mne.simulation.metrics.region_localization_error

mne.simulation.metrics.region_localization_error(stc_true, stc_est, src, threshold='90%', per_sample=True)

Compute region localization error (RLE) between 2 source estimates.

$$RLE=\frac{1}{2Q}\sum _{k\in I}\underset{l\in \hat{I}}{min}||r_k-r_l||+\frac{1}{2\hat{Q}}\sum _{l\in \hat{I}}\underset{k\in I}{min}||r_k-r_l||$$

where $I$ and $\hat{I}$ denote respectively the original and estimated indexes of active sources, $Q$ and $\hat{Q}$ are the numbers of original and estimated active sources. $r_k$ denotes the position of the k-th source dipole in space and $||\cdot ||$ is an Euclidean norm in ${\mathbb{R}}^3$.

Parameters:

stc_trueinstance of (Vol|Mixed)SourceEstimate

The source estimates containing correct values.

stc_estinstance of (Vol|Mixed)SourceEstimate

The source estimates containing estimated values e.g. obtained with a source imaging method.

srcinstance of SourceSpaces

The source space on which the source estimates are defined.

thresholdfloat | str

The threshold to apply to source estimates before computing the dipole localization error. If a string the threshold is a percentage and it should end with the percent character.

per_samplebool

If True the metric is computed for each sample separately. If False, the metric is spatio-temporal.

Returns:

metricfloat | array, shape (n_times,)

The metric. float if per_sample is False, else array with the values computed for each time point.

Notes

Papers [1] and [2] use term Dipole Localization Error (DLE) for the same formula. Paper [3] uses term Error Distance (ED) for the same formula. To unify the terminology and to avoid confusion with other cases of using term DLE but for different metric [4], we use term Region Localization Error (RLE).

New in v1.2.

References

[1]

Kostiantyn Maksymenko, Bernard Giusiano, Nicolas Roehri, Christian-G. Bénar, and Jean-Michel Badier. Strategies for statistical thresholding of source localization maps in magnetoencephalography and estimating source extent. Journal of Neuroscience Methods, 290:95–104, October 2017. doi:10.1016/j.jneumeth.2017.07.015.

[2]

H. Becker, L. Albera, P. Comon, J. -C. Nunes, R. Gribonval, J. Fleureau, P. Guillotel, and I. Merlet. SISSY: An efficient and automatic algorithm for the analysis of EEG sources based on structured sparsity. NeuroImage, 157:157–172, August 2017. doi:10.1016/j.neuroimage.2017.05.046.

[3]

Jun Yao and Julius P. A. Dewald. Evaluation of different cortical source localization methods using simulated and experimental EEG data. NeuroImage, 25(2):369–382, April 2005. doi:10.1016/j.neuroimage.2004.11.036.

[4]

Molins A, Stufflebeam S. M., Brown E. N., and Hämäläinen M. S. Quantification of the benefit from integrating MEG and EEG data in minimum l2-norm estimation. Neuroimage, 42(3):1069–1077, 2008. doi:10.1016/j.neuroimage.2008.05.064.

Examples using mne.simulation.metrics.region_localization_error

Compare simulated and estimated source activity

Compare simulated and estimated source activity