- Signal & Image Processing.
- Bayesian estimation.
- Applied Mathematics.
Bayesian M/EEG Source Localization with possible Joint Skull Conductivity Estimation
M/EEG is a powerful non-invasive technique that measures the electric potentials at the scalp and the magnetic fields around the head. These measurements depend upon 1) the underlying brain activity and 2) the geometric composition of the head. The recovery of the brain activity from the measurements is an ill-posed inverse problem. Thus, a regularization is needed to narrow the search space. This regularization should typically be chosen to constrain the solution to have some realistic properties.
In this thesis we propose new Bayesian approaches to solve the M/EEG source localization problem. Our work is specifically focused on cases where the brain activity is spatially concentrated, such as in certain forms of epilepsy. The main idea underlying the thesis is to develop Baysian sparse models for the sources. More precisely, three contributions are presented:
1) Developing a hierarchical Bayesian model that solves the source localization problem for one instant of time by promoting sparsity using Bernoulli-Laplacian priors.
2) Investigating a Bayesian structured sparsity model to exploit the temporal dimension of the M/EEG measurements.
3) Developing Metropolis-Hasting sampling scheme that improves significantly the speed of the convergence of our MCMC algorithm.
4) Expanding the model to estimate the skull conductivity jointly with the brain activity.
Areas: Signal Processing, Image Processing, Medical imaging, Bayesian models
EVE2 Microprocessor Design
Design of a superscalar RISC microprocessor with a six stages pipeline implemented on FPGA.
Areas: Microprocessor Architecture, FPGA, VHDL
Evaluation of pre-processing algorithms PCA and DFT in ECG analysis with neural networks
Assement of detection of anormalities in electrocardiogram pulses using a neural network with a PCA or DFT pre-processing stage.
Presented at the 18th Argentine Congress on Bioengineering, Mar del Plata, Buenos Aires, 2011.
Areas: Medical imaging, Signal Processing, Neural networks
Implementation of the SURF algorithm to detect moving objects on a video.
Presented at the 2010 Buenos Aires Institute of Technology (ITBA) Electronics Fair.
Areas: Image Processing, Movement detection