Example Dataset#
CaliBrain workflows require local forward-model and leadfield files before larger simulation studies can be run. This page describes the example dataset layout expected by the current codebase and how to point CaliBrain to that data.
Note
These example data are provided solely to help users become familiar with CaliBrain workflows and data handling. They are not intended for evaluating the performance of any EEG, MEG, MRI, forward-modeling, or acquisition system.
Dataset access in the current workflow is local and explicit. The workflow configuration points to the directories that contain forward solutions and leadfield files.
Note
See the Dataset Notice for the distinction between the CaliBrain software license and third-party dataset terms.
Data root#
The local data root is expected to contain precomputed forward solutions and leadfield matrices. In practice, the workflow configuration should point to the directories that contain those files.
Available local example data#
The current workflow configurations assume an example source space with 1284 sources and multiple subjects.
Forward solutions versus leadfields#
The dataset layout distinguishes between two related but different objects:
Forward solution (
fwd): an MNE forward-model file, typically stored as*-fwd.fif. It contains the full source-to-sensor model together with source-space and geometry information. It is used when source coordinates or other source-space metadata must be recovered.Leadfield (
leadfield): a matrix or tensor extracted from the forward solution. It contains the numerical source-to-sensor mapping used directly by the simulation and inverse solvers. It is stored in compact.npzfiles for workflow use.
Full versus reduced representations#
The local data may also distinguish between full and reduced representations:
Full forward solution: retains the original forward-model object and associated metadata. It is appropriate when source-space geometry, orientation structure, or coordinate recovery is needed.
Reduced leadfield: stores only the numerical leadfield needed by the workflow. It is smaller and faster to load than the full forward solution and is intended for routine simulation and source-estimation runs.
Modality-specific overview#
The workflow uses the same directory structure for EEG and MEG, but the geometric interpretation differs by modality and orientation setting.
EEG#
Typical EEG-related files in the local example dataset include:
Path |
Contents |
|---|---|
|
Forward solutions such as |
|
Reduced fixed and free-orientation leadfields such as
|
In free-orientation EEG, each source location is represented with three local
components. This is the setting in which CaliBrain can work with both
marginal and full_cov uncertainty representations.
MEG#
Typical MEG-related files use the same local folders, but the reduced leadfields are interpreted differently:
Path |
Contents |
|---|---|
|
Forward solutions used when source coordinates or other source-space information must be recovered. |
|
Reduced fixed and reduced free-orientation leadfields used in simulation and inverse estimation. |
In reduced free-orientation MEG, each source location is represented with two tangential components rather than the full three-dimensional orientation used for EEG. This changes both the leadfield shape and the geometry of the local uncertainty representation.
Subject identifiers#
The default local example dataset commonly uses:
CC120166
CC120264
CC120309
CC120313
fsaverage
The exact subject list is controlled by the workflow configs. If a config requests a subject whose forward or leadfield file is missing, data generation or calibration will fail with a file-not-found error.
Minimal check#
Use this short check before running data generation:
from pathlib import Path
data_root = Path("/path/to/calibrain/data")
print(data_root)
print(sorted((data_root / "1284src_leadfield").glob("*_fixed_leadfield.npz")))
For a configured example dataset, the second line should print at least one fixed-orientation leadfield file.