo as the primary field map.

The other set was a multi-echo spin-echo PEpolar-type field map (acq-MESE). We have also created a copy of the first echo for each direction as a standard field map.

The single-echo copies of both the acq-MEGE and the acq-MESE field maps have B0FieldIdentifier fields and IntendedFor fields, though we used the acq-MESE field maps for the B0FieldSource fields of the multi-echo fMRI scans. Therefore, tools which leverage the B0* fields, such as fMRIPrep, should use the single-echo acq-MESE scans for distortion correction.

Single-echo PEpolar-type EPI field maps (acq-SESE) with parameters matching the single-echo fMRI data were also acquired for distortion correction.

Dataset idiosyncrasies

Multi-echo field maps

There are two sets of PEpolar-style field maps for the multi-echo BOLD scans: one gradient echo and one spin echo. Each field map set contains five echoes, like the BOLD scans. However, because distortion shouldn't vary across echoes (at least not at 3T), there is no need for multi-echo PEpolar-style field maps, and tools like fMRIPrep can't use them. As such, we have made a copy of the spin echo field map's first echo without the echo entity for BIDS compliance, as well as a copy of the gradient echo field map's first echo's single-band reference image.

No radio frequency excitation scans

The multi-echo BOLD scans included three no-radio-frequency noise scans acquired at the end of the scan, which have been split off into files with the _noRF suffix. These noise scans can be used to suppress thermal noise with NORDIC denoising. BOLD runs that were stopped early or failed to fully reconstruct may be missing these noise scans.

The _noRF suffix is not (as of 2024/03/22) supported within BIDS, but there is an open pull request adding it (https://github.com/bids-standard/bids-specification/pull/1451).

NORDIC-denoised BOLD runs

We have run NORDIC on the multi-echo scans, using the noRF files when available. The NORDIC-denoised data have the rec-nordic entity in the filenames.

We have made copies of the associated single-band reference images as well.

sub-08 ses-noHC

Subject 08's ses-noHC was accidentally acquired without the head coil plugged in. We included the session in the dataset in case anyone might find it useful, but do not recommend using the data for analyses.

sub-04 ses-2 and ses-3

Subject 04 had to stop session 2 early, so a separate session was acquired to finish acquiring the remaining scans.

Excluded data

Physio (PPG + chest belt) data were acquired for a subset of the scans, but, due to equipment issues, the data were unusable and have been excluded from the dataset.

There was also an MEGRE field map sequence in the protocol, provided by Dr. Andrew Van, but there were reconstruction errors at the scanner, so these field maps were not usable. We've chosen to exclude the remaining files from the dataset.

In some cases, we noticed reconstruction errors on final volumes in the multi-echo BOLD runs. When that happened, we dropped any trailing volumes, so that all files from a given run are the same length. For some runs, this involved entirely removing the noRF scans.

Penn Fractal N-Back events files

The events files for the fractal n-back task are not included in version 1.0.0 of the dataset. We will add them in a future patch release.

Notes about acquisition

The multi-echo BOLD scans were acquired on a 3T Siemens Prisma scanner running VE11C. The same protocol has exhibited consistent reconstruction errors on XA30.

Effect size (in Pearson’s r) quantiles for analytic variations.

Effect sizes (Cohen’s d) with 95% Cis, t-statistics, p-values, ln(Bayes Factor) (lnBF), group means with standard deviations (SD), Shapiro-Wilk statistics (W-statistic), and F-statistics are reported. (XLSX)

Comparison of sociodemographic measures between controls and those with anhedonia in the ABCD 4.0 release, excluding ABCD 1.0 release, sub-sample.

subs-32_effect-onesample_contrast-Lgain-Neut_stat-cohensd_sample-ABCD_type-site19.nii.gz

Collection description

This collection includes the output files, z- and/or t-statistics (uncorrected), cohen-s D statistical maps (calculated: t-statistic / sqrt(n)) for subjects from the Adolescent Brain Cognitive Development (ABCD), Adolescent Health Risk Behavior (AHRB) and Michigan Longitudinal Sample (MLS). The analyses are for the modified monetary incentive delay task. ABCD and AHRB are multiband acquisitions (TR = 800ms, vox = 2.4mm; MB factor = 6). MLS is a spiral acquisition (TR = 2000ms; vox = 4mm). All datasets were preprocessed using fMRIprep v23.1.0 (fmap correction in AHRB/ABCD and no slice-time correction; MLS no fieldmap correction with slice-time correction).

For ABCD, AHRB and MLS, the following [six[ contrasts types are included for the anticipation phase (*_contrast-[type]*):
Lgain-Neut: Large Gain ($5) > Neutral (no money at stake)
LSgain-Neut: Large Gain ($5, .5) & Small Gain ($0.20, .5) > Neutral (no money at stake)
Lgain-Lloss: Large Gain ($5) > Large Loss (-$5)
Lloss-Neut: Large Loss (-$5) > Neutral (no money at stake)
LSloss-Neut: Large Loss (-$5, .5) & Small Loss (-$0.20, .5) > Neutral (no money at stake)
Lloss-Lgain: Large Loss (-$5) > Large Gain ($5)

For the ABCD (N = 346) , AHRB (N = 97) and MLS (N = 112) there are group level contrasts for each of the two runs, e.g. *_type-run-01.nii.gz, *_type-run-02.nii.gz. For ABCD samples maps are included across different scanners and sites. Maps include site specific (e.g, *_type-site06.nii.gz) and scanner specific (e.g, *_type-SIEMENS.nii.gz) maps. Note, the N for each map is indicated in the first field, e.g subs-347 the sample size N = 347.

Subject species

homo sapiens

Map type

Other

The proportion (%) of participants in each group for each measure are shown. Student’s t-test was done to compare age (in months) between the “removed” and “retained” groups. Chi-square tests of independence were done for all other measures. P-values < 0.05 are bolded. (XLSX)

Average Generalized Variance Inflation Factor (GVIF) values for predictors from multiple linear regressions controlling for sociodemographic covariates across the ABCD samples.

b'stat of contrast [1.]'

Collection description

This collection includes the output files, z- and/or t-statistics (uncorrected), cohen-s D statistical maps (calculated: t-statistic / sqrt(n)) for subjects from the Adolescent Brain Cognitive Development (ABCD), Adolescent Health Risk Behavior (AHRB) and Michigan Longitudinal Sample (MLS). The analyses are for the modified monetary incentive delay task. ABCD and AHRB are multiband acquisitions (TR = 800ms, vox = 2.4mm; MB factor = 6). MLS is a spiral acquisition (TR = 2000ms; vox = 4mm). All datasets were preprocessed using fMRIprep v23.1.0 (fmap correction in AHRB/ABCD and no slice-time correction; MLS no fieldmap correction with slice-time correction).

For ABCD, AHRB and MLS, the following [six[ contrasts types are included for the anticipation phase (*_contrast-[type]*):
Lgain-Neut: Large Gain ($5) > Neutral (no money at stake)
LSgain-Neut: Large Gain ($5, .5) & Small Gain ($0.20, .5) > Neutral (no money at stake)
Lgain-Lloss: Large Gain ($5) > Large Loss (-$5)
Lloss-Neut: Large Loss (-$5) > Neutral (no money at stake)
LSloss-Neut: Large Loss (-$5, .5) & Small Loss (-$0.20, .5) > Neutral (no money at stake)
Lloss-Lgain: Large Loss (-$5) > Large Gain ($5)

For the ABCD (N = 346) , AHRB (N = 97) and MLS (N = 112) there are group level contrasts for each of the two runs, e.g. *_type-run-01.nii.gz, *_type-run-02.nii.gz. For ABCD samples maps are included across different scanners and sites. Maps include site specific (e.g, *_type-site06.nii.gz) and scanner specific (e.g, *_type-SIEMENS.nii.gz) maps. Note, the N for each map is indicated in the first field, e.g subs-347 the sample size N = 347.

Subject species

homo sapiens

Map type

Other