One of the major aims of Systems Neuroscience is to understand how the nervous system transforms sensory inputs into appropriate motor reactions. In very simple cases sensory neurons are immediately coupled to motoneurons and the entire transformation becomes a simple reflex, in which a noxious signal is immediately transformed into an escape reaction. However, in the most complex behaviours, the nervous system seems to analyse in detail the sensory inputs and is performing some kind of information processing (IP). IP takes place at many different levels of the nervous system: from the peripheral nervous system, where sensory stimuli are detected and converted into electrical pulses, to the central nervous system, where features of sensory stimuli are extracted, perception takes place and actions and motions are coordinated. Moreover, understanding the basic computational properties of the nervous system, besides being at the core of Neuroscience, also arouses great interest even in the field of Neuroengineering and in the field of Computer Science. In fact, being able to decode the neural activity can lead to the development of a new generation of neuroprosthetic devices aimed, for example, at restoring motor functions in severely paralysed patients (Chapin, 2004). On the other side, the development of Artificial Neural Networks (ANNs) (Marr, 1982; Rumelhart & McClelland, 1988; Herz et al., 1981; Hopfield, 1982; Minsky & Papert, 1988) has already proved that the study of biological neural networks may lead to the development and to the design of new computing algorithms and devices. All nervous systems are based on the same elements, the neurons, which are computing devices which, compared to silicon components, are much slower and much less reliable. How are nervous systems of all living species able to survive being based on slow and poorly reliable components? This obvious and naïve question is equivalent to characterizing IP in a more quantitative way. In order to study IP and to capture the basic computational properties of the nervous system, two major questions seem to arise. Firstly, which is the fundamental unit of information processing: 2 single neurons or neuronal ensembles? Secondly, how is information encoded in the neuronal firing? These questions - in my view - summarize the problem of the neural code. The subject of my PhD research was to study information processing in dissociated neuronal cultures of rat hippocampal neurons. These cultures, with random connections, provide a more general view of neuronal networks and assemblies, not depending on the circuitry of a neuronal network in vivo, and allow a more detailed and careful experimental investigation. In order to record the activity of a large ensemble of neurons, these neurons were cultured on multielectrode arrays (MEAs) and multi-site stimulation was used to activate different neurons and pathways of the network. In this way, it was possible to vary the properties of the stimulus applied under a controlled extracellular environment. Given this experimental system, my investigation had two major approaches. On one side, I focused my studies on the problem of the neural code, where I studied in particular information processing at the single neuron level and at an ensemble level, investigating also putative neural coding mechanisms. On the other side, I tried to explore the possibility of using biological neurons as computing elements in a task commonly solved by conventional silicon devices: image processing and pattern recognition. The results reported in the first two chapters of my thesis have been published in two separate articles. The third chapter of my thesis represents an article in preparation.
|Autori interni:||Bonifazi, Paolo|
|Titolo:||Information processing in dissociated neuronal cultures of rat hippocampal neurons|
|Relatore/i interni:||Torre, Vincent|
|Data di pubblicazione:||15-dic-2005|
|Appare nelle tipologie:||8.1 PhD thesis|