deep brain stimulation (DBS) is an established method for symptom control in movement disorders such as Parkinson’s disease (PD) and essential tremor (ET). The treatment consists in delivering electrical stimulation in the deep brain via multi-contact electrodes. The potential for improving quality of life is important and the precision of the electrode implantation is crucial. Despite the success of the therapy, mechanisms of action of DBS are still an intense topic of study. One promising approach is to analyze the effect of stimulation relative to the anatomy. In the clinical process, the effect of stimulation on the symptoms is evaluated at different stages, sometimes during surgery (intraoperative) and systematically in the months following implantation (postoperative). This data can be combined with magnetic resonance imaging (MRI) acquired as part of the clinical routine for individuals or groups of patients to create probabilistic stimulation maps (PSMs). The work in this thesis aimed at developing image and data processing workflows to analyze and visualize DBS outcome in the ventro-intermediate nucleus of the thalamus (Vim) and the Zona Incerta (Zi) which are the two typical ET targets. This was done first in a patient-specific manner, and then at a group level using data collected intra- and postoperatively. Computer electrical field modeling was used to estimate the extent of stimulation within brain tissue and to compare intraoperative and postoperative stimulation (Paper I). The simulation method was then applied to clinical data, combining quantitative tremor assessment collected using acceleration sensors during intraoperative testing in the Vim, to create patient-specific stimulation maps based on high resolution objective data (Paper II). Stimulation maps constitute a partial answer to the aim by providing a visual summary of individual patients’ stimulation tests that can be useful in the clinic. In order to summarize several patients’ data, neuroimaging workflows were developed, and a range of non-linear image registration tools were evaluated to create group-specific anatomical templates from MRI using data from 19 patients (Paper III). The normalization workflow was improved and the settings optimized, resulting in the creation of an anatomical atlas defining the outline of 58 structures of the deep brain (Paper IV). The normalization method was transferred to a second group of 77 patients with postoperative stimulation tests in Zi, and combined with patient specific electric field modeling to create PSMs. The influence of different methodological choices such as input data type and voxel clustering method on the resulting maps were investigated (Paper V). The deep brain atlas from the first group was then combined with electric field simulations and the acceleration-sensor-based quantitative tremor assessment. A stimulation atlas based on intraoperative test data in Vim was created (Paper VI) to evaluate of the potential of intraoperative test stimulation for PSM creation.

In conclusion, reproducible pipelines for the generation of group-specific brain templates and high precision PSMs were developed and applied to two different patient groups. The template created for one group resulted in a deep brain atlas defining 58 anatomical structures. The PSMs created for both groups combine group-specific MRI templates, patient-specific electric field simulation and symptom assessment for two common DBS targets that are challenging to locate. The PSMs have the potential to improve the understanding of the mechanisms of action, support clinical planning and aid follow-up by providing visualizations of the therapeutic effect of stimulation in relation to anatomy. In the future, the pipeline will be applied to more patients and diseases. The resulting PSMs will be integrated in a visualization tool for intuitive data exploration and used to predict programming settings.

Djup hjärnstimulering (DBS) är en etablerad metod för behandling av symtom vid rörelsesjukdomar såsom Parkinsons sjukdom (PD) och essentiell tremor (ET). Behandlingen består av elektrisk stimulering av djupa hjärnstrukturer via små elektroder. Stimuleringens effekt på patientens symtom utvärderas kliniskt, ibland under operationen (intraoperativt) och systematiskt under månaderna efter implantationen (postoperativt). Data som samlas in vid dessa utvärderingar kan kombineras med magnetresonans (MR) bilder och simuleringar av elektriska fält runt DBS elektroderna. Detta kan utföras för enskilda patienter eller grupper av patienter, det senare för att skapa statistiska stimuleringskartor. Arbetet i denna avhandling syftar till att utveckla arbetsflöden för bild- och databehandling för att analysera och visualisera DBS-resultat. Två typiska målstrukturer för implantering av DBS elektroder har studerats, ventro-intermediate kärnan (Vim) i talamus och zona incerta (Zi).

Modellering och simulering av elektriska fält användes för att jämföra storleken av stimuleringsvolymen baserad på intraoperativ testelektrod och postoperativ DBS elektrod (Papper I). Simuleringsmetoden tillämpades sedan på kliniska data och kombinerades med kvantitativ förbättring av tremor med hjälp av accelerometrar under intraoperativ testning i Vim, och patientspecifika stimuleringskartor skapades (Papper II). För att sammanfatta flera patienters data utvecklades ett arbetsflöde för normalisering av MR bilder. En rad verktyg för icke-linjär bildregistrering utvärderades för att skapa gruppspecifika anatomiska mallar från MR bilder med hjälp av data från 19 patienter (Papper III). Arbetsflödet för normalisering förbättrades och inställningarna optimerades, vilket resulterade i en anatomisk atlas som definierar konturerna av 58 strukturer i den djupa hjärnan (Papper IV). Normaliseringsmetoden överfördes till en större patientgrupp på 77 patienter med postoperativa stimuleringstester i Zi och kombinerades med patientspecifik modellering av elektriska fält för att skapa stimuleringskartor. Påverkan av metodologiska val, t.ex. typ av indata och klustermetod, på de resulterande kartorna undersöktes i Papper V. Den anatomiska atlasen från Papper III kombinerades sedan med simuleringar av elektriska fält och den kvantitativa tremorförbättningen. Detta resulterade i en stimuleringskarta baserad på intraoperativa testdata i Vim (Papper VI).

Sammanfattningsvis utvecklades arbetsflöden för generering av stimuleringskartor för två patientgrupper med DBS implantat. Stimuleringskartorna kombinerar MR-bilder, patientspecifik simulering av elektriska fält och symtombedömning för två vanliga DBS-mål som är svåra att lokalisera. Stimuleringskartorna har potential att förbättra förståelsen av DBS mekanismerna, stödja klinisk planering och underlätta uppföljning genom att visualisera stimuleringens terapeutiska effekt i förhållande till anatomin.

Electrical stimulation of the deep parts of the brain is the standard answer for patients subject to drug-refractory movement disorders. Collective analysis of data collected during surgeries are crucial in order to provide more systematic planning assistance and understanding the physiological mechanisms of action. To that end, the process of normalizing anatomies captured with Magnetic Resonance imaging across patients is a key component. In this work, we present the optimization of a workflow designed to create group-specific anatomical templates: a group template is refined iteratively using the results of successive non-linear image registrations with refinement steps in the in the basal-ganglia area. All non-linear registrations were executed using the Advanced Normalization Tools (ANTs) and the quality of the normalization was measured using spacial overlap of anatomical structures manually delineated during the planning of the surgery. The parameters of the workflow evaluated were: the use of multiple modalities sequentially or together during each registration to the template, the number of iterations in the template creation and the fine settings of the non-linear registration tool. Using the T1 and white matter attenuated inverse recovery modalities (WAIR) together produced the best results, especially in the center of the brain. The optimal numbers of iterations of the template creation were higher than those from the literature and our previous works. Finally, the setting of the non-linear registration tool that improved results the most was the activation of the registration with the native voxel sizes of images, as opposed to down-sampled version of the images. The normalization process was optimized over our previous study and allowed to obtain the best possible anatomical normalization of this specific group of patient. It will be used to summarize and analyze peri-operative measurements during test stimulation. The aim is that the conclusions obtained from this analysis will be useful for assistance during the planning of new surgeries. © 2021, Springer Nature Switzerland AG.

Electrical stimulation of the deep parts of the brain is the standard answer for patients subject to drug-refractory movement disorders. Collective analysis of data collected during surgeries are crucial in order to provide more systematic planning assistance and understanding the physiological mechanisms of action. To that end, the process of normalizing anatomies captured with Magnetic Resonance imaging across patients is a key component. In this work, we present the optimization of a workflow designed to create group specific anatomical templates: a group template is refined iteratively using the results of successive non-linear image registrations. I norder to improve the results in the basal-ganglia area, the process is refined in this specific volume of interest. All non-linear registrations were executed using the Advanced Normalization Tools (ANTs). The quality of the normalization was measured using the manual delineation of anatomical structures produced during the planning of the surgery and their spacial overlap after trans- formation in the template space by means of Dice coefficient and mean surface distance. The parameters of the workflow evaluated were: the use of multiple modalities sequentially or together during each registration to the template, the number of iterations in the template creation and the fine settings of the non-linear registration tool. Using the T1 and white matter attenuated inverse recovery modalities together produced the best results, especially in the center of the brain. The optimal numbers of iterations of the template creation were higher than those advised in the literature and our previous works. Finally, the setting of the nonlinear registration tool that improved results the most was the activation of the registrationwith the native voxel sizes of images, as opposed to down-sampled version of the images. Theuse of the delineation of the anatomical structures as a mean to measure the quality of the anatomical template of a group of patient allowed to optimize the normalization process and obtain the best possible anatomical normalization of this specific group of patient. The most crucial points were the combination of multiple modalities in order to maximize the quality of information available during image registration and the activation of the registration with native voxel size. The anatomical template of the group will be used to summarize and analyze peri-operative measurements during test stimulation. The aim is that the conclusions obtained from this analysis will be useful for assistance during the planning of new surgeries.

INTRODUCTION

The success of the deep brain stimulation (DBS) therapy relies primarily in the localization of the implanted electrode, implying the need of utmost accuracy in the targeting process. Intraoperative microelectrode recording and stimulation tests are a common procedure before implanting the permanent DBS lead to determine the optimal position with a large therapeutic window where side effects are avoided and the best improvement of the symptoms is achieved. Differences in dimensions and operating modes exist between the exploration and the permanent DBS electrode which might lead to different stimulation fields, even when ideal placement is achieved. The aim of this investigation is to compare the electric field (EF) distribution around the intraoperative and the chronic electrode, assuming that both have exactly the same position.

METHODS

3D models of the intraoperative exploration electrode and the chronically implanted DBS lead 3389 (Medtronic Inc., USA) were developed using COMSOL 5.2 (COMSOL AB, Sweden). Patient-specific MR images were used to determine the conductive medium around the electrode. The exploration electrode and the first DBS contact were set to current and voltage respectively (0.2mA(V) - 3 mA(V) in 0.1 mA(V) steps). The intraoperative model included the grounded guide tube used to introduce the exploration electrode; for the chronic DBS model, the outer boundaries were grounded and the inactive contacts were set to floating potential considering a monopolar configuration. The localization of the exploration and the chronic electrode was set according to the planned trajectory. The EF was visualized and compared in terms of volume and extension using a fixed isocontour of 0.2 V/mm.

RESULTS

The EF distribution simulated for the exploration electrode showed the influence of the parallel trajectory and the grounded guide tube. For an amplitude of e.g. 2 mA/2 V, the EF extension of the intraoperative was 0.6 mm larger than the chronic electrode at the target level; the corresponding difference in volume was 76.1 mm³.

CONCLUSION

Differences in the EF shape between the exploration and the chronic DBS electrode have been observed using patient-specific models. The larger EF extension obtained for the exploration electrode responds to its higher impedance and the use of current controlled stimulation. The presence of EF around the guide tube and the influence of the parallel trajectory require further experimental and clinical evaluation.