University of Cambridge > > Clinical Neurosciences Seminars > Engineered neural networks as self-organised computational substrates in the healthy and lesioned CNS

Engineered neural networks as self-organised computational substrates in the healthy and lesioned CNS

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Dr Romina Vuono.

Neural networks demonstrate self-organising behaviour underscored by emergence/morphogenesis and self-organised criticality (SoC). Emergence involves cell behaviours driven by local cell-cell interactions and the spontaneous appearance of a highly ordered structure or function as a whole, not simply explained by the sum of the elements’ complexity. At the same time, SoC represents a universal characteristic of neural systems, being a spontaneous dynamic state established in networks of moderate complexity. In these networks, cascades of spontaneous activity are typically characterised by power-law distributions and rich, stable spatiotemporal patterns (neuronal avalanches). Thus SoC plays a functional role in neural computation by putatively maximising information transmission, the number of stable patterns, information capacity, and the range of usable inputs in a neural system.

A developed/mature neural network is expected to reach a fine balance of neural excitation and inhibition, consistent with normal function. Intrinsic or extrinsic perturbations to the network (e.g.evolving disease-related pathology; age-associated and/or epigenetic changes/ DNA damage) will result in remodelling of structural and functional connectomes. The associated changes can be adaptive or maladaptive in nature and involve an interplay between homeostatic and Hebbian plasticity contingent on the inherent state of the neuron, nature of the perturbation, and state of the microenvironment. A better understanding of such processes can be instrumental for our ability to elucidate fundamental mechanisms of neural network dynamics in the healthy and lesioned CNS . A highly promising perspective towards a better understanding of neural network behaviour is the application of advanced in vitro and in silico modelling tools incorporating deep learning principles.

Specifically, by applying advanced computational tools and machine learning principles, we can identify patterns that can reveal principal components influencing adaptive or maladaptive network responses as well as critical vulnerability states. This is crucial for early diagnosis, and timing/type of intervention (e.g. gene repair, DNA repair, or pharmacological treatment), as well as assessment of their efficacy. Fundamental research questions which we can address may thus include: can we associate and/or predict morphology-activity relationships at the neural network and synaptic level with progressive disease-related pathology? Do the most vulnerable neural networks self-regulate intrinsic responses to perturbations in an attempt to preserve normal function? If so, to what extent can homeostatic plasticity mask manifestation of disease pathology at the prodromal disease state? Are we able to identify critical vulnerability states that might instruct the timing and/or nature of appropriate interventions? Which aspects of complex neural network dynamics and at which level/scale are relevant for restoration of functional connectivity?

In this talk, I will present an overview of our relevant research activities at NTNU , providing examples of how we apply morphogenetic neuroengineering, computational tools and deep learning principles in an advanced modelling platform for the study and elucidation of evolving neural network dynamics in healthy and perturbed conditions.

This talk is part of the Clinical Neurosciences Seminars series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.


© 2006-2018, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity