Characterization of the glial scar tissue in a murine model of spinal cord compression, with focus on hyaluronan

Spinal cord injury (SCI) is a devastating neurological disorder that affects thousands of people each year. Although in recent decades significant progress has been made in relation to understanding the molecular and cellular events underlying the nervous damage, spinal cord injury is still a highly disabling condition for which there is no curative therapy. People affected by spinal cord injuries manifested dysfunction or loss, temporary or permanent, of motor, sensory and / or autonomic functions depending on the spinal lesion damaged. Currently, the incidence rate of this type of injury is approximately 15-40 cases per million people worldwide. At the origin of these lesions are: road accidents, falls, interpersonal violence and the practice of sports. In this work we placed the hypothesis that HA is one of the component of the scar tissue formed after a compressive SCI, that it is likely synthetised by the perilesional glial cells and that it might support the permeation of the glial scar during the late phase of SCI. Nowadays, much focus is drawn on the recovery of CNS function, made impossible after SCI due to the high content of sulfated proteoglycans in the extracellular matrix. Counterbalancing the ratio between these proteoglycans and hyaluronic acid could be one of the experimental therapy to re-permeate the glial scar tissue formed after SCI, making possible axonal regrowth and functional recovery. Therefore, we established a model of spinal cord compression in mice and studied the glial scar tissue, particularly through the characterization of the expression of enzymes related to the metabolism of HA and the subsequent concentration thereof at different distances of the lesion epicenter. Our results show that the lesion induced in mice shows results similar to those produced in human lesions, in terms of histologic similarities and behavioral results. but these animals demonstrate an impressive spontaneous reorganization mechanism of the spinal cord tissue that occurs after injury and allows for partial recovery of the functions of the CNS. As regards the study of the glial scar, changes were recorded at the level of mRNA expression of enzymes metabolizing HA i.e., after injury there was a decreased expression of HA synthases 1-2 (HAS 1-2) and an increase of the expression HAS3 synthase mRNA, as well as the enzymes responsible for the HA catabolism, HYAL 1-2. But the amount of HA measured through the ELISA test was found unchanged after injury, it is not possible to explain this fact only with the change of expression of enzymes. At two weeks and in response to SCI, we found synthesized HA by reactive astrocytes and probably by others like microglial cells as it was advanced by the HA/GFAP+ and HA/IBA1+ cells co-location.

Techniques for batch reinforcement learning in robotics

This thesis addresses the Batch Reinforcement Learning methods in Robotics. This sub-class of Reinforcement Learning has shown promising results and has been the focus of recent research. Three contributions are proposed that aim to extend the state-of-art methods allowing for a faster and more stable learning process, such as required for learning in Robotics. The Q-learning update-rule is widely applied, since it allows to learn without the presence of a model of the environment. However, this update-rule is transition-based and does not take advantage of the underlying episodic structure of collected batch of interactions. The Q-Batch update-rule is proposed in this thesis, to process experiencies along the trajectories collected in the interaction phase. This allows a faster propagation of obtained rewards and penalties, resulting in faster and more robust learning. Non-parametric function approximations are explored, such as Gaussian Processes. This type of approximators allows to encode prior knowledge about the latent function, in the form of kernels, providing a higher level of exibility and accuracy. The application of Gaussian Processes in Batch Reinforcement Learning presented a higher performance in learning tasks than other function approximations used in the literature. Lastly, in order to extract more information from the experiences collected by the agent, model-learning techniques are incorporated to learn the system dynamics. In this way, it is possible to augment the set of collected experiences with experiences generated through planning using the learned models. Experiments were carried out mainly in simulation, with some tests carried out in a physical robotic platform. The obtained results show that the proposed approaches are able to outperform the classical Fitted Q Iteration.