Biomag 2016

ChildBrain BIOMAG2016 Symposium

 

Symposium Title

Improving source reconstruction for MEG and EEG in children

Chairperson(s)

PD Dr. Carsten Wolters (Institute for Biomagnetism and Biosignalanalysis, University of Münster, Germany, carsten.wolters@uni-muenster.de)

and

Prof. Dr. Robert Oostenveld (Donders Institute for Brain, Cognition and Behaviour, Nijmegen, The Netherlands, r.oostenveld@donders.ru.nl)

They are both the lead beneficiaries of the methodological work package WP3.

Symposium Summary

The European Brain Council (EBC) has recommended disorders of the brain to be prioritized for funding. One successful example of this is the Marie Curie Innovative Training Network ChildBrain which aims on the one hand to train young researchers and on the other hand to utilize evidence-based neuroscientific knowledge for helping children, especially those at high risk for dropout due to neurocognitive disorders, to meet future educational and societal demands.

In the ChildBrain network we develop new, innovative brain imaging-based tools in collaboration between research and industry and that can be applied by researchers and clinicians.

MEG and EEG data acquisition (movements and SNR) and modelling (volume conduction) are specifically challenging in children. In this proposed Biomag2016 session we will highlight the ChildBrain brain research methods work package. This provides not only value to the use of MEG and EEG in children, but will also contribute to improving the application in adults.

Speakers Information

Affiliation

BESA GmbH, Gräfelfing, Germany (Beneficiary)

Title

An automatic Markov Random Field-based approach for segmentation of volume conductor models of the human head

Abstract

The accurate solution of the EEG and MEG forward problem requires taking into account information about a subject's individual anatomy. This information is contained in the volume conductor model which describes the electrical properties of the subject's head. Commonly, individual models are constructed by first segmenting the head into the different tissue compartments based on the available medical image data. Next, a suitable discretization of the head domain is computed, and previously published tissue conductivities are assigned.

Here, we propose a new automatic segmentation approach for segmenting the tissues of the head. The approach is formulated in a Bayesian framework. A-priori knowledge about the anatomy is included from two sources. A Markov Random Field model encodes our knowledge about the general arrangement of the head tissues. Secondly, a custom probabilistic tissue atlas further facilitates the segmentation.

Validation studies versus CT-based and manual segmentations were performed. Results prove the accuracy and reliability of the proposed approach. Average accuracies for the skull segmentation reached values of 88%.

In the ChildBrain project an approach for segmenting volume conductor models of infant subjects and patients will be developed. An outlook will discuss the related challenges and how we are aiming to solve them.