Carsten Stahlhut © 2008 • Email: cs(at)imm.dtu.dk or carsten.stahlhut(at)gmail.com • Phone: +45 4525 3894 

 

News

• Visiting BIL at UCSF » 
Visiting Prof. Srikantan Naga-rajan's Biomagnetic Imaging Lab at UCSF, Feb-June 2010.

• MLSP 2009 » 
Acceptance of "Hierarchical Bayesian model for simultaneous EEG source and forward model reconstruction (SOFOMORE)", 12th June 2009.

• ISBI 2009 » 
Acceptance of "SOFOMORE: Combined EEG source and forward model reconstruction", 7th April 2009.

 


Research

My primarily research area is machine learning for biomedical signals which cur-rently focuses on functional brain imaging using electroencephalography (EEG) signals.

My PhD project concerns functional brain imaging, where modeling in the context of EEG is my current focus. EEG is a measurement technique of the electrical activity in the brain. EEG is directly related to neuronal signaling and is measured by a set of sensors placed on the scalp in a structured way. Compared to the widely used fMRI and PET, EEG offers high temporal resolution, in fact below 100 ms, and is therefore highly interesting for neuropsychological applications.

Functional brain imaging using EEG is a relatively new research field, which aims at a better understanding of the human brain by localizing so-called current sources with the high temporal resolution. In modern EEG imaging methods, the sources are represented as a dipole field, where each dipole is assigned to a network of tens of thousands cortical neurons.

To accomplish this EEG source localization problem, which is highly underdetermined, my PhD project deals with topics like:

  • Variational Bayesian (VB) inference methods
     
  • Markov Chain Monte Carlo (MCMC) methods
     
  • Sparsity (e.g. ARD priors for the current sources in the brain)
     
  • Spatio-temporal basis functions to incorporate spatial and temporal dependence of the sources
     
  • Clustering