arXiv subjects

Stefano De Blasi

Publications and source records attributed to Stefano De Blasi.

2 recordsLinked to original sources

arXiv2020

Connectivity estimation of high dimensional data recorded from neuronal cells

The main result of this thesis is the development of a novel connectivity estimation method, called Total Spiking Probability Edges (TSPE). Based on cross-correlation and edge filtering at different time scales this method is proposed and the theoretical framework is outlined in this work. TSPE enables the classification between inhibitory and excitatory connections by using recorded action potentials. To compare this method learning about state of the art algorithms to estimate connectivity is necessary. After a research, promising algorithms are implemented and evaluated for further research topics, among others in the biomems lab of UAS Aschaffenburg. To evaluate these algorithms in silico networks are used, because of their known connectivity. This makes it possible to validate the correctness of our algorithm results. Therefore, a biophysically representative neuronal network simulation is needed first. Datasets were simulated in different ways and analysed in order to develop an evaluation framework. After a successful evaluation with in silico networks, in vitro experiments and their analyses complete this project.

eess.SP↗

arXiv2018

Simulation of Large Scale Neural Networks for Evaluation Applications

Understanding the complexity of biological neural networks like the human brain is one of the scientific challenges of our century. The organization of the brain can be described at different levels, ranging from small neural networks to entire brain regions. Existing methods for the description of functionally or effective connectivity are based on the analysis of relations between the activities of different neural units by detecting correlations or information flow. This is a crucial step in understanding neural disorders like Alzheimers disease and their causative factors. To evaluate these estimation methods, it is necessary to refer to a neural network with known connectivity, which is typically unknown for natural biological neural networks. Therefore, network simulations, also in silico, are available. In this work, the in silico simulation of large scale neural networks is established and the influence of different topologies on the generated patterns of neuronal signals is investigated. The goal is to develop standard evaluation methods for neurocomputational algorithms with a realistic large scale model to enable benchmarking and comparability of different studies.

q-bio.NC↗