Investigating, implementing, and creating methods for analysing large neuronal ensembles

  • Thomas J Delaney

Student thesis: Doctoral ThesisDoctor of Philosophy (PhD)


Since the use of multi-electrode recording in neuroscience began, the number neurons being recorded in parallel has been increasing. Recently developed methods using calcium or voltage imaging have also contributed to the growth in neuronal datasets. As datasets grow, the need for new analysis methods also grows. In this research we attempted to address some of the problems associated with reading from large neuronal ensembles using fluorescent calcium indicators, and some of the problems with analysing data read from large neuronal ensembles.
We created a biophysical model for the fluorescence trace produced by a calcium indicator responding to a given spike train. Our model reproduced the characteristics of a real fluorescence trace recognised by spike inference algorithms. This model will be useful for anyone using or considering calcium imaging.
To find order in the correlated behaviour of a large multi-region neuronal ensemble, we
applied a novel method from network science to detect structure and communities in correlated behaviour. We investigated the similarities between these communities and their brain anatomy. Our results indicate local correlated networks function at shorter timescales ( 100ms). This result agrees with previous findings from EEG data, but has not been shown before using spiking data.
We developed a statistical model for the number of neurons spiking in a neuronal ensemble based on the Conway-Maxwell-binomial distribution. Our aim was to capture correlated activity in a neuronal population without measuring correlation coefficients directly. The model captured correlated activity at very short timescales better than measuring correlation coefficients. We also replicated one of the findings of Churchland et al. (2010) relating to the quenching of neural variability at stimulus onset. We propose a connection between this result and the changes in association captured by our model.
Date of Award19 Aug 2020
Original languageEnglish
Awarding Institution
  • University of Bristol
SupervisorCian O'Donnell (Supervisor) & Michael C Ashby (Supervisor)

Cite this