Connectomic Deep Brain Stimulation

Connectomic Deep Brain Stimulation (DBS) covers this highly efficacious treatment option for movement disorders such as Parkinson’s Disease, Essential Tremor and Dystonia. The book examines its impact on distributed brain networks that span across the human brain in parallel with modern-day neuroimaging concepts and the connectomics of the brain. It asks several questions, including which cortical areas should DBS electrodes be connected in order to generate the highest possible clinical improvement? Which connections should be avoided? Could these connectomic insights be used to better understand the mechanism of action of DBS? How can they be transferred to individual patients, and more.

This book is suitable for neuroscientists, neurologists and functional surgeons studying DBS. It provides practical advice on processing strategies and theoretical background, highlighting and reviewing the current state-of-the-art in connectomic surgery.

• Written to provide a "hands-on" approach for neuroscience graduate students, as well as medical personnel from the fields of neurology and neurosurgery
• Includes preprocessing strategies (such as co-registration, normalization, lead localization, VTA estimation and fiber-tracking approaches)
• Presents references (key articles, books and protocols) for additional detailed study
• Provides data analysis boxes in each chapter to help with data interpretation

• Language: English
• Publisher: Academic Press
• Release date: Sep 10, 2021
• ISBN: 9780128218624

Book preview

Part I

Deep brain stimulation and connectomics: a fruitful marriage?

Chapter 1: Connectomic DBS: An introduction

Andreas Horn; Bassam Al-Fatly; Wolf-Julian Neumann; Clemens Neudorfer    Movement Disorders and Neuromodulation Section, Department of Neurology, Charité—Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Berlin, Germany

Abstract

Treating brain disorders with the aim of modulating brain connections or networks has been a concept since before 1900. However, with the introduction of the human connectome in 2005, a more formal way of analyzing whole-brain connectivity measures was introduced. Shortly after, the concept of connectomic surgery was proposed, harnessing the power of modern neuroimaging methods to derive brain connectivity graphs that may serve as brain stimulation targets. In this chapter, we briefly recap historical developments and introduce the concept of the connectome and predominant imaging methods commonly used to map it. We then transfer the connectome concept to deep brain stimulation and delve into how the concept may transform the way we currently think about the application of this surgical intervention.

Keywords

Deep brain stimulation; Connectivity; Sensorimotor; Associative; Limbic; Cortico-striato-thalamo-cortical circuitopathy; Cognitive; Affective; Networks; Side effects

Connectomic deep brain stimulation: Aim of this book

Deep brain stimulation (DBS) is an established and highly efficacious treatment option for movement disorders such as Parkinson’s Disease, essential tremor, and dystonia, but promising results have been shown in a growing number of brain diseases such as depression, obsessive compulsive disorder, Alzheimer’s Disease, and pain. Currently, DBS is experiencing a paradigm shift away from studying the focal effects of stimulation on the target structure (such as, e.g., the subthalamic nucleus in Parkinson’s Disease) toward studying the impact on distributed brain networks that span across the whole scale of the human brain. This development is paralleled by a development in the field of neuroimaging—where the term connectome was introduced in 2005 and had a tremendous impact on the field. The connectome, i.e., the formal description of parts of the brain and their interconnections is applicable to DBS if the surgical practice is combined with modern-day neuroimaging concepts. This fosters a marriage between connectomics and DBS and allows us to ask the following questions:

•To which cortical areas should DBS electrodes be connected in order to achieve the highest possible clinical improvement?

•Which connections should be avoided in order to reduce side effects?

•Could these connectomic insights be used to better understand the mechanism of action of DBS?

•How can they be transferred to the individual patient undergoing surgery—will this bring forth a more personalized approach of DBS?

•And finally, could these network modulation concepts be transferred to noninvasive treatment options such as transcranial electric stimulation or transcranial magnetic stimulation?

Our book aims to address the current state of these questions and to give an outlook about how the field of connectomic DBS may develop in the future. We envision a target audience consisting of functional surgeons, neurologists, neuroradiologists, and neuroscientists studying DBS. The book will give practical advice on processing strategies and theoretical background where needed. It will highlight and review the current state-of-the-art in connectomic surgery and provide an outlook of what is to come.

We would like to thank you for your interest in our book—but now, let the journey begin. Connectomic DBS, here we come!

Surgical modulation of networks: Where do we come from?

The concept of retuning brain function by modulating brain connections is all but new. The late 19th century was marked by radical advances in our understanding of the brain’s functional anatomic relationships. Lesional studies by Wernicke and Broca, which helped to localize speech faculties in the brain, and electrical stimulation studies of the dog cortex by Fritsch and Hitzig led to a localizationist and associationist understanding of brain function. Dedicated brain areas were presumed to receive input via sensory systems that were modulated and processed in connecting systems and subsequently fed to the motor system. Based on this premise the Swiss psychiatrist, Gottlieb Burckhardt, reasoned that surgical separation of these respective systems may alleviate pathologic mental or psychiatric states. In 1888, he performed the first psychosurgery of the modern era, performing an experimental topectomy on patients with intractable psychiatric diseases. In 1954, Ernest Spiegel and Henry Wycis introduced the concept of ansotomy for the treatment of Parkinsonian tremor,¹ for the first time using a stereotactic device for human brain surgery.² Based on an earlier proposal by Russel Meyers in 1951, the aim was to cut pallidofugal efferent fibers traversing from the internal pallidum to the thalamus within the ansa lenticularis³.a The early success of these procedures sparked the lesional era, which saw a widespread application of ablative procedures; however, many surgeons recognized the invasive and irreversible ramifications of these procedures. Aiming to find reversible alternatives, the teams around Lawrence Pool, Robert Heath, and Jose Delgado in Columbia, Tulane, and Yale introduced the concept Electrical Stimulation of the Brain (ESB) in the 1950sb⁷,⁸⁹—nowadays we call the same procedure DBS.

To reiterate: already in these very beginnings of stereotactic neurosurgery, brain connections were targeted in form of ansotomy, campotomy, and Forel-H-tomy which aimed at disrupting information flow between the pallidum and thalamus (Fig. 1.1). Around the same time, Jean Talairach and Lars Leksell (independently) began lesioning the anterior limb of the internal capsule yet again with the aim of disrupting a network between limbic regions and the prefrontal cortex.¹² Knight began lesioning white matter tracts below the caudate (subcaudate tractotomy) to disrupt the connection between orbitofrontal and limbic regions.¹³ Using DBS, Heath and colleagues deliberately targeted the medial forebrain bundle as the outflow pathway from the septal region into the interpeduncular nuclei of the mesencephalon to modulate the reward system in the brain as early as 1959.⁷, ⁸ Some of these concepts, such as modulation of the ansa lenticularis¹⁴ or medial forebrain bundle¹⁵ are currently being rediscovered as surgical targets by means of advanced neuroimaging methods.

Fig. 1.1

Fig. 1.1 Functional anatomy as envisioned around the turn of the 20th century. (A) Detailed anatomical knowledge which is largely still valid today based on work by Oskar and Cecile Vogt that was translated into clinical practice by Otfrid Foerster as early as 1921 . ¹⁰ (B) Drawing of the dentatorubrothalamic tract with Wernekinck’s decussation by Ludwig Edinger in 1896. ¹¹ C.c. Cortex centralis; C.p.e. Cortex praecentalis; C. th. Centro-thalamic tract; Py. Pyramidal tract; F. th. Fronto-thalamic tract; F. p. c. Fronto-ponto-cerebellar tract; N.c. Caudate nucleus; Put. Putamen; Pal. Pallidum; Th. o. optical thalamus; F. th. Fasciculus thalamicus; F.l. Fasciulus lenticularis; A.I. Ansa lenticularis; C. L. Subthalamic nucleus (Corpus Luys); S. n. Substantia nigra; P.p. Pes pedunculi; F.A. Fasciculus Arnoldi (fronto-ponto cerebellar tract migrating into the anterior limb of the internal capsule); N. R. Red nucleus; N. Da. Nucleus Darkschewitschi; C.q. Corpora quadrigemina; O. Oculus; F.r.s. Fasciculus rubrospinalis; F. d. Fasciculus Darkschewitschi; F.t.s. Fasciculus tectospinalis; F. s. c. Fasciculi spino-cerebellares; C.r. Corpus restiforme; Cbl. Cerebellum; N. De. Nucleus dentatus; Jgr. c. Brachium conjunctivum; N. D. Nucleus Deitersi; N. B. Nucleus Bechterewi; V. Vestibularis; C. cochlearis; G. p. Griseum pontis; Br. p. Brachium cerebelli ad pontem.

It may not be a coincidence that some of the early concepts (such as ansotomy) are not widely known by DBS researchers of the present day.⁴ The invention of magnetic resonance imaging in 1973 by Paul Lauterbur and Mansfield¹⁶, ¹⁷ introduced a powerful tool to study the brain’s anatomy without opening the skull. Lauterbur and Mansfield definitely deserved the Nobel Prize for their discovery, but—hypothetically—a potential side-effect of the increased use of MRI may have heralded a decline of enthusiasm for invasive brain anatomy. Could the concept of direct targeting of deep brain structures as they are shown on MRI have led to a diminished desire to collaborate with anatomists when operating on the brain? If one could see deep structures in the brain specific to the individual patient, why bother consulting histological sections derived from other brains. In contrast, in earlier days of stereotactic surgery, it was essential to have an anatomist as part of the surgical team. For instance, the emergence of stereotactic surgery in Europe was likely as much facilitated by Rolf Hassler (a neuroanatomist) as by Traugott Riechert and Fritz Mundinger (neurosurgeons). Similarly, early work by Oskar and Cecile Vogt (neuroanatomists) motivated translational work performed by the surgical neurologist Otfrid Foerster in 1922 (Fig. 1.1).

So, was there potentially a time and generation (between 1990 and 2020) that may have relied slightly too much on MRI and partly forgotten about anatomy? Is there a modern trend to rediscover surgical white-matter targets described before? In other words, is the concept of connectomic deep brain stimulation anything new, at all? If so, we should not fall into the same trap when using the MRI to assess brain connectivity.

In this book, we aim to show that there are indeed new developments that go beyond what was done by pioneering surgeons from the 1950s onward while honoring and building upon their efforts. However, we do make the point that collaboration with anatomists and the use of anatomical references above and beyond MRI remain critical (Chapter 14). While the book will focus on MRI (Chapters 2–9) and the use of advanced neuroimaging methods to analyze brain connectivity (Chapters 10–13), we will touch upon limitations and potential alternatives (Chapters 14 and 16) throughout its course. If the MRI is to become our main method, we should still ask each time we address a question, whether it is the best tool to answer it—or whether we could use other resources to augment the insights we may gain.

The connectome: A new era of measuring brain connectivity?

The title of our book reads Connectomic Deep Brain Stimulation. But what is the difference between brain connectivity and the connectome? While the two are sometimes treated as two sides of the same coin in contemporary work, the connectome has a proper definition and to formally introduce it, we will take a short detour before getting back to DBS. A more formal introduction to the connectomic revolution can be found in Chapter 2.

In 2005, Olaf Sporns developed the concept of the connectome with late Rolf Kötter alluding to the human genome.¹⁸ The idea behind the concept, most authoritatively outlined in his seminal book Networks of the Brain,¹⁹ involves parcellating the brain into distinct regions and formally describing a wiring diagram between those regions. In this framework, two ideas are crucial: First, the degree of parcellation can range from single neurons to large brain regions, resulting in micro-, meso-, or macroscale descriptions of the connectome, only the latter of which is truly accessible by MRI research. Second, when describing wiring diagrams mathematically, we engage in graph theory. Mathematical graphs consist of nodes and edges connecting them. Most graphs can be represented by adjacency or connectivity matrices with rows and columns representing nodes and matrix entries representing strength of the connection (i.e., the strength of each edge). If matrices are symmetric, we speak of undirected graphs (edges are often represented by sticks between balls that represent nodes, Fig. 1.2A). If they are asymmetric, resulting graphs are considered directed (where instead of sticks we often draw arrows to show directionality of connections, Fig. 1.2B).

Fig. 1.2

Fig. 1.2 Different representations of graphs. (A) Diffusion-MRI tractography–based structural connections between nodes that were activated in a task-fMRI experiment. The left panel features reconstructions of anatomical tracts as estimated by probabilistic tractography, while the right panel displays a schematic representation of the underlying graph. (B) Directed graph estimated on brain-wide voxel-wise psychophysiological interactions calculated on task-fMRI data in a spatiovisual attention task. (C) Common graph representation featuring nodes (numbered circles) and their connections (edges), some of them highlighted in colors. Panel (A) modified from Horn A. DTI-Fibre-Tracking funktioneller Netzwerke der Sprach- und Musik-Verarbeitung. October 2012:1–69.

Needless to say, the study of brain connectivity predates the first description of the connectome by Sporns and Hagmann in 2005 by centuries (Fig. 1.1). However, when it was introduced, the connectome concept conquered the field of neuroimaging and neuroscience by storm. In the preface to his second book on the connectome,²⁰ Olaf Sporns wrote:

When I googled the term ‘connectome’ [… in 2005] I remember getting around 10 hits, none of them relevant to the brain. In fact, some of them were oddly irrelevant – I recall finding ‘connect-to-me’ (a dating site, I believe) and ‘connect-home’ among the search results. As of April 2012, the same Google search returns nearly a quarter million hits. What happened?

What made the concept of connectomic research so powerful? First, the mathematical representation of graphs allowed to calculate with them and to apply statistical measures that allowed formal comparisons. Second, the idea of describing properties of whole-brain connectivity was compelling to our field. In comparison to investigating connections between specific regions, we were now able to assess and integrate, to compare, and condense whole-brain networks in a formal way. By constructing a formal graph, we can describe specific abstract properties such as centrality (how strongly connected specific nodes in the network are), groups of nodes as motifs and specific nodes as hubs, clustering coefficients, shortest path lengths (the shortest way to hop from node A to B) or network-wide properties such as global efficiency (the average inverse shortest path length across the network). Most of these measures have been around in different fields for ages. For instance, an adaptation of a smart way to measure centrality by use of eigenvectors made the first version of the Google search engine more efficient than their competitors (e.g., AltaVista and Excite) and was part of Larry Page’s Ph.D. project.²¹

New technologies to measure brain connectivity without opening the skull

In order to widely apply brain connectomics to living humans, technology to measure brain connectivity noninvasively is essential. While multiple methods for measuring brain connectivity exist (such as electroencephalography, magnetoencephalography, local field potential, or electrocorticography recordings), two of them play a prominent role in the present context. First, functional magnetic resonance imaging (fMRI) was discovered by Jack Belliveau in 1991²² and extended to be fully noninvasive by Kenneth Kwong a year later.²³ Both endeavors were led by Bruce Rosen’s team at the Martinos Center in Boston and entered science history. fMRI is commonly employed to study brain function in subjects that perform tasks in the scanner by estimating the blood-oxygen-level–dependent (BOLD) signal. However, seminal work by Bharat Biswal demonstrated that spontaneous fluctuations of the BOLD-signal recorded when subjects were at rest (rs-fMRI) could be used to estimate functional brain connectivity (i.e., functional couplings between different brain regions; see Fig. 1 in Chapter 10).²⁴ A more in-depth introduction of resting-state functional MRI is found in Chapter 10.

Second, after Denis Le Bihan had introduced diffusion-MRI in 1986,²⁵ Peter Basser’s breakthrough in the same group introduced the concept of the diffusion tensor in 1994.²⁶ While diffusion tensorsc are largely avoided nowadays (they inherently lack the capability of resolving fiber crossings in the brain), similar enhanced models are widely used and the fundamental concept of applying diffusion-imaging (dMRI) to estimate structural connectivity has remained unchanged: Multiple diffusion-weighted acquisitions with gradients along different directions are acquired and aggregated using voxel-wise models that account for the probability of water to diffuse in each direction. This is useful since diffusion is larger along fiber bundles than perpendicular to them (which constitutes hindered diffusion). By measuring the degree of anisotropic diffusion across voxels, tractography algorithms are then able to make inferences about diffusion directionality. This information is ultimately used to calculate streamlines that estimate structural connectivity. A more in-depth introduction of diffusion-weighted imaging–based tractography can be found in Chapter 11.

Despite the many inherent limitations and derived nature of brain connectivity estimated by fMRI and dMRI, these methods constitute powerful tools to map the functional and structural connectome in the living human brain (Fig. 1.3).

Fig. 1.3

Fig. 1.3 Magnetic Resonance Imaging (MRI) sequences that estimate brain connectivity. Most applications of functional MRI (left) measure the blood-oxygen-level–dependent (BOLD) signal which indicates the relationship between oxygenated and deoxygenated hemoglobin. Spontaneous fluctuations (acquired while the subject is at rest in the scanner) between some brain regions are correlated, speaking of ultra-slow (< 0.1 Hz) coupling of brain activity across regions. Series of diffusion-weighted imaging (right) with different gradient directions can be aggregated to estimate probabilities of water diffusion in various directions at each voxel in the brain. This data can be used to fit orientation distribution functions (ODF) which in sequence serve as guidance for tractography experiments. Starting from a seed-region, a tracking-algorithm will move from voxel to voxel following main or secondary orientation directions of ODFs. By those means, streamlines will be created that estimate structural brain connectivity. If noise is added to the ODFs, we speak of probabilistic tractography, without the added noise of deterministic tractography. Both concepts have advantages and disadvantages in terms of sensitivity (false negative connections) and specificity (false positive connections). ²⁷ Please note that both fMRI and dMRI are highly derived methods that have serious limitations in their accuracy to map brain connectivity. Their main elegance and power should be seen in the ability to noninvasively measure connectivity in the living human brain.

From brain circuitopathies to connectomic surgery

The above paragraphs highlighted the old and successful concept of altering brain connectivity by functional neurosurgery, the framework of the connectome, and noninvasive methods to estimate brain connectivity in living humans. Although the field had been thinking about combinations of these concepts much earlier, it was not until 2012 that all three were formally combined when Jamie Henderson proposed the idea of connectomic surgery in 2012.²⁸ Multiple groups had already applied diffusion-imaging–based tractography or results from fMRI in combination with deep brain stimulation before 2012.¹⁵, ²⁹–³¹, However, in Henderson’s article, a specific marriage of the fields of connectome research and functional neurosurgery was proposed. Here was a view that postulated the translation from the flourishing field of connectomics into clinical practice, ultimately by finding ways to modulate brain networks in a tailored, deliberate fashion. In his article, Henderson used words we knew from the connectome literature, such as nodes and hubs, and proclaimed that network surgery could be a solution to modulate distributed networks in the minimally conscious state:

Many of the subcortical structures that have been targeted for the surgical treatment of movement disorders show dense, widespread connectivity, and may represent this kind of integrative hub within an overall less densely connected network. Disruption of integration or segregation within these hubs could lead to abnormalities in connectivity, creating pathological brain states.

Could it be possible that deep nuclei (as the ones targeted by DBS) would exactly form the type of integrator-hubs that were proposed by Rubinov, Sporns, and other proponents of the connectome field? Most certainly. The subthalamic nucleus (STN), the most targeted brain area in functional neurosurgery, receives widespread axonal afferents from the entire frontal cortex.³² The striato-pallidofugal system (the main axis of the basal-ganglia rate model, see Chapter 4) is anatomically constructed to compress information from widespread input at each stage of the loop (Figs. 1.4 and 1.5). Namely, striatal zones are spatially compressed when entering the external pallidum and further so when entering its internal segment (GPi).³⁵, ³⁸ Furthermore, the basal ganglia compress information in the form of a dimensionality reduction system—as explored in depth by Izhar Bar-Gad and Hagai Bergman in 2003.³⁶ Hence, it is not a coincidence that surgeons chose to modulate the brain here. The concept of the basal ganglia as an information funnel has been described for decades (³⁴, ³⁵ Fig. 1.4) and has been excellently captured in oral communication with Pierre Pollak:

Fig. 1.4

Fig. 1.4 Basal ganglia as an information funnel. Information openly available at the level of the cortex is spatially compressed to smaller brain regions. Moreover, when seen from an information theory perspective, the basal ganglia compress the information content itself. (A) Seminal work led by Parent and Percheron shows the striatopallidofugal system to integrate data from various cortical regions. Parent and colleagues reviewed the parallel loops (left) and funnel (right) frameworks in 1995. ³³ The figure taken from Percheron and colleagues ³⁴ shows how pallidal neurons have disc-like receptive terminals oriented orthogonally to the striatopallidofugal fibers. As discs are of the same size at each level along the axis, neurons at the end of the pallidal funnel integrate information from increasingly larger striatal zones. (B) Coronal polarized light imaging section of a vervet monkey. ³⁵ Image courtesy by Dr. Markus Axer and Prof. Karl Zilles, Forschungszentrum Jülich, INM-1. In the transition zone between striatum, external and internal pallidum, striatopallidofugal neurons are rewired to compress data similar to layers of artificial neural networks. ³⁶ W: Wilson’s pencils, Put: Putamen, IC: Internal Capsule, c: Edinger’s comb system, ZI: Zona incerta, Thal: Thalamus, LE: Lamina externa, LI: Lamina interna, GPi & PaM: internal pallidum, GPe & PaL: external pallidum, Cd: Caudate.

Fig. 1.5

Fig. 1.5 (A) The basal ganglia as an information-compression system. When targeting integrator hubs (in which spatial compression is high), we are able to modulate large distributed brain networks with single electrodes. Established DBS targets such as the STN and GPi serve as a funnel that integrates and spatially compresses information stemming from increasingly larger brain regions. Similarly, they project back to larger regions, spreading out their projection fields across distributed brain networks. In graph-theoretical language, they serve as integrator-hubs, in information-theoretical language as dimensionality reduction systems. Interestingly, the same brain targets have beneficial effects on a wide array of brain diseases. (B) Effect of stimulation in the VIM across different frequencies to suppress tremor. When applying frequencies between ~ 100 and 2000 Hz (log scale), tremor is suppressed at low amplitudes. Frequencies below or above this rate are less efficient requiring current increases to effectively control symptoms. Panel (B) adapted from Benabid AL, Pollak P, Gervason C, et al. Long-term suppression of tremor by chronic stimulation of the ventral intermediate thalamic nucleus. Lancet 1991;337(8738):403–406. https://doi.org/10.1016/0140-6736(91)91175-t.

"One point to add, which is not generally discussed in papers. It’s the deep of deep brain stimulation. Why does it work, when it’s deep and it doesn’t work well when we stimulate cortically or superficial? And this is true with many targets and new indications. I think because as we know about functioning of brain networks: They are anatomically like a funnel. So, the deeper you are, the greater network you can impact." Pierre Pollak in 2020—https://doi.org/10.6084/m9.figshare.12611939.v1

Modulating brain networks at the level of their integrator-hubs

Following the concept of the basal ganglia as an information funnel, we could be excited but also disappointed. Excited because we are able to modulate large brain networks with single electrodes if we place them at sites of integrator-hubs (the narrow end of the funnel, Fig. 1.5). But disappointed as our neuromodulatory effect would come at the expense of specificity. Along these lines, it is surprising that STN-DBS has clear effects on symptoms in PD,³⁸ focal dystonic syndromes such as cervical dystonia³⁹ or Meige syndrome,⁴⁰ obsessive–compulsive disorder (OCD)⁴¹ and Tourette’s Syndrome.⁴² In addition, the internal pallidum is also a comparably efficient target for PD,⁴³, ⁴⁴ primary dystonia,⁴⁵ Tourette’s Syndrome,⁴⁶ or Lesch–Nyhan syndrome.⁴⁷ Figuratively speaking, it seems we can enter different syndromes into the funnel and always receive a beneficial effect (Fig. 1.5).

In this context, it is crucial to note the existence of parallel functional circuits (Fig. 1.6). Similar to the thalamus and striatum (where functional zones have been mapped in much greater detail), both STN and GPi are known to have a spectrum of functional subzones⁵¹, ⁵² which can partly explain their effects on symptoms and diseases of such variety. Different anatomical zones of the STN relate to different brain functions, and this knowledge is deliberately exploited by surgeons. For instance, in OCD, surgeons target the anteromedial STN while the classical target for movement disorders is located in the posterolateral aspect. However, when projecting the frontal cortex to the subthalamic nucleus, distances of multiple centimeters will be miniaturized to single millimeters. Stated differently, the peak receptive field relevant for tremor, bradykinesia, tics, obsessive or affective symptoms may each reside a mere millimeter away from each other (the total length of the STN is but a centimeter). Does it seem likely that surgeons will always hit the exact correct millimeter within the STN? Despite their marvelous skill, discipline, and rigor, they will not.⁵³ What about two motor disorders (such as PD and dystonia)? It is surprising that the same part of the nucleus is such an effective target for both diseases. What would happen in a dystonia patient if their receptive field relevant for treating obsessive–compulsive symptoms is co-stimulated? Reports of psychiatric side-effects in STN-DBS when treating PD patients are frequent—but not frequent enough and rare in dystonia. Not every anteromedially located electrode will provoke them. Hence, while some differences in effects may be explained by slight variations in stimulation location within the narrow end of the basal ganglia funnel which leads to an adjustment of the predominant network stimulated, it seems unlikely that this will explain the full extent of this matter. The reason that DBS works well and robustly in many diseases—even when targeting the same brain area—is in its mechanism of action, which disrupts pathological information flow. In other words, pathological information flow needs to be present for DBS—in its current form—to work. We will devote the rest of the chapter to address this