Finding Optimal GPi Target for DBS in Dystonia Through Microelectrode Recording Analysis

Dystonia is a movement disorder that manifests itself by uncontrollable muscle contractions of individuals or multiple muscle groups at one of several body parts ranging from mild to severe symptoms. Deep Brain Stimulation of a specific brain region became a standard procedure for alleviating the symptoms of drug-resistant movement disorders such including primary Dystonia. Even though the clinical efficacy of DBS on hypokinetic and hyperkinetic movement disorders is demonstrated with multiple studies, the therapeutical mechanism of DBS is still a highly debated topic. Target localization of DBS electrodes for acquiring the intended clinical outcome and prevention of DBS-induced adverse effects are two important challenges that need to be handled to optimize the treatment of drug-resistant primary Dystonia. Since the pathological functioning of basal ganglia is considered for the generation of dystonia, the link between pathological neural dynamics of Globus Pallidus Interna (GPi) neurons and primary Dystonia is investigated in the thesis work. For that investigation procedure, we developed a quantitative pipeline. The pipeline extracts neural biomarkers from raw MER recordings from dystonic GPi, analyses these biomarkers with multiple statistical approaches and finally exploit them to classify neurons for multiple classification problems in two separate spatial domain: relative depths and MNI space. The results of this master thesis mark that neural dynamics inside dystonic GPi, previously considered as a homogenous subcortical nucleus, varies with statistical significance, and these variations can be elucidated to locate neurons within GPi. The other important finding is that firing rate and firing regularity-related neural biomarkers are the ones that matter in the case of neural localization. Biomarkers that represent oscillatory and bursting neural dynamics within dystonic GPi are statistically not relevant for neural localization. At the end of the developed pipeline, multiple classification algorithms are trained with cross-validation, and we reached up to 92.5% performance level for some classification routines. From the clinical point of view, the findings of this thesis can contribute to the development of complementary clinical toolboxes that analyze the neural activity in real-time and provide insights in terms of electrode localization to neurosurgeons during DBS surgeries. Additionally, the results can pave the way for describing the varying neural dynamics and relating them to clinical outcomes to define sweet spots inside GPi for dystonic patients and identification of neural biomarkers that can be used as a feedback mechanism for adaptive DBS applications