FIFF File Format Reference

The Functional Imaging File Format (FIFF) is the standard data format used throughout the MNE software ecosystem. Originally developed at the Neuromag/Elekta/MEGIN MEG systems group, FIFF stores MEG, EEG, and MRI data along with all associated metadata in a single, structured binary format.

Overview

FIFF files use a tag-based hierarchical structure. Each piece of information — a data array, a string, a coordinate transformation matrix — is stored as a tag identified by a numeric kind and type. Tags are organized into blocks (also called trees) that group related information together.

The MNE-CPP FIFF library (FIFFLIB) provides complete read/write support for FIFF files. See the Library API for programmatic access.

File Types

The MNE software uses FIFF files for many different purposes. The file naming conventions help identify the contents:

Suffix	Contents
_raw.fif	Raw MEG/EEG data
-ave.fif	Averaged (evoked) data
-cov.fif	Noise-covariance matrix
-fwd.fif	Forward solution
-inv.fif	Inverse operator
-src.fif	Source space
-bem.fif	BEM geometry
-bem-sol.fif	BEM geometry + solution
-trans.fif	Coordinate transformation
COR.fif	MRI description (for Neuromag MRIlab)

Key Data Blocks

Measurement Info

The measurement info block is present in raw, averaged, and processed data files. It contains:

• Channel information — channel names, types (MEG/EEG/STI/EOG/ECG), locations, orientations, unit, calibration, coordinate frame
• Digitization points — head shape, fiducials, HPI coil locations, EEG electrode positions
• Projection operators — SSP vectors for artifact rejection
• Compensation data — software gradient compensation (CTF/4D)
• Bad channels — list of channels marked as bad
• Sampling frequency
• Measurement date and subject information

Raw Data

Raw data files contain continuous time-series data from all channels. The data are stored in buffers (typically 600 samples per buffer) and can optionally be split across multiple files for very large recordings.

Evoked Data

Evoked (averaged) data files contain one or more averaged responses. Each response is stored with:

• The averaged data matrix (channels × time points)
• The number of averages
• Baseline information
• Stimulus category and comment

Forward Solution

Forward solution files contain:

• Source space description (vertex locations and orientations)
• The gain matrix $G$ mapping source currents to sensor signals
• Coordinate transformations (head-to-MRI, device-to-head)
• BEM model information
• Sensor and electrode locations

Inverse Operator

Inverse operator files contain the SVD decomposition of the inverse operator. This decomposition allows the regularization parameter to be adjusted without recomputing the operator. The file includes:

• The SVD components
• Source and noise covariance information
• Source space description
• Forward solution reference

Source Space

Source space files describe the locations and orientations of the current dipoles used in the forward and inverse calculations. They may also contain:

• Cortical patch statistics (neighborhood information)
• Triangle connectivity for surface-based source spaces
• The complete surface mesh for reference

BEM Model

BEM files contain the triangulated surfaces and conductivity values for the boundary-element forward model. The -bem-sol.fif variant also includes the geometry-dependent solution matrices.

Other File Formats Used by MNE

In addition to FIFF, the MNE software works with several other file formats:

FreeSurfer Formats

Format	Extension	Contents
Surface	(no extension, e.g., lh.white)	Triangulated cortical surface mesh
Annotation	.annot	Cortical parcellation (atlas labels)
Label	.label	Vertex list for a cortical region of interest
Curvature	.curv	Per-vertex curvature values
MRI volume	.mgz / .mgh	3D MRI volume data
Morphometry	.thickness, .area	Per-vertex morphometry

Source Estimate Formats

Format	Extension	Contents
STC	.stc	Dynamic source estimates (time series on cortex) — one file per hemisphere (-lh.stc, -rh.stc)
W file	.w	Static overlay data (single time point on cortex) — one file per hemisphere (-lh.w, -rh.w)

STC files contain:

• Vertex indices (subset of the source space)
• Time points (start time, sample period)
• Data matrix (vertices × time points)

Text Formats

Format	Extension	Contents
Event file	-eve.fif or .eve	Event timestamps and codes
Averaging script	.ave	Description file for offline averaging
Covariance script	.cov	Description file for noise-covariance computation
Projection file	.proj	SSP operator definition
Point file	.pnt	Surface points in text format
Dipole file	.dip	Dipole locations and orientations
Triangle file	.tri	ASCII surface triangulation

Data Conversion Tools

MNE-CPP provides several tools for converting data from other formats to FIFF:

Tool	Source Format
mne_edf2fiff	EDF / EDF+ / BDF
mne_brain_vision2fiff	BrainVision (.vhdr)
mne_ctf2fiff	CTF MEG (.ds)
mne_kit2fiff	KIT / Yokogawa (.sqd / .con)

For exporting FIFF data:

Tool	Target Format
mne_raw2mat	MATLAB (.mat)

Inspecting FIFF Files

Use mne_show_fiff to list the contents of a FIFF file:

mne_show_fiff --in sample_audvis_raw.fif

Use mne_compare_fif_files to compare two FIFF files tag by tag:

mne_compare_fif_files --in1 file_a.fif --in2 file_b.fif

Conventions

Command-Line Options

All MNE-CPP command-line tools employ the double-dash (--) option convention. Two universal options are available in all programs:

Option	Description
--help	Print concise usage information
--version	Print the program module name, version number, and compilation date

Environment Variables

Variable	Description
MNE_ROOT	Location of the MNE software installation
FREESURFER_HOME	Location of the FreeSurfer software
SUBJECTS_DIR	Location of the MRI data (FreeSurfer subject directories)
SUBJECT	Name of the current subject

See Also

• Tools Overview — Complete list of MNE-CPP command-line tools
• Workflow — The MEG/EEG processing workflow
• Library API — Programmatic access to FIFF files via FIFFLIB