This section covers various informational files and other files relevant to how the NSD data were pre-processed.
This is a detailed, comprehensive list of all known data problems. Most of these problems are very minor, but we are deliberately comprehensive so that the user understands what is in the data.
A table that provides an overview of all of the NSD data collected.
A table that provides demographic information (age, sex) on the NSD subjects as well as basic information concerning vision- and language-related abilities. The table also provides behavioral data for TOWRE and VVIQ.
A table that provides information at the level of individual scanning sessions. Includes information such as time of session, notes on eyetracking and physiology, sleepiness ratings, mood, stress, hunger, general notes on scanning, and subject feedback.
This file contains some data quality metrics that are computed at the level of individual NSD runs. There are two variables:
- ‘runmetrics’ is 8 subjects x 40 sessions x 12 runs x 7 metrics. The seven metrics, in order, are (1) tSNR (this was computed by taking the raw functional volumes, computing the voxel-wise mean of each voxel divided by the voxel-wise standard deviation after quadratic detrending of each voxel’s time-series, and the calculating the median value observed across a liberal whole-brain mask), (2) FD (this was computed on the 1.8-mm version of the pre-processed data by computing the absolute value of the temporal derivative of each of the six motion parameters in each run, computing a weighted sum according to weights [1 1 1 50 50 50], and calculating the mean FD across the volumes in the run), (3) ON-OFF R^2 (for the 1.8-mm version of the data, we fit a simple ON-OFF GLM to the voxel time-series, and we extract the variance explained for each run and compute the median variance explained across voxels in the nsdgeneral ROI), (4) response rate (percentage of trials on which a button was pressed), (5) percent correct (percentage of trials for which the subject pressed the correct response), (6) easy trials (percentage of easy trials (trials that are memory events for an image seen earlier in the scan session) for which the subject pressed the correct response; can be NaN for cases where there are zero easy trials), and (7) number of easy trials (the number of easy trials that actually occurred; this is useful because some runs might have zero or very few easy trials).
- ‘runmetricsRS’ is 8 subjects x 40 sessions x 2 runs x 2 metrics. The two metrics, in order, are tSNR and FD, as described above. When acquired, the resting-state runs were acquired as the very first and very last run in a given session.
Note that because not all subjects participated in all 40 sessions, some of the values in ‘runmetrics’ are NaN. Also, note that because resting-state data were acquired in only certain sessions, some of the values in ‘runmetricsRS’ are NaN. Also, note that for subject 8’s second NSD session, the fourth run was actually split across two distinct scan sessions (on different days); when computing FD, we compensated for this discontinuity (by dropping the appropriate volume), and when computing tSNR, we considered only the first segment of the fourth run. Also, note that for subject 1, session 2, run 2, there was complete MR signal loss for a few volumes in the middle of the run, and for this reason the tSNR values are abnormally low for that run (in the pre-processing of the data, compensation was applied to appropriately deal with this issue).
This text file contains a 2D matrix of dimensionality 40 sessions x 8 subjects. The entries indicate the number of nuisance regressors chosen by GLMdenoise for each NSD scan session. NaNs indicate scan sessions that subjects did not participate in.
FreeSurfer configuration file that was used.
Information file copied from the FreeSurfer software package.
Each of these files contains a variable ‘hrfs’ that has dimensions time-points x 20 HRFs. The first time point is coincident with trial onset. There are 20 different HRFs comprising the library of HRFs used to estimate voxel-specific HRFs. The ‘func1mm’ version has a sampling rate of 1-s whereas the ‘func1pt8mm’ version has a sampling rate of 1.333-s.
Contains HRF parameters (using the parametric function implemented in spm_hrf.m) that were determined by fitting each of the HRFs in the library of HRFs (as described above). The variable ‘params’ is 20 different HRFs x 7 parameters.
An example MATLAB script that generates an figure illustrating the contents of the hrfparams.mat file.
MNI template files copied from fsl-5.0.7/fsl/data/standard. These were used in the pre-processing of the NSD data.
Configuration file used in the T1-to-MNI alignment procedure.