Data analysis

Research on and application of data analysis methods in our laboratory focuses on the following topics:

FUNCTIONAL CONNECTIVITY

Two important principles of brain organization are functional specialization and integration. Functional integration refers to the mechanisms by which several subfunctions supported by different brain areas are combined to carry out a basic sensory/cognitive function. A number of connectivity measures have been proposed that capture the statistical dependence among time series reflecting the activity of given brain areas. There is a trade-off between how general and how data demanding these measures are. A publicly available toolbox, Hermes, has been developed by the laboratory

HERMES (Connectivity Toolbox)

The analysis of the interdependence between brain activity time-courses has become an important field of research in recent years. The CCNL has developed HERMES (http://hermes.ctb.upm.es/), a toolbox for the Matlab® environment, designed to study functional and effective brain connectivity from MEG and EEG multivariate recordings. A large number of connectivity measures are included such as the synchronization likelihood, phase synchronization, mutual information, Grainger causality, coherence and cross-correlation. HERMES also provides visualization tools and statistical methods to address the problem of multiple comparisons.

SOURCE RECONSTRUCTION

MEG offers the best combined spatiotemporal resolution to study brain function noninvasively in humans. A key analysis step to realize this potential is the estimation of sources of brain activity from the data recorded at the sensors. The sensor data does not uniquely determine a single solution. Additional assumptions are needed and different reconstruction methods exist such as beamforming, minimum-norm estimation or those based on Bayesian methods.

MULTIVARIATE ANALYSIS

Classification algorithms within the framework of machine learning can be applied to exploit the multivariate nature of MEG and EEG recordings. Considering the different sensors in a combined fashion can be advantageous both to understand neural coding and to separate experimental conditions and patient groups. Of particular interest for the laboratory is the early diagnosis in Alzheimer’s Disease.

PREPROCESSING METHODS

MEG and EEG recordings are contaminated by environmental noise and physiological artefacts. Researching and applying optimal preprocessing methods is essential to improve the signal to noise ratio of the data

BAYESIAN INFERENCE

The Bayesian framework is proving to be very valuable both for understanding brain function and as an analytical tool. According to Bayesian inference, the brain efficiently combines sensory information and prior knowledge according to their reliability to infer the functionally relevant causes of the sensory input.

Selected publications

Bruña R, Fuggetta G, Pereda E. One Definition to Join Them All: The N-Spherical Solution for the EEG Lead Field. Sensors (Basel). 2023 Sep; 23(19):. PubMed ID: 37836967.

Bruña R, Vaghari D, Greve A, Cooper E, Mada MO, Henson RN. Modified MRI Anonymization (De-Facing) for Improved MEG Coregistration. Bioengineering (Basel). 2022 Oct; 9(10):. PubMed ID: 36290559.  PDF file.

Bruña R, Maestú F, Pereda E. Phase locking value revisited: teaching new tricks to an old dog. J Neural Eng. 2018 10; 15(5):056011. PubMed ID: 29952757.

Dimitriadis SI, López ME, Bruña R, Cuesta P, Marcos A, Maestú F, Pereda E. How to Build a Functional Connectomic Biomarker for Mild Cognitive Impairment From Source Reconstructed MEG Resting-State Activity: The Combination of ROI Representation and Connectivity Estimator Matters. Front Neurosci. 2018; 12:306. PubMed ID: 29910704.  PDF file.

Niso G, Bruña R, Pereda E, Gutiérrez R, Bajo R, Maestú F, del-Pozo F. HERMES: towards an integrated toolbox to characterize functional and effective brain connectivity. Neuroinformatics. 2013 Oct; 11(4):405-34. PubMed ID: 23812847