Validation Framework For Source Localization

This repository contains all the codes and data relative to the following paper:

La Fisca et al., A Versatile Validation Framework for ERP and Oscillatory Brain Source Localization Using FieldTrip, ICBEA '21: 4th International Conference on Biometric Engineering and Applications, May 2021, Pages 7–12

Tutorial

Requirements

• MATLAB 2016+
• FieldTrip Toolbox
• SEREEGA Toolbox

Benchmark (plug and play)

1. Clone this github repository : git clone https://github.com/LucaLaFisca/Validation-Framework-Source-Reconstruction.git in command prompt
2. Download template data from Zenodo to template subfolder
3. Set benchmark parameter in config.json to true and type to ERP or OSCIL depending on the signal type you want to analyze
4. Run the main script validation_framework.m (it is a very long process... Don't worry)
5. Run your own source reconstruction pipeline on the newly generated pseudo-EEG stored in pseudo_eeg folder (one mat file by session). Please, use the template atlas to compute the final region-by-region signal.
6. Save the n_sessions x 1 structure with fields signaland label corresponding to the reconstructed sources and set the corresponding path in config.json as reconstr_source parameter (ensure label order is the same than atlas labels)
7. Set evaluation parameter to true and run again the main script. The benchmark score appears in the second figure.

Custom analysis

1. Clone this github repository : git clone https://github.com/LucaLaFisca/Validation-Framework-Source-Reconstruction.git in command prompt
2. Set the configuration parameters with respect to their definition (see below) and set benchmark and evaluation parameters to false
3. Run the main script validation_framework.m (it is a very long process... Don't worry)
4. Run your own source reconstruction pipeline on the newly generated pseudo-EEG stored in pseudo_eeg folder (one mat file by session). Please, use the atlas defined in the config to compute the final region-by-region signal.
5. Save the n_sessions x 1 structure with fields signaland label corresponding to the reconstructed sources and set the corresponding path in config.json as reconstr_source parameter (ensure label order is the same than atlas labels)
6. Set evaluation parameter to true and run again the main script. The final score appears in the second figure.

Description of configuration parameters

Parameter	Pattern	Definition
benchmark	true/false	run benchmark (template parameters)
evaluation	true/false	run the evaluation (set to true when reconstructed sources are computed)
PATH_TO_FIELDTRIP	path name	Path to FieldTrip toolbox
PATH_TO_SEREEGA	path name	Path to SEREEGA toolbox
n_sessions	int	number of sessions over which to generate pseudo-data (different dipoles for each session)
n_dipoles	int	number of dipoles defined as source (recommended between 2 and 5)
n_trials	int	number of occurences of source activation within one session
n_artifacts	int	number of artifactual segments occuring within one session
fsample	int	sampling frequency of generated EEG
session_duration	float	duration in minutes of each session
pseudo_length	float	duration in seconds of each trial (source activation)
event	n_sessions x n_trials matrix	matrix defining the starting time of each trial
type	ERP/OSCIL	definition of source signal type (event-related or oscillatory)
ERP.peaks	Pxxx/Nxxx series	definition of ERP peaks as series of positive/negative peaks (e.g. P100,N200,P300)
ERP.ampli	float series	maximum amplitude of each peak
ERP.width	int	width of each peak corresponding to 6 standard deviation
OSCIL.freq	"f1": [min_f, max_f], "f2": [min_f2,max_f2],...	definition of the frequency band of each desired oscillation
OSCIL.max_freq	int	maximum default frequency if OSCIL.freq is not defined
OSCIL.ampli	float series	maximum amplitude of each oscillation
OSCIL.modulation	none/ampmod	enable or not amplitude modulation of the predefined osciillations
atlas	mesh structure (FieldTrip)	FieldTrip atlas structure defining targeted brain regions (cf. ft_read_atlas)
dipoles_selection	n_sessions x n_dipoles matrix	matrix defining the index of the selected dipoles (relatively to the atlas structure)
avoid_2nd_neighb	true/false	avoid or not the selected dipoles to be second neighbors (i.e., neighbor of neighbor) of each other
headmodel	headmodel structure (FieldTrip)	head model FieldTrip structure corresponding to the volume conduction model (cf. ft_prepare_headmodel)
elec	elec structure (FieldTrip)	electrode FieldTrip structure describing the EEG sensors (cf. ft_datatype_sens)
channels	1 x n_channels vector	vector defining which channels to work with (relatively to the elec structure)
reconstr_source	n_sessions x 1 structure	signal and label of the reconstructed sources from the generated pseudo-EEG (cf. template/reconstr_source.mat)
artifacts	struct(artf,time,fsample,label)	structure containing the artifactual segments (artf field) with specific longest time vector (1 x n_sample), sampling frequency (fsample field) and channel labels (n_channel x 1 cell) (cf.template/artf_template.mat)
snr_source	int	signal-to-noise ratio of generated source signal
snr_eeg	int	signal-to-noise ratio of generated EEG signal