\(\newcommand{L}[1]{\| #1 \|}\newcommand{VL}[1]{\L{ \vec{#1} }}\newcommand{R}[1]{\operatorname{Re}\,(#1)}\newcommand{I}[1]{\operatorname{Im}\, (#1)}\)

Example datasets

We will be using data from two OpenFMRI projects – ds114 and ds107. Data from both projects are covered by a Public Domain Dedication and License version 1.0 (PDDL 1.0).

Dataset ds114

We make most use of some files from the ds114 dataset.

The cited article for this study is:


There are more details of the analysis in:


In particular, we will be using the data from the covert verb generation task. Here is the description from the second (NeuroImage) paper linked above:

Subjects were asked to think of a verb complementing a noun presented to them visually. The following instructions were used: “When a word appears it will be a noun. Think of what you can do with it and then imagine saying ‘With that I can …’ or ‘That I can …’ “. A block design with 30 s activation and 30 s rest blocks was employed, with eight trials used for training. During the activation blocks, ten nouns were presented for 1 s each followed by a fixation cross during which subject had to generate the response. The nouns were chosen at random from a set of 70 nouns (mean lexical frequency: 0.000087, min: 0.000005, max: 0.000392, std: 0.000092). Rest blocks had an analogous structure but with each word replaced by scrambled visual patterns generated by scrambling the phase of the ‘picture’ of each word, i.e. the control patterns were matched in the amplitude spectrum. Seven activation/rest blocks were presented for a total scan time of 7 min 12.5 s.

There are more acquisition parameters listed at http://gigascience.biomedcentral.com/articles/10.1186/2047-217X-2-6#Tab1

The analysis scripts are available at: https://github.com/chrisfilo/2010-Reliability-Study

From the links above, for the covert verb generation task:

  • GE Signa 1.5T scanner
  • EPI (functional images)
    • TR = 2.5 s
    • TE = 50 ms
    • FOV = 256 x 256 mm
    • matrix = 64 × 64
    • 4 x 4 x 4 mm voxel size
    • 0 slice gap
    • 173 volumes per run
    • ‘Interleaved’ slice timing acquisition order
  • 3D Inversion Recovery Prepared (anatomical images)
    • TR = 10 s
    • TE = 4 s
    • 1 x 1 x 1.3 mm voxel size
    • 0 slice gap

The original analysis assumed ascending interleaved slice acquisition order - see the analysis script.

The original analysis dropped 4 dummy scans from the functional runs before continuing the analysis.

The onset times in the OpenFMRI model specification files almost certainly take time = 0 to be the beginning of the first dummy scan.

Dataset ds107

We will also be using data from the ds107 dataset.

Here is the article describing dataset and analysis:


From the methods section of that article, we find these bits of information:

  • Siemens 1.5 T MR scanner
  • EPI (functional images)
    • TR = 3000 ms
    • TE = 50 ms
    • FOV = 192 × 192
    • matrix = 64 × 64
    • “notional resolution” 3 × 3 × 3 mm
    • 164 volumes per run
  • T1 FLASH (anatomical images)
    • TR = 12 ms
    • TE = 5.6 ms
    • 1 mm3 resolution

Missing information

At least - I couldn’t find this information at a quick scan - please correct me if I’m wrong:

  • slice order (are they interleaved? sequential?);
  • volume gap in time (any gap between the last slice from volume N and the first slice of volume N+1?);
  • Slice width (was there a gap between the slices?).