mne_simu

Overview

mne_simu generates simulated MEG/EEG raw data from a forward solution. A sinusoidal signal is placed at a specified source location and projected through the forward model, with optional additive noise from a noise covariance matrix. This is useful for testing inverse methods and validating processing pipelines.

This is a C++ port of the original MNE-C tool by Matti Hämäläinen.

Usage

mne_simu [options]

Options

Option	Description
--fwd <file>	Forward solution FIFF file
--raw <file>	Raw FIFF file (for channel info and sampling rate)
--cov <file>	Noise covariance FIFF file (optional)
--source <idx>	Source index to activate (default: 0)
--snr <value>	Signal-to-noise ratio in dB (default: 20)
--duration <sec>	Duration of simulation in seconds (default: 1.0)
--freq <Hz>	Signal frequency in Hz (default: 10.0)
--out <file>	Output simulated raw FIFF file
--help	Print help
--version	Print version

Description

The simulation generates a sinusoidal time course at a single source location, projects it through the forward solution to produce sensor-level data, and optionally adds Gaussian noise scaled to achieve the specified SNR. The channel configuration and sampling rate are taken from the template raw file.

Workflow Context

Simulation is used for:

• Inverse method validation: Generate data with known source location and verify that inverse methods recover it correctly.
• Pipeline testing: Test processing pipelines end-to-end with controlled data.
• SNR studies: Evaluate how source localization accuracy depends on noise levels.

The typical workflow is:

1. Compute forward solution → mne_forward_solution
2. Generate simulated data → mne_simu
3. Apply inverse method → mne_compute_mne or mne_dipole_fit
4. Compare recovered source location with ground truth

Example

# Simulate 2 seconds of 10 Hz activity at source 42 with 15 dB SNR
mne_simu --fwd sam-fwd.fif --raw sam-raw.fif --cov noise-cov.fif \
    --source 42 --snr 15 --duration 2.0 --freq 10.0 --out simulated-raw.fif