PERSONALISED ELECTRICAL BRAIN IMAGING (PEBI)
Aim of the project was to develop next-generation computational tools for high spatial-resolution electroencephalography (EEG) based brain imaging technologies. “Personalisation” means here the consideration of individually varying geometry and electrical properties of the different head tissues, particularly the skull, that affect the recorded signals and thus can interfere with the diagnostics & treatment. In the future, we envision that portable EEG-based devices will enable remote & home-based brain monitoring and communication and become as common in personalised health tracking as smart watches.
Goal 1. Software that determines individual variations in the electric conductivities of the skull with the help of Electrical Impedance Tomography (EIT) imaging has been developed [1].
Goal 2. Software that takes personally tailored conductivity profiles and high spatial-resolution promoting models as inputs for the EEG brain imaging has been developed.
Goal 3. Testing of the multimodal EIT-EEG software.
Future work. We’ll utilize 3D machine vision and EIT technology to get information on the geometry of the head
Project link: https://attract-eu.com/showroom/project/personalised-electrical-brain-imaging-pebi/
EEG Brain Imaging solvers
In this project we have develop different techniques for the reconstruction of focal sources inside the brain using EEG potential measurements.
Super-resolution in sparse peak deconvolution on grids (application in fluorescent microscopy)
Recovery of spikes (location/amplitude) from measurements obtained by a convolution of the sparse signal with a kernel e.g. Gaussian or Lorentzian. The solution space is a discrete grid (low resolution). For the estimation of the actual locations, we solve the discrete L1-norm minimization problem, which results in numerical solutions with pair of peaks on the grid points around the locations of the actual peaks. Subsequently, we employ an extrapolation scheme to approximate the peak locations.
Maps development for the monitoring of ionospheric scintillation and quantifying its effect on the ground using Bayesian Filtering and Data Assimilation.
Automatic Segmentation of the Abdominal Aorta
The proposed method intends to segment the abdominal aortic lumen. The first step of the method is based on a “rough” approximation of the aortic lumen position. Assuming that the aortic-cross sections are circular we try to track the position of the abdominal aorta in each slice using the Kalman Filter. The observation (measurement) needed by the Kalman procedure is extracted from the Hough Circle Transformation. When the recursive Kalman procedure ends a level set method is applied.
Thoracic organs segmentation for RT planning
We focus on the automatic segmentation of the whole thoracic cavity (major organs e.g. lungs, heart) from CT images. The segmentation scheme, followed by the registration process, could enable the estimation of the motion of the organs during the breath cycle and thus can provide useful information to the radiotherapist about the position of the organs and the specification of the dose distribution that should be used between the tumour and the organs at risk.