Definition of Neurophysiological Indices to Describe and Quantify the Cortical Plasticity Induced by Neuro-Rehabilitation

The general objective of the PhD project was to develop a methodology for the definition and analysis of neurophysiological indices able to provide a stable and reliable measure of changes induced by a rehabilitative intervention in the brain activity and organization, with the aim to: i) provide a neurophysiological description of the modifications subtending a functional recovery; ii) allow the evaluation of the effects of rehabilitation treatments in terms of brain reorganization; iii) describe specific properties in the brain general organization to be correlated with the outcome of the intervention, with possible prognostic/decision support value. For this purpose, the research activity was focused on the development of an approach for the extraction of neurophysiological indices from non-invasive estimation of the cerebral activity and connectivity based on electroencephalographic recordings. Brain activity and its changes in time were investigated at three different interconnected levels: spectral analysis, connectivity estimation and graph theory. For each of these, the state of the art methods were evaluated and methodological advancements were proposed on the basis of open problems presented by the nature of the data and by the clinical problem. Experimental data were acquired from 56 stroke patients subjected to a rehabilitative intervention based on Motor Imagery (MI). A subgroup of randomly selected patients were trained in the MI task with the support of Brain Computer Interface. New spectral and functional indices were defined and computed from EEG recorded during the execution of specific tasks (e.g. motor execution), but also from resting state brain activity, to capture both specific and general brain functional modifications. The application of the developed methods allowed to return a proof of the nature, quality and properties of the brain description and quantitative indices that can be derived from data easily recordable from a wide range of patients.