I finished my 5 year Diploma degree on September 2014 from the University of Patras, Greece in the department of Computer Engineering and Informatics with a focus mainly on Computer Vision and Signal and Image Processing. I also did a research internship at INRIA with the Perception team, working with the problem of Shape Correspondence. I just finished my MRes with distinction from UCL. I am now starting my 3rd year of the 4 year MRes + MPhil + PhD course at UCL in the Translational Imaging Group and I am working on Dr. Jason Warren's Nexopathy project. My supervisors are Prof. Sebastien Ourselin, Prof. Jason Warren and Dr. Marc Modat.
I am working on understanding the mechanisms that drive the pathogenic proteins (e.g. Ab-42, tau, etc.) present in most neurodegenerative diseases to display different spatiotemporal patterns of spread in the brain's neural network. These proteins are hypothesized to induce the misfolding of normally folded proteins when in close proximity and that there are many mechanisms by which they could be spreading around in the brain. The brain's clearance mechanisms are either completely unable to remove the misfolded proteins (insoluble) or after a certain instigating event they are unable to cope with the amount of misfolded protein that is being accumulated (soluble, but too much).
During my MRes year and 1st PhD year we used the NEURON simulator and simulated a neural network comprised of 3 cortical columns. We added an abstract pathogenic protein to the network and varied the parameters of the protein's mechanisms. Analysing the results, we reached certain conclusions: the network's connectivity structure plays a more important role with faster moving proteins (i.e. hub/bottleneck areas of the brain are more vulnerable), whereas the distance from the protein seed plays a more important role with slower moving proteins (i.e. neurons close to the seed are more vulnerable). This can be roughly confirmed by the real-life patient cases with neurodegenerative diseases.
In the next years of the PhD, we will create similar computational models for the full brain that we will be able to validate based on longitudinal atrophy imaging. Following that, we will construct an atrophy or pathogenic protein concentration prediction software.
Georgiadis K, Selina W, Ourselin S, Warren J and Modat M: Simulation of Pathogenic Protein in an Artificial Neural Network.10th ICFTD Munich 2016