One of the most fundamental features of a neural circuit is its connectivity since the single neuron activity is not due only to its intrinsic properties but especially to the direct or indirect influence of other neurons1. It is fundamental to elaborate research strategies aimed at a comprehensive structural description of neuronal interconnections as well as the networks’ elements forming the human connectome. The connectome will significantly increase our understanding of how functional brain states emerge from their underlying structural substrate, and will provide new mechanistic insights into how brain function is affected if this structural substrate is disrupted. The connectome is characterized by three different types of connectivity: structural, functional and effective connectivity. It is evident that the final goal of a connectivity analysis is the reconstruction of the human connectome, thus, the application of statistical measures to the in vivo model in both physiological and pathological states. Since the system under study (i.e. brain areas, cell assemblies) is highly complex, to achieve the purpose described above, it is useful to adopt a reductionist approach. During my PhD work, I focused on a reduced and simplified model, represented by neural networks chronically coupled to Micro Electrodes Arrays (MEAs). Large networks of cortical neurons developing in vitro and chronically coupled to MEAs2 represent a well-established experimental model for studying the neuronal dynamics at the network level3, and for understanding the basic principles of information coding4 learning and memory5. Thus, during my PhD work, I developed and optimized statistical methods to infer functional connectivity from spike train data. In particular, I worked on correlation-based methods: cross-correlation and partial correlation, and information-theory based methods: Transfer Entropy (TE) and Joint Entropy (JE). More in detail, my PhD’s aim has been applying functional connectivity methods to neural networks coupled to high density resolution system, like the 3Brain active pixel sensor array with 4096 electrodes6. To fulfill such an aim, I re-adapted the computational logic operations of the aforementioned connectivity methods. Moreover, I worked on a new method based on the cross-correlogram, able to detect both inhibitory and excitatory links. I called such an algorithm Filtered Normalized Cross-Correlation Histogram (FNCCH). The FNCCH shows a very high precision in detecting both inhibitory and excitatory functional links when applied to our developed in silico model. I worked also on a temporal and pattern extension of the TE algorithm. In this way, I developed a Delayed TE (DTE) and a Delayed High Order TE (DHOTE) version of the TE algorithm. These two extension of the TE algorithm are able to consider different temporal bins at different temporal delays for the pattern recognition with respect to the basic TE. I worked also on algorithm for the JE computation. Starting from the mathematical definition in7, I developed a customized version of JE capable to detect the delay associated to a functional link, together with a dedicated shuffling based thresholding approach. Finally, I embedded all of these connectivity methods into a user-friendly open source software named SPICODYN8. SPICODYN allows the user to perform a complete analysis on data acquired from any acquisition system. I used a standard format for the input data, providing the user with the possibility to perform a complete set of operations on the input data, including: raw data viewing, spike and burst detection and analysis, functional connectivity analysis, graph theory and topological analysis. SPICODYN inherits the backbone structure from TOOLCONNECT, a previously published software that allowed to perform a functional connectivity analysis on spike trains data

Development of statistical and computational methods to estimate functional connectivity and topology in large-scale neuronal assemblies

PASTORE, VITO PAOLO
2018-02-13

Abstract

One of the most fundamental features of a neural circuit is its connectivity since the single neuron activity is not due only to its intrinsic properties but especially to the direct or indirect influence of other neurons1. It is fundamental to elaborate research strategies aimed at a comprehensive structural description of neuronal interconnections as well as the networks’ elements forming the human connectome. The connectome will significantly increase our understanding of how functional brain states emerge from their underlying structural substrate, and will provide new mechanistic insights into how brain function is affected if this structural substrate is disrupted. The connectome is characterized by three different types of connectivity: structural, functional and effective connectivity. It is evident that the final goal of a connectivity analysis is the reconstruction of the human connectome, thus, the application of statistical measures to the in vivo model in both physiological and pathological states. Since the system under study (i.e. brain areas, cell assemblies) is highly complex, to achieve the purpose described above, it is useful to adopt a reductionist approach. During my PhD work, I focused on a reduced and simplified model, represented by neural networks chronically coupled to Micro Electrodes Arrays (MEAs). Large networks of cortical neurons developing in vitro and chronically coupled to MEAs2 represent a well-established experimental model for studying the neuronal dynamics at the network level3, and for understanding the basic principles of information coding4 learning and memory5. Thus, during my PhD work, I developed and optimized statistical methods to infer functional connectivity from spike train data. In particular, I worked on correlation-based methods: cross-correlation and partial correlation, and information-theory based methods: Transfer Entropy (TE) and Joint Entropy (JE). More in detail, my PhD’s aim has been applying functional connectivity methods to neural networks coupled to high density resolution system, like the 3Brain active pixel sensor array with 4096 electrodes6. To fulfill such an aim, I re-adapted the computational logic operations of the aforementioned connectivity methods. Moreover, I worked on a new method based on the cross-correlogram, able to detect both inhibitory and excitatory links. I called such an algorithm Filtered Normalized Cross-Correlation Histogram (FNCCH). The FNCCH shows a very high precision in detecting both inhibitory and excitatory functional links when applied to our developed in silico model. I worked also on a temporal and pattern extension of the TE algorithm. In this way, I developed a Delayed TE (DTE) and a Delayed High Order TE (DHOTE) version of the TE algorithm. These two extension of the TE algorithm are able to consider different temporal bins at different temporal delays for the pattern recognition with respect to the basic TE. I worked also on algorithm for the JE computation. Starting from the mathematical definition in7, I developed a customized version of JE capable to detect the delay associated to a functional link, together with a dedicated shuffling based thresholding approach. Finally, I embedded all of these connectivity methods into a user-friendly open source software named SPICODYN8. SPICODYN allows the user to perform a complete analysis on data acquired from any acquisition system. I used a standard format for the input data, providing the user with the possibility to perform a complete set of operations on the input data, including: raw data viewing, spike and burst detection and analysis, functional connectivity analysis, graph theory and topological analysis. SPICODYN inherits the backbone structure from TOOLCONNECT, a previously published software that allowed to perform a functional connectivity analysis on spike trains data
13-feb-2018
File in questo prodotto:
File Dimensione Formato  
phdunige_3389519.pdf

accesso aperto

Dimensione 7.19 MB
Formato Adobe PDF
7.19 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11567/931946
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact