Standardizing indirect targeting and building electrophysiological maps for deep brain stimulation surgery after accounting for brain shift
Pallavaram Srinivasan, Srivatsan
Chronic Deep Brain Stimulation (DBS) has been a rapidly evolving area of neurotherapeutics since its initial introduction for the treatment of Parkinson’s disease and essential tremor in the 1990s. In the recent past, there has been active research to improve the outcome of the procedure as well as to make it more accessible to patients. This dissertation is broadly categorized into two parts. The first is motivated by a lack of standardization in the localization of popular anatomical landmarks used to indirectly localize as well as communicate stereotactic targets. Inter-surgeon variability in manually selecting these landmarks and its impact on target localization is shown to be substantial. A method based on non-rigid image registration is used for automatic prediction of the landmarks and its accuracy is shown to be sub-millimetric in both clinical and non-clinical settings. The second part is motivated by shortcomings and inaccuracies in existing methods to populate statistical atlases of electrophysiological data acquired intra-operatively during DBS surgeries. A Gaussian smoothed spherical shell kernel is proposed as an improvement over an existing method to model stimulation response in order to build accurate statistical maps. The effect of intra-operative brain shift on the creation of electrophysiological atlases is investigated and shown to be substantial. An approach to build low-shift atlases is proposed and statistical maps of stimulation response built using data from such an atlas are shown to correlate strongly with a statistical ground truth as well as with an anatomical atlas. Finally, in a preliminary study, it is shown that statistical maps of adverse effects combined with statistical maps of efficacious stimulation response could be clinically useful for post-operative programming assistance in DBS.