|Ph.D Student||Dubin Uri|
|Subject||Multi-Dimensional Data Analysis in Cortical Networks|
|Department||Department of Medicine||Supervisor||Professor Jackie Schiller|
|Full Thesis text|
Understanding brain functionality is a great challenge. It not only fulfils our curiosity but promises better treatment of diseases, improved life quality and improvement of the world around us.
The brain is very complex and its complexity is present on several levels. It starts on the molecular and biochemical level, where we are trying to understand interaction chains between molecular mechanisms, find out how and when proteins are expressed, and how they affect other such components. The cellular level is another tier of complexity. Here we see different neurons with different characteristics, dendrite structures and electro-chemical properties of intracellular mechanisms. Internal computations, short term and long term memories are related to a single neuron and still are not understood. The next higher complexity level comes from myriad of connections between different neurons and their electrochemical interactions. Recent technology affords monitoring of many neuron units, using advanced microscope and electrode array methods. The final ultimate complexity level is present in the observed animal behavior. Under various experiments and stimulations, we observe animal responses and try to infer the underlying activity of the cells and networks.
In this thesis we have developed tools and performed analyses of single neuron and multi neuron networks on anaesthetized and awake behaving animal in-vivo. The enormous-sized data were collected from animals during multiple experiments that at times lasted several months. This vast amount of data required massive preprocessing before we can browse it. Then, using the tools we created, the most important information is extracted and concisely represented in a smart browser with a flexible GUI that we created. We are now able to observe and analyze the imaging, electrophysiology and behavioral data and perform on them multiple calculations. Especially, we were interested in discovering the spatio-temporal patterns in the observed neural networks. These multi-dimensional patterns can suggest about data representation and about the functionality of the network.
In a first project, we developed tools to answer questions of how texture is represented in the somato-sensory cortex and how texture discrimination is performed by the barrel whisker system. A second project dealt with the network dynamics and functional reorganization during initiation of epileptic seizures. We have developed dedicated algorithms to de-convolve the continuous calcium signals into event trains. In addition, we have developed statistical tools to analyze the network functional structure based on synchrony of the neurons.
The last project concentrated on understanding how movement is represented in the network of primary motor cortex of mice. We used in-vivo two photon calcium imaging during a hand reach task to acquire neuron and behavioral data. The acquired data is complex and high dimensional and require sophisticated analysis tools. We have developed a dedicated, complex software to analyze the vast amount of the calcium imaging data as well as the behavioral data.
In this work, we try to understand the neuron activity patterns and connections that emerge between the high level animal behavior and the cellular-level events of neural networks.