Neural Engineering Seminar by Daniel Miklody. Oct 31

Daniel Miklody. Oct 31 at 2 pm on the 5th floor BME conference room.

Individualized head models through electrical impedance measurements

In EEG source localization,transcranial current stimulation (tCS) and other topics in neuroscience, a model of the volume conduction properties of the head is needed to estimate e.g. the sources of activity in EEG or the areas of stimulation for tCS.

To create a model, two approaches seem predominant: either an individual MRI is obtained out of which a model is created or an average head model is used. An MRI is expensive and the process to receive a head model very time consuming and labour intensive. Average head models are much cheaper but systematic errors occur through the individual differences in anatomy and conductivities. The anatomy can be morphed to fit the outer head shape as measured by localization devices. The results are acceptable but modeling errors still occur.

I am introducing a new approach to this topic, which involves Electrical Impedance Tomography (EIT) to individualize the headmodel. Therefor small electric currents are injected through a standard set of EEG electrodes. The pattern of injection is alternating between injection pairs and the resulting scalp potential is measured on the rest. The measured data is then used to individualize an average head model involving non-linear optimization techniques from machine learning.

In my talk I will give a gentle introduction to different ways of modeling the head (concentric spheres, spherical harmonics, BEM and FEM), their advantages and drawbacks.

I will continue with an introduction to Electrical Impedance Tomography and different approaches within. I will also present the EIT device (virtually only) that I have designed an constructed within my Master’s thesis and some of the results.

Currently I am working on two algorithms for this problem in parallel: a leadfield based approach and a geometry based approach involving dimensionality reduction through PCA. I will introduce some basic concepts of those.

I will also present some preliminary results on 4-shell (scalp,skull,CSF,brain) BEM head modeling and the leadfield based approach.

Neural Engineering