Special issue on Computational Methods and Modeling in Psychiatry
Personalized transcranial magnetic stimulation in psychiatry Luca Cocchi1, Andrew Zalesky2,3 1
QIMR Berghofer Medical Research Institute, Brisbane, Australia. Melbourne Neuropsychiatry Centre, University of Melbourne, Melbourne, Australia. 3 Department of Biomedical Engineering, University of Melbourne, Melbourne, Australia. 2
Corresponding Author: Luca Cocchi, QIMR Berghofer Medical Research Institute, Brisbane, Australia. Email:
[email protected]. Phone: +61738453008 Running head: Personalised TMS in psychiatry Number of figures: 5 Number of tables: zero Supplementary material: zero
Key words: TMS, Brain stimulation, Connectivity, Brain networks, Personalised medicine, Biotypes
Cocchi & Zalesky
1
Abstract Transcranial magnetic stimulation (TMS) is a non-invasive brain stimulation technique that allows for modulating the activity of local neural populations and related neural networks. TMS is touted as a viable intervention to normalize brain activity and alleviate some psychiatric symptoms. However, TMS interventions are known to be only moderately reliable and the efficacy of such therapies remains to be proven for psychiatric disorders other than depression. Here we review new opportunities to personalize TMS interventions using neuroimaging and computational modeling, aiming to optimize treatment to suit particular individuals and clinical subgroups. Specifically, we consider the prospect of improving the efficacy of existing TMS interventions by parsing broad diagnostic categories into biologically and clinically homogenous biotypes. Biotypes can provide distinct treatment targets for optimized TMS interventions. We further discuss the utility of computational models in refining TMS personalization and efficiently establishing optimal cortical targets for distinct biotypes. Personalizing cortical stimulation targets, treatment frequencies and intensities can improve the therapeutic efficacy of TMS and potentially establish non-invasive brain stimulation as a viable treatment for psychiatric symptoms.
Cocchi & Zalesky
2
Introduction Psychiatric disorders are characterized by distinct behavioral symptoms and abnormalities in brain function and structure. These symptoms can be mitigated with a number of behavioral and pharmacological interventions (1, 2). However, existing interventions only partially offset the burden of disease and a number of individuals remain clinically symptomatic after treatment (3, 4). New therapeutic avenues that may complement existing pharmacological and psychological therapies are therefore required. Transcranial magnetic stimulation (TMS) is a non-invasive brain stimulation technique that offers the potential to successfully complement existing behavioral and pharmacological treatments for, at least some, psychiatric disorders (5). TMS is a widely used technique for non-invasively modulating neural activity, via the application of a rapidly changing magnetic field over the scalp. Depending on the stimulation protocol, TMS can either increase or inhibit local neural activity, consistent with the processes of long-term potentiation and depression, respectively (6). In principle, TMS can therefore be used to balance altered local neural activity and restore related complex changes in brain networks activity underpinning psychiatric symptoms (7). Unfortunately, the impact of existing TMS interventions on both brain activity and symptoms is highly variable (for a review see (7)). While some individuals respond positively to treatment, with a commensurate normalization of brain activity, other individuals with similar clinical profiles often show no response to TMS intervention (8). Can this heterogeneity in treatment efficacy be improved through personalizing interventions to suit an individual’s sex, age, clinical profile, brain anatomy, connectivity of the simulated region or other physiological indicators? Currently, cortical stimulation targets, stimulation intensities, burst intervals and intervention frequencies have been established largely based on trial-and-error as well as anecdotal evidence supporting choices and parameters that have been effective in a majority of individuals. Personalizing interventions according to neural mechanisms of action can potentially benefit the many individuals for which current TMS interventions have proven to be ineffective. Here, we consider new opportunities to improve the reliability of TMS interventions in psychiatry by taking advantage of emerging knowledge about the impact of TMS on brain regions that are distant from the local stimulation site. To understand the neural mechanisms of action underlying TMS, in addition to the local effects of TMS evident in cortical neuropil within the stimulation vicinity (9-11), it is crucial to account for downstream effects that may be distant to the stimulation site (12-14). Advances in the field of neural connectomics (15) herald new opportunities to take into account heterogeneity in brain connectivity when planning TMS interventions.
Cocchi & Zalesky
3
In this review, we consider two distinct opportunities to personalize TMS interventions that have been enabled through developments in the field of connectomics and computational neuroscience. First, we consider the prospect of biotyping individuals according to brain connectivity and behavioral symptoms, with the goal of delineating unique interventions for each specific biotype. Secondly, we consider recently developed computational models of brain networks that aim to predict downstream effects of local cortical stimulation. These models can be used to evaluate the efficacy of an exhaustive number of stimulation targets in silico.
TMS TMS is a validated non-invasive brain stimulation technique for modulating the activity of neurons within a region of the cortex (16-18). TMS is based on the principle of electromagnetic induction and delivers a strong, but short-lived, magnetic field that induces a perpendicular electric field (19). The resulting electrical currents can subsequently depolarize neuronal axons (Figure 1A). The local effects of TMS are typically studied in the motor system because motor evoked potential (MEP) provides a direct behavioral correlate that can be easily measured (20-24). An increased motor response (motor evoked potential, MEP) in the targeted muscle following TMS is generally considered a proxy of local neural excitation, while a reduction of the MEP is thought to indicate inhibition (17, 21). However, the effect of TMS on the targeted neuropil is not necessarily a categorical effect and may be better quantified along a response spectrum. Furthermore, the MEP represents a relatively coarse measure of complex neural changes. These considerations are critical, but often overlooked when using the MEP as a proxy of TMS-induced changes in neural activity. More generally, given that gyral geometry and tissue conductivity can differ between the motor cortices and other regions, using the MEP to guide selection of TMS parameters in regions distant from the motor corticies is challenging. Neuroimaging indices such as electroencephalography (EEG) signal power may provide more principled alternatives. A typical TMS system is shown in Figure 1B, including a figure-of-eight stimulation coil (Figure 1C).
Cocchi & Zalesky
4
Figure 1. A. Representation of current flow in a standard figure-of-eight TMS coil. Electrical current in the stimulation coil is used to generate a magnetic field. The magnetic field induces secondary currents (eddy currents) within cortical neuropil. Eddy currents can modify neural activity (25), resulting in effects that are consistent with long-term potentiation and depression. B. Illustration of a typical TMS system, including coil (positioned on scalp), pulse generator and neuro-navigation system. C. TMS is delivered via a stimulation coil that is positioned close to the scalp of the individual receiving treatment. Typically, the individual remains in a seated position throughout the stimulation. The operator navigates the TMS coil to the desired cortical target using scalp landmarks or with guidance from a neuro-navigation system (as depicted in Panel B). The intensity of the stimulation is increased until a motor cortex threshold is reached. This threshold is determined by stimulating (single-pulse) the primary motor cortex and measuring the resulting MEP in the contralateral hand muscle of the individual using electromyography (EMG). Although the individual may sense a tapping on the head and other muscle-related sensations that may cause mild discomfort, the treatment is painless and considered safe as long as the appropriate safety guidelines are followed (26). TMS interventions for psychiatric disorders such as depression and obsessive-compulsive disorders are typically repeated daily for a period of 3-4 weeks.
In addition to acute effects, repetitive TMS (rTMS) can change local neural activity for a period that outlasts the duration of the stimulation (27). Due to its lasting effects and low risk of adverse side effects, rTMS has become the protocol of choice for clinical trials assessing the potential use of TMS as a therapeutic tool to alleviate symptoms of psychiatric disorders (for a review see (27)). A popular protocol to chronically inhibit or excite neural activity is high (5–20 Hz) and low (1 Hz) frequency rTMS (28, 29). High frequency TMS is used to excite neural activity in the targeted area whereas low frequency TMS is used to inhibit neural activity. The use of these protocols has been approved by the US Food and Drug Administration as a therapy for treatment-resistant depression and has since proven to be an efficacious and safe adjunctive intervention for the disorder (30). Theta burst stimulation (TBS) is a newer protocol that allows for the modulation of neural activity in a shorter time compared to high and low frequency rTMS (~1-3 min versus 15-20 min). The development of TBS was borne from animal studies showing that continuous or intermittent patterns of stimulation in the theta frequency range induce long-term potentiation (LTP) and depression (LTD) (31-34). The neurobiological foundations, short duration, and relatively low intensity of stimulation (70% resting motor threshold) represent major advantages of TBS protocols
Cocchi & Zalesky
5
and make them readily translatable to clinical settings. More recent stimulation protocols such as the quadripulse magnetic stimulation (QPS) have been developed (35, 36) but their clinical safety and efficacy in psychiatry remains to be established.
TMS in psychiatry: Challenges and opportunities The use of TMS as a potential tool to evaluate neural function and improve symptoms of psychiatric disorders can be traced to the 1990s (37-40). TMS is now recognised for its role in managing symptoms of major depression in patients failing to respond to at least one or two courses of pharmacological treatment (41). Stimulation of the dorsolateral prefrontal cortex (DLPFC) with rTMS has consistently shown significant efficacy at the group level. Compared to placebo, four-weeks of rTMS was reported to achieve a reduction in the severity of depressive symptoms with a mean weighted effect size of 0.55 (42). However, response to rTMS remains variable in patients with depression (43-45). Moreover, it remains unclear whether rTMS can provide significant clinical benefits as a standalone intervention and whether the positive effects generalize to patients with milder forms of depression. Currently, rTMS is primarily administered as an adjunct therapy to patients who have not responded to pharmacological interventions for depression, resulting in a selection bias for studies investigating these cohorts and in disambiguating potential interactions between pharmacological and rTMS effects. The clinical use of TMS for disorders other than depression has not been approved by regulatory authorities. Several clinical trials targeting different cortical regions including the DLPFC and the supplementary motor area (SMA) have been conducted in psychiatric disorders including schizophrenia (46), OCD (42), and PTSD (47). These clinical trials report mixed evidence for TMS (27), underscoring the need to improve the efficacy and the reliability of TMS interventions. The cortical stimulation target is a key choice for TMS intervention. Efficacious cortical targets are likely to vary between disorders as well as between individuals with the same disorder (7, 48-50). Fitzgerald et al. (50) show that the efficacy of rTMS in treatment-resistant depression can be significantly improved by using a personalised stimulation of the DLPFC based on neuronavigation (Figure 1B). However, the choice of cortical targets for TMS remains largely determined by trial-and-error, previous evidence of efficacy, and knowledge about local abnormalities in cortical activity gathered from neuroimaging. For example, clinical trials of TMS in OCD have often used interventions adapted from TMS protocols for depression (51). These studies have so far failed to provide unequivocal evidence supporting the use of rTMS of DLPFC to alleviate symptoms of OCD. Group-level neuroimaging analyses have been used to inform the choice of stimulation targets in many psychiatric
Cocchi & Zalesky
6
disorders, including schizophrenia, depression, and OCD (45, 52, 53). Cortical regions found to show altered neural activity in a group of patients compared to healthy controls provide candidate targets, under the assumption that TMS can normalize local neural activity and this normalization would be sufficient to improve behavior. The development of neuroimaging methodologies to map whole-brain connectivity provides new opportunities to improve the efficacy of TMS interventions by targeting brain-wide neural abnormalities. It is now recognized that the phenomenology of many psychiatric disorders including schizophrenia (54, 55), OCD (56-58), and depression (59, 60) is associated with distinct changes in connectivity between functionally specialized brain regions. For example, schizophrenia is linked to widespread abnormalities in frontostriatal, frontotemporal, thalamocortical and cortico-cerebellar networks in the resting state (54, 55), whereas OCD is associated with hyper-connectivity between the anterior insula and the dorsal anterior cingulate (dACC) when people engage in an external task (58). Promising recent research suggests that putative responders and non-responders to TMS in psychiatric conditions including OCD (8) and depression (48, 49) can be predicted a priori based on patterns of functional connectivity evident in an individual’s functional magnetic resonance imaging (fMRI) scan. In the future, neuroimaging is therefore likely to play a greater role in planning and administering TMS therapies.
TMS effects at regions distant to the stimulation site It has long been posited that TMS can impact brain activity in regions that are distant to the stimulation site as well as modulate functional connectivity between pairs of brain regions that are not directly stimulated (61-63). More recently, neuroimaging techniques have been used to study the impact of TMS on the activity of whole-brain networks ((64-67); for recent summaries see (7, 68)). For example, Eldaief et al. (64) showed that 1Hz and 20Hz rTMS of the left posterior inferior parietal pole induced two distinct changes in functional brain connectivity between the target regions and other regions comprising the default mode network: 1-Hz stimulation increased functional connectivity between the inferior parietal lobule and the left hippocampal formation. On the other hand, 20-Hz stimulation decreased functional connectivity between the medial prefrontal cortex, posterior cingulate cortex/ventral precuneus, and the targeted site. This suggests that inhibition or excitation of local neuropil can result in multifaceted changes in functional connectivity that extend beyond the primary stimulation site (Figure 2A). More recently, by combining resting-state fMRI and TMS, Cocchi et al. (65) showed that the remote effects of an inhibitory stimulation applied to visual cortical areas can be predicted by the position of the targeted areas within the visual cortical hierarchy (Figure
Cocchi & Zalesky
7
2B). Specifically, inhibitory TBS (continuous TBS, cTBS, (21)) on the right early visual cortex increased coupling between neural activity in this sensory region and activity in higher visual areas. On the other hand, inhibition of the frontal eye fields (FEF) reduced coupling between neural activity in this core visual region and early visual cortices. These findings accord with recent neurophysiological (69, 70), neuroimaging (71), and computational models (72) suggesting that neural systems are organized according to a spatio-temporal hierarchy, with highly interconnected associative regions (hub regions) showing a slower regime of activity compared to sensory-peripheral regions (13).
Figure 2. A. Changes in functional connectivity within the default mode network following TMS of the left posterior inferior parietal lobule (lpIPL). The effects of TMS on functional connectivity were assessed by contrasting resting-state fMRI before and after 1Hz and 20Hz rTMS. Stronger connections are shown with thicker lines. Blue (red) indicates decreased (increased) connectivity. Dashed lines denote a significant change in connectivity strength (pre versus post). mPFC = Medial prefrontal cortex, pCC = posterior cingulate cortex/ventral precuneus, pIPL = the posterior inferior parietal lobule, and HF = hippocampal formation. Note that scaling of line thickness is different between right and left. Adapted with permission from (64). B. Remote effects of local TMS. Changes in resting-state functional connectivity between the target cortical region and the rest of the brain following continuous
Cocchi & Zalesky
8
(inhibitory) theta-burst stimulation (cTBS). The stimulation target is indicated with a lightning symbol. Inhibition of the right early visual cortex increases positive connectivity with other regions within the visual hierarchy. On the other hand, inhibition of the right frontal eye fields reduces functional interactions between this high-order visual area and early visual cortices. Adapted with permission from (65). C. Impact of stimulation target choice on clinical efficacy. (1) shows DLPFC regions that have been used in different studies and their relative efficiency in reducing symptoms of depression. Difference in seed to whole-brain functional connectivity between the more effective site and the less effective (more minus less effective) (2). Each panel depicts the level of correlation of each DLPFC site with the subgenual cingulate (indicated with the arrow and red contours in (2)). The more effective DLPFC site showed greater anticorrelation with the subgenual cingulate compared to the less effective site. Fitzgerald Target: Coordinates are taken from (50) and correspond to x=−46, y=45, z=38 (MNI space). Adapted with permission from (49).
The development of new TMS interventions should take into account the propagation of stimulation effects beyond the immediate stimulation vicinity. Choosing stimulation targets based only on local cortical abnormalities may result in unknown and potentially detrimental downstream effects. To determine distant regions that might encounter downstream effects, non-invasive fiber tractography can be performed from the proposed stimulation site to map white matter fiber bundles that can propagate the stimulation to distant regions (73). In addition to using diffusion MRI to map the anatomical connectivity of candidate stimulation targets, fMRI can be used to provide complementary maps of functional connectivity. These connectivity maps can be used to infer the brain-wide effects of stimulation relative to a local stimulation site. Therefore, therapeutic TMS should be evaluated in light of the network of brain regions in which the targeted cortical area belongs. In line with this approach, the antidepressant effect of TMS (high frequency rTMS) has recently been linked to resting-state patterns of functional connectivity between the left DLPFC (TMS target site) and the subgenual cingulate ((49), but see also (74)) (Figure 2C). Specifically, higher anticorrelation between fMRI activity in sub-regions of the DLPFC and the subgenual cingulate was associated with improved clinical efficacy.
Personalization of TMS interventions Personalizing cortical stimulation targets is motivated by the observation that patterns of abnormal brain activity vary markedly between psychiatric disorders and even between individuals with the same diagnosis (7, 48, 75). To improve reliability and efficacy, TMS interventions can be tailored to each individual to account for inter-individual differences in brain pathology and anatomy. The cortical stimulation target is perhaps the parameter that is most commonly considered for personalization, but other parameters such as the stimulation frequency, intensity, and duration are also amenable to optimization. For example, the stimulation intensity can be adjusted as a function of the scalp-neuropil distance, which can
Cocchi & Zalesky
9
vary markedly between the motor cortex and other regions such as the frontal poles (76). Not accounting for scalp-brain variability will result in suboptimal stimulation protocols that may generate heterogeneous effects across sites. Moreover, a careful manipulation of stimulation parameters and evaluation of the behavioural effects may also facilitate the study of the neural underpinnings of paradoxical behavioural facilitations (77, 78) and network-wide dysfunctions in mechanisms such as intracortical inhibition (79, 80). It is important to note that some degree of personalization is currently common; in that a patient’s structural brain scan is used to guide positioning of the TMS coil at the desired cortical target. The use of anatomical imaging to guide positioning of the TMS coil allows adjustment for inter-individual heterogeneity in neuroanatomy, and ensures that precisely the same cortical region is targeted across therapeutic sessions. Imaging guided neuronavigation to target a cortical region has been shown to increase the response to rTMS treatment in psychiatric disorders including depression [e.g., (50), see also (81)]. There is substantial scope to extend this basic form of personalization to account for inter-individual variation in sex, age, clinical symptoms, brain connectivity, brain chemistry, and genomic information. In addition to improving treatment efficacy, this level of personalization may lead to a better understanding of heterogeneity in neuropathology across patients and provide a deeper understanding of the biological mechanisms of action underpinning TMS. If TMS shows efficacy in a certain patient biotype but not in another, characterizing the salient neurobiological differences between these two groups can provide insight into factors that facilitate TMS efficacy and elucidate potential mechanisms of actions. Personalization can also be undertaken at the level of patient subgroups. It is increasingly recognized that current diagnostic categories in psychiatry are broad and that any given category can encompass individuals with diverse variation in clinical symptoms, genetic risk markers and brain pathology (82, 83). This variation between individuals within broad diagnostic categories can be parsed into patient subgroups, or so-called biotypes, using cluster analysis and knowledge discovery algorithms. Biotypes comprise patient subgroups that are more homogenous relative to their broader diagnostic category with respect to patterns of clinical symptoms, behavioral measures, and/or biological markers including brain connectivity.
Biotype-based
personalization
is
less
challenging
than
individual
personalization and provides an intermediate goal for the field. Importantly, biotype-based personalization is not intended to substitute current clinical nosologies with alternative categories. It rather seeks to parse/dissect the heterogeneity across individuals within these nosolgies into homogenous and consistent patient subtypes. Once biotypes have been established, new patients can be assigned to a biotype according to the results of clinical assessments and neuroimaging.
Cocchi & Zalesky
10
Two kinds of biotypes can be delineated: dimensional and categorical. A dimensional biotype characterizes a continuous relation across individuals with respect to a set of clinical, behavioral or biological characteristics (Figure 3A). A dimensional biotype simplifies to a pairwise association in the case that only a single pair of characteristics is considered. In general though, a dimensional biotype characterizes a spectrum across individuals in a highdimensional space, where each dimension represents a patient characteristic. Dimensional biotypes can be delineated with statistical procedures for discovering continuous relations in high-dimensional data, such as canonical correlation analysis (84) and Gaussian mixture models (85). A categorical biotype characterizes a discrete subgroup of individuals that is segregated with respect to a set of clinical, behavioral or biological characteristics from other individuals comprising a broad diagnostic category (Figure 3B). Categorical biotypes can be delineated with cluster (86) and latent class analysis (87). Borsboom et al. (88) provide a thoughtful and accessible survey of modeling approaches for delineating categorical and dimensional constructs in psychiatry from a psychometric perspective. Most clustering algorithms will segregate data into distinct clusters irrespective of whether segregated clusters are truly present. Therefore, it is important to assess whether the number of categorical biotypes delineated with cluster analysis is supported by appropriate null data (89). A common misconception is that replicating categorical biotypes across independent datasets is sufficient to establish that a categorical representation is parsimoniously supported by the data. Clustering algorithms can consistently partition a continuous relation across multiple datasets into discrete clusters, yet this does not establish the existence of a separating gap where few data points reside. Importantly, categorical and dimensional biotypes are not mutually exclusive, given that a dimensional relation can be found across a segregated cluster of individuals. This might be construed as a hybrid biotype. Figure 3 shows two-dimensional examples that elucidate the difference between dimensional (Figure 3A) and categorical (Figure 3B) biotypes. An example of spuriously delineated biotypes is also shown (Figure 3C).
Cocchi & Zalesky
11
Figure 3. Example biotypes delineated in a hypothetical two-dimensional space. The two dimensions could represent symptoms, behaviors or biological characteristics. Each cross represents an individual. A. Dimensional biotype characterizing a continuous relation across individuals. In two dimensions, dimensional biotypes simplify to pairwise associations (e.g., linear correlation). B. Three distinct categorical biotypes characterizing segregated clusters of individuals with respect to the two dimensions. One of the three biotypes (lowermost) also characterizes a dimensional relation. C. Example of three spurious biotypes. Most clustering algorithms can segregate data into distinct clusters irrespective of whether segregated clusters are truly present. This is referred to as overfitting. Consideration of null data (gap statistic) indicates absence of any categorical or dimensional biotypes in Panel C. For all three examples, Gaussian mixture models comprising 1, 3, and 3 components were fitted, respectively. Contour lines show the probability distribution of the fitted Gaussian distributions. Warm (cool) colors indicate regions of high (low) probability mass.
TMS interventions can be optimized to suit categorical biotypes. For example, if a set of biotypes primarily segregates individuals based on variation in the cortical location of brain pathology, the TMS coil position can be adjusted to target the distinct pathological location associated with each biotype. Individuals might also be excluded from treatment if their goodness-of-fit to an efficacious biotype is marginal (Figure 3B) or they belong to a biotype that has been established to yield low efficacy. Once a set of biotypes have been established, stimulation targets, intensities and frequencies can be progressively optimized for each biotype based on treatment efficacy in new individuals. Biotyping is not the only approach to TMS treatment personalization. If data on treatment efficacy is available in an adequately large group of individuals, pattern recognition techniques (machine learning) can be trained to learn high-dimensional patterns among clinical symptoms and biological markers that may predict an individual’s treatment response. An individual’s likely response to treatment can therefore be predicted in advance and the planned TMS intervention adjusted, or excluded in the case of poor predicted efficacy. The disadvantage of this approach is that it does not provide insight into how an intervention might be adjusted. Drysdale et al. demonstrate the utility of biotype-based personalization of TMS therapy for intractable depression (48). A large cohort of individuals with depression was segregated into four categorical biotypes based on variation in behavioral and functional neuroimaging measures (Figure 4A). The efficacy of high-frequency TMS varied markedly across the four categorical biotypes. TMS intervention was more efficient for individuals in a categorical biotype defined Biotype 1 (82.5% showed >25% improvement in the HAMD depression scale), as compared to Biotype 2 (25.0%), 3 (61%), and 4 (29.6%). It was further shown that patterns of functional connectivity between dorsomedial prefrontal stimulation (TMS target), left amygdala, left DLPFC, bilateral orbitofrontal cortex, and posterior cingulate cortex were highly predictive of TMS response.
Cocchi & Zalesky
12
Biotypes have been delineated in psychiatric disorders other than depression, although it remains to be established whether these biotypes provide viable targets for treatment personalization. Clementz et al (90) delineated three categorical psychosis biotypes using cluster analysis applied to a diverse array of neuropsychological data acquired in individuals with schizophrenia, schizoaffective disorder, and bipolar disorder with psychosis (Figure 4B). While these psychosis biotypes did not simply recapitulate existing diagnostic categories and were independently validated using measures of gray matter density (91), Barch (92) has argued that they may represent a dimension of severity rather than providing segregated categories. Kaczkurkin et al. (93) also used a constrained factor analysis applied to psychiatric symptoms in a large cohort of youth to delineate orthogonal dimensions of psychopathology (i.e., anxious-misery, psychosis, behavioural, fear; Figure 4C). Importantly, the biotypes delineated in these seminal studies can be explicitly related to pathology in specific brain structures and circuits, potentially providing biotype-personalized stimulation targets for TMS.
Figure 4. Categorical and dimensional biotypes delineated in psychiatric populations. A. Drysdale et al (48) delineated four categorical biotypes among individuals with depression. Biotypes were delineated with cluster analysis applied to a two-dimensional space representing linear combinations of interregional functional connectivity strengths that associated with anhedonia-related (horizontal axis) and anxiety-related (vertical axis) symptoms. These two symptom-connectivity dimensions were identified with canonical correlation analysis. Response to high-frequency TMS treatment significantly differed between individuals according to biotype (barplot). Treatment was most effective for individuals in Biotype 1 (Cluster 1), which was characterized by reduced functional connectivity in frontoamygdala networks. B. Clementz et al (90) delineated three categorical Cocchi & Zalesky
13
psychosis biotypes among individuals with schizophrenia (blue triangles), schizoaffective disorder (mauve crosses) and bipolar disorder with psychosis (orange circles). Biotypes were delineated with dimensionality reduction and cluster analysis applied to a wide array of cognitive and neuropsychological data. Ordering individuals according to a schizo-bipolar scale demonstrates that the biotypes do not simply recapitualte standard clinical diagnoses and yield distinctive neurobiological dimensions. C. Kaczkurkin et al (93) applied factor analysis to psychiatric symptom ratings acquired in a large cohort of youth. Four orthogonal dimensions of psychopathology (anxious-misery, psychosis, behavioral, fear) were delineated as well as a dimension of general psychopathology. The barplot shows mean factor scores for each dimension stratified according to diagnostic screening categories (ADHD: attentiondeficit hyperactivity disorder; GAD: generalized anxiety disorder; MDD: major depressive disorder; OCD: obsessive-compulsive disorder; ODD: oppositional definat disorder; PTSD: post-traumatic stress disorder; TD: typically developing). Panel A reproduced from Drysdale et al (48), B reproduced from Celementz et al (90), and C reproduced from Kaczkurkin et al (93), with permission.
Improving the prediction and personalization of TMS using computational modeling and network simulation Stimulation targets that are optimized to individual biotypes will ultimately need to be tested. Moreover, biotype analysis may indicate more than one candidate target, requiring the evaluation of different cortical sites to establish a single clinical target. Computational models to predict the local and distributed effects of TMS can help to offset the cost of empirical testing and assist with personalizing TMS protocols (13). Models for TMS fall within two broad categories: (i) biophysical models of electric currents and neural dynamics that are local to the stimulation site; and, (ii) systems-level (whole-brain) models of functional connectivity and neural dynamics. Computational models for the spatial distribution of the electric fields induced by TMS within cortical neuropil adjacent to the stimulation coil have been widely studied (94-96). These biophysical models enable the impact of inter-individual differences in anatomy on the local electric field distribution to be estimated. Structural brain scans are first used to delineate subject-specific volume meshes of different cortical tissue segments and boundaries (Figure 5A). Each tissue segment is endowed with a distinct conductivity value and finite-element methods are then used to calculate the electric field at each mesh location. The position of the diploes used to model the stimulation coil can be rotated to investigate the impact of coil type, position and orientation. This approach has revealed that the highest field strengths occur at gyral crowns perpendicular to the local electric field orientation (Figure 5B; (96)). Electric field modeling suggests that inter-individual variation in cortical anatomy can substantially impact the strength and distribution of TMS-induced fields. Field calculations may therefore benefit individual stimulation planning and analysis (95). Electric field models are limited to characterizing the local effects of TMS. Effects that are distant to the simulation site can be modeled under a mean-field approximation involving
Cocchi & Zalesky
14
networks of interconnected neural masses (97). Each neural mass represents a node and characterizes the ensemble behavior of a large population of neurons. Nodes are
interconnected according to an individual’s connectome. Typically, a specific neural mass is perturbed to model the impact of TMS applied to the corresponding cortical region, after which neural dynamics are simulated for each neural mass comprising the network to elucidate which distant regions are impacted. How to perturb a single neural mass to parsimoniously model the impact of TMS remains an open issue. Various approaches have been considered, including mean-field models that incorporate a calcium-dependent plasticity component, where the postsynaptic calcium concentration describes long-term potentiation and depression (98, 99); and models that incorporate spike-timing dependent plasticity (99, 100). Modeling each of 513 regions spanning the entire cortex as an interconnected network of Kuramoto oscillators, Gollo et al. (13) comprehensively mapped the distributed effects of TMS on functional connectivity for different stimulation targets (Figure 5C).
Figure 5. Computational modeling of the local and distributed effects of TMS. Finite-element methods can be used to estimate the spatial distribution of TMS-induced electric fields within cortical tissue that is adjacent to the stimulation site. This involves delineating meshes of different cortical tissue segments and then estimating the electric scalar potential for each element in the mesh (A). Each tissue segment (colored distinctly in A) is endowed with a unique conductivity value. B. Variation in the electric field, denoted with E, due to changes in the coil orientation for stimulation of the primary motor cortex. For each panel, the coil is oriented perpendicular to the solid arrow. The distribution of the electric fields was determined with finite-element methods. The strongest field strengths occur at gyri that are
Cocchi & Zalesky
15
perpendicular to the local electric field orientation. C. The distributed impact of TMS on whole-brain functional connectivity simulated with networks of neural masses. Changes in functional connectivity within each brain network following simulated TMS of 513 brain regions. Excitatory or inhibitory TMS are hypothesized to increase or decrease local neural dynamics, respectively (65). Because of the temporal organization of whole-brain systems (70, 72), local changes in regional activity alter coupling between the targeted region and the rest of the brain. Green colors show the impact that simulated TMS has on functional connectivity within the network in which the region belongs (e.g., default mode network). Conversely, purple colors indicate the impact that simulated TMS can have on brain networks outside the resting-state network comprising the targeted region. Darker colors represent greater effects of simulated TMS on brain networks. µ=mean change of functional connectivity within the network, σ=standard deviation. (13) Panel A and B reproduced from Theilescher et al (96), and C reproduced from Gollo et al (13) with permission.
Simulating brain network activity in silico using network models can also assist with predicting the impact that abnormal brain characteristics, such as white matter pathology, may have on the whole-brain effect of local TMS. In summary, computational modeling and network simulation provides a means to efficiently estimate the impact of candidate TMS interventions at different spatial scales, from locally induced electric fields, to brain network dynamics and whole-brain functional connectivity. Simulation targets and other simulation parameters can be iteratively optimized in silico until the desired mechanism of action is achieved. However, in silico experimentation cannot inform clinical efficacy, since even if computational modeling reveals that a particular stimulation target achieves a desired normalization of network dynamics, this does imply that TMS applied to the indicated target will necessarily provide therapeutic benefit.
Conclusion and future directions Here, we argued that the efficacy and reliability of TMS can be enhanced with intervention personalization. We considered two complementary approaches to personalization; namely, personalizing interventions at the i) biotype and, ii) single-subject level. Biotypes delineate patient subgroups that exhibit greater homogeneity with respect to brain chemistry, connectivity, and electrophysiology as well as behavior and clinical symptoms, relative to current broad diagnostic categories. Parsing inter-individual heterogeneity within existing clinical diagnostic categories represents an achievable first step toward the development of individualized TMS interventions. Here, we discussed recent work highlighting the feasibility of biotyping individuals with complex psychiatric symptoms and reviewed how TMS interventions could be tailored to suit an individual’s biotype. We also reviewed how biotypebased personalization of TMS interventions can be further optimized based on the consideration of individual factors including anatomy and cortical excitability. As established
Cocchi & Zalesky
16
in other branches of medicine (101-103), personalised interventions are likely to significantly advance the reliability and efficacy of TMS as a therapy for psychiatric disorders. If the clinical rollout of biotype-guided TMS personalization is endorsed by future research, standardized neuroimaging protocols will need to be established and validated software will need to be developed to accurately and reliably determine an individual’s biotype. Proof-ofconcept trials assessing the efficacy of biotype-based TMS over standard FDA approved TMS intervention for depression could provide initial motivation to undertake larger trials and extend the approach to other disorders. Ultimately, the utility of implementing personalized TMS interventions in the clinic will require a cost-benefits analysis. However, before these economic issues should be addressed further evidence attesting to the utility of biotype-guided personalised TMS will be crucial.
Cocchi & Zalesky
17
Acknowledgments L.C. and A.Z were supported by the Australian National Health Medical Research Council (L.C. APP1099082, A.Z. APP1047648).
Cocchi & Zalesky
18
Financial Disclosures None (both LC and AZ)
Cocchi & Zalesky
19
References 1. March JS, Silva S, Petrycki S, Curry J, Wells K, Fairbank J, et al. (2007): The Treatment for Adolescents With Depression Study (TADS): long-term effectiveness and safety outcomes. Archives of general psychiatry. 64:1132-1143. 2. Leichsenring F, Rabung S (2008): Effectiveness of long-term psychodynamic psychotherapy: A meta-analysis. Jama. 300:1551-1565. 3. Shaw M, Hodgkins P, Caci H, Young S, Kahle J, Woods AG, et al. (2012): A systematic review and analysis of long-term outcomes in attention deficit hyperactivity disorder: effects of treatment and non-treatment. BMC medicine. 10:99. 4. Bloch MH, Landeros-Weisenberger A, Kelmendi B, Coric V, Bracken MB, Leckman JF (2006): A systematic review: antipsychotic augmentation with treatment refractory obsessive-compulsive disorder. Mol Psychiatry. 11:622-632. 5. Fox MD, Buckner RL, Liu H, Chakravarty MM, Lozano AM, Pascual-Leone A (2014): Resting-state networks link invasive and noninvasive brain stimulation across diverse psychiatric and neurological diseases. Proceedings of the National Academy of Sciences of the United States of America. 111:E4367-4375. 6. Malenka RC, Bear MF (2004): LTP and LTD: an embarrassment of riches. Neuron. 44:5-21. 7. Sale MV, Mattingley JB, Zalesky A, Cocchi L (2015): Imaging human brain networks to improve the clinical efficacy of non-invasive brain stimulation. Neuroscience and biobehavioral reviews. 57:187-198. 8. Dunlop K, Woodside B, Olmsted M, Colton P, Giacobbe P, Downar J (2016): Reductions in Cortico-Striatal Hyperconnectivity Accompany Successful Treatment of Obsessive-Compulsive Disorder with Dorsomedial Prefrontal rTMS. Neuropsychopharmacology. 41:1395-1403. 9. Wilson M, Fung P, Robinson P, Shemmell J, Reynolds J (2016): Calcium dependent plasticity applied to repetitive transcranial magnetic stimulation with a neural field model. Journal of computational neuroscience.1-19. 10. Wilson M, Goodwin D, Brownjohn P, Shemmell J, Reynolds J (2014): Numerical modelling of plasticity induced by transcranial magnetic stimulation. Journal of computational neuroscience. 36:499-514. 11. Wilson MT, Fulcher BD, Fung PK, Robinson PA, Fornito A, Rogasch NC (2017): Biophysical modeling of neural plasticity induced by transcranial magnetic stimulation. bioRxiv.175893. 12. Esser SK, Hill SL, Tononi G (2005): Modeling the effects of transcranial magnetic stimulation on cortical circuits. Journal of Neurophysiology. 94:622-639. 13. Gollo LL, Roberts JA, Cocchi L (2017): Mapping how local perturbations influence systems-level brain dynamics. NeuroImage. 14. Cona F, Zavaglia M, Massimini M, Rosanova M, Ursino M (2011): A neural mass model of interconnected regions simulates rhythm propagation observed via TMS-EEG. NeuroImage. 57:1045-1058. 15. Bullmore E, Sporns O (2009): Complex brain networks: graph theoretical analysis of structural and functional systems. Nature Reviews Neuroscience. 10:186-198. 16. Barker AT, Jalinous R, Freeston IL (1985): Non-invasive magnetic stimulation of human motor cortex. The Lancet. 325:1106-1107. 17. Hallett M (2000): Transcranial magnetic stimulation and the human brain. Nature. 406:147-150. 18. Hallett M (2007): Transcranial magnetic stimulation: a primer. Neuron. 55:187-199. 19. Barker AT (1991): An introduction to the basic principles of magnetic nerve stimulation. Journal of Clinical Neurophysiology. 8:26-37. 20. Bohning D, Shastri A, McConnell K, Nahas Z, Lorberbaum J, Roberts D, et al. (1999): A combined TMS/fMRI study of intensity-dependent TMS over motor cortex. Biological psychiatry. 45:385-394.
Cocchi & Zalesky
20
21. Huang YZ, Edwards MJ, Rounis E, Bhatia KP, Rothwell JC (2005): Theta burst stimulation of the human motor cortex. Neuron. 45:201-206. 22. de Lara LIN, Tik M, Woletz M, Frass-Kriegl R, Moser E, Laistler E, et al. (2017): High-sensitivity TMS/fMRI of the Human Motor Cortex Using a Dedicated Multichannel MR Coil. NeuroImage. 150:262-269. 23. Casula EP, Tarantino V, Basso D, Arcara G, Marino G, Toffolo GM, et al. (2014): Low-frequency rTMS inhibitory effects in the primary motor cortex: Insights from TMSevoked potentials. NeuroImage. 98:225-232. 24. Murphy SC, Palmer LM, Nyffeler T, Müri RM, Larkum ME (2016): Transcranial magnetic stimulation (TMS) inhibits cortical dendrites. eLife. 5:e13598. 25. Edgley S, Eyre J, Lemon R, Miller S (1997): Comparison of activation of corticospinal neurons and spinal motor neurons by magnetic and electrical transcranial stimulation in the lumbosacral cord of the anaesthetized monkey. Brain: a journal of neurology. 120:839-853. 26. Rossi S, Hallett M, Rossini PM, Pascual-Leone A, Safety of TMSCG (2009): Safety, ethical considerations, and application guidelines for the use of transcranial magnetic stimulation in clinical practice and research. Clinical neurophysiology : official journal of the International Federation of Clinical Neurophysiology. 120:2008-2039. 27. Lefaucheur J-P, André-Obadia N, Antal A, Ayache SS, Baeken C, Benninger DH, et al. (2014): Evidence-based guidelines on the therapeutic use of repetitive transcranial magnetic stimulation (rTMS). Clinical Neurophysiology. 125:2150-2206. 28. Peinemann A, Reimer B, Löer C, Quartarone A, Münchau A, Conrad B, et al. (2004): Long-lasting increase in corticospinal excitability after 1800 pulses of subthreshold 5 Hz repetitive TMS to the primary motor cortex. Clinical neurophysiology. 115:1519-1526. 29. Brückner S, Kiefer M, Kammer T (2013): Comparing the after-effects of continuous theta burst stimulation and conventional 1Hz rTMS on semantic processing. Neuroscience. 233:64-71. 30. Connolly KR, Helmer A, Cristancho MA, Cristancho P, O'Reardon JP (2012): Effectiveness of transcranial magnetic stimulation in clinical practice post-FDA approval in the United States: results observed with the first 100 consecutive cases of depression at an academic medical center. The Journal of clinical psychiatry. 73:e567-573. 31. Hess G, Aizenman CD, Donoghue JP (1996): Conditions for the induction of longterm potentiation in layer II/III horizontal connections of the rat motor cortex. J Neurophysiol. 75:1765-1778. 32. Otani S, Blond O, Desce JM, Crepel F (1998): Dopamine facilitates long-term depression of glutamatergic transmission in rat prefrontal cortex. Neuroscience. 85:669-676. 33. Urban J, Kossut M, Hess G (2002): Long-term depression and long-term potentiation in horizontal connections of the barrel cortex. The European journal of neuroscience. 16:1772-1776. 34. Thickbroom GW (2007): Transcranial magnetic stimulation and synaptic plasticity: experimental framework and human models. Experimental Brain Research. 180:583-593. 35. Hamada M, Terao Y, Hanajima R, Shirota Y, Nakatani‐Enomoto S, Furubayashi T, et al. (2008): Bidirectional long‐term motor cortical plasticity and metaplasticity induced by quadripulse transcranial magnetic stimulation. The Journal of physiology. 586:3927-3947. 36. Hanajima R, Tanaka N, Tsutsumi R, Enomoto H, Abe M, Nakamura K, et al. (2017): The effect of age on the homotopic motor cortical long-term potentiation-like effect induced by quadripulse stimulation. Experimental Brain Research.1-6. 37. Conca A, Koppi S, Konig P, Swoboda E, Krecke N (1996): Transcranial magnetic stimulation: a novel antidepressive strategy? Neuropsychobiology. 34:204-207. 38. Geller V, Grisaru N, Abarbanel JM, Lemberg T, Belmaker RH (1997): Slow magnetic stimulation of prefrontal cortex in depression and schizophrenia. Prog Neuropsychopharmacol Biol Psychiatry. 21:105-110. 39. Klein E, Kolsky Y, Puyerovsky M, Koren D, Chistyakov A, Feinsod M (1999): Right prefrontal slow repetitive transcranial magnetic stimulation in schizophrenia: a double-blind sham-controlled pilot study. Biol Psychiatry. 46:1451-1454. Cocchi & Zalesky
21
40. Puri B, Davey N, Ellaway P, Lewis S (1996): An investigation of motor function in schizophrenia using transcranial magnetic stimulation of the motor cortex. The British journal of psychiatry. 169:690-695. 41. Horvath JC, Mathews J, Demitrack MA, Pascual-Leone A (2010): The NeuroStar TMS device: conducting the FDA approved protocol for treatment of depression. JoVE (Journal of Visualized Experiments).e2345-e2345. 42. Slotema CW, Dirk Blom J, Hoek HW, Sommer IE (2010): Should we expand the toolbox of psychiatric treatment methods to include Repetitive Transcranial Magnetic Stimulation (rTMS)? A meta-analysis of the efficacy of rTMS in psychiatric disorders. Journal of Clinical Psychiatry. 71:873. 43. Avery DH, Isenberg KE, Sampson SM, Janicak PG, Lisanby SH, Maixner DF, et al. (2008): Transcranial magnetic stimulation in the acute treatment of major depressive disorder: clinical response in an open-label extension trial. Journal of Clinical Psychiatry. 69:441-451. 44. Luborzewski A, Schubert F, Seifert F, Danker-Hopfe H, Brakemeier EL, Schlattmann P, et al. (2007): Metabolic alterations in the dorsolateral prefrontal cortex after treatment with high-frequency repetitive transcranial magnetic stimulation in patients with unipolar major depression. J Psychiatr Res. 41:606-615. 45. George MS, Nahas Z, Molloy M, Speer AM, Oliver NC, Li XB, et al. (2000): A controlled trial of daily left prefrontal cortex TMS for treating depression. Biol Psychiatry. 48:962-970. 46. Dougall N, Maayan N, Soares‐Weiser K, McDermott LM, McIntosh A (2015): Transcranial magnetic stimulation (TMS) for schizophrenia. The Cochrane Library. 47. Trevizol AP, Barros MD, Silva PO, Osuch E, Cordeiro Q, Shiozawa P (2016): Transcranial magnetic stimulation for posttraumatic stress disorder: an updated systematic review and meta-analysis. Trends in psychiatry and psychotherapy. 38:50-55. 48. Drysdale AT, Grosenick L, Downar J, Dunlop K, Mansouri F, Meng Y, et al. (2016): Resting-state connectivity biomarkers define neurophysiological subtypes of depression. Nat Med. 49. Fox MD, Buckner RL, White MP, Greicius MD, Pascual-Leone A (2012): Efficacy of TMS targets for depression is related to intrinsic functional connectivity with the subgenual cingulate. Biol Psychiatry. 72:595-603. 50. Fitzgerald PB, Hoy K, McQueen S, Maller JJ, Herring S, Segrave R, et al. (2009): A randomized trial of rTMS targeted with MRI based neuro-navigation in treatment-resistant depression. Neuropsychopharmacology. 34:1255. 51. Sarkhel S, Sinha VK, Praharaj SK (2010): Adjunctive high-frequency right prefrontal repetitive transcranial magnetic stimulation (rTMS) was not effective in obsessive– compulsive disorder but improved secondary depression. Journal of anxiety disorders. 24:535-539. 52. George MS, Wassermann EM (1994): Rapid-rate transcranial magnetic stimulation and ECT. Convulsive therapy. 53. Pascual-Leone A, Rubio B, Pallardo F, Catala MD (1996): Rapid-rate transcranial magnetic stimulation of left dorsolateral prefrontal cortex in drug-resistant depression. Lancet. 348:233-237. 54. Cocchi L, Harding IH, Lord A, Pantelis C, Yucel M, Zalesky A (2014): Disruption of structure-function coupling in the schizophrenia connectome. Neuroimage Clin. 4:779-787. 55. Fornito A, Harrison BJ, Goodby E, Dean A, Ooi C, Nathan PJ, et al. (2013): Functional dysconnectivity of corticostriatal circuitry as a risk phenotype for psychosis. JAMA psychiatry. 70:1143-1151. 56. Harrison BJ, Pujol J, Cardoner N, Deus J, Alonso P, Lopez-Sola M, et al. (2013): Brain corticostriatal systems and the major clinical symptom dimensions of obsessivecompulsive disorder. Biol Psychiatry. 73:321-328. 57. Harrison BJ, Soriano-Mas C, Pujol J, Ortiz H, Lopez-Sola M, Hernandez-Ribas R, et al. (2009): Altered corticostriatal functional connectivity in obsessive-compulsive disorder. Arch Gen Psychiatry. 66:1189-1200.
Cocchi & Zalesky
22
58. Cocchi L, Harrison BJ, Pujol J, Harding IH, Fornito A, Pantelis C, et al. (2012): Functional alterations of large-scale brain networks related to cognitive control in obsessivecompulsive disorder. Hum Brain Mapp. 33:1089-1106. 59. Greicius MD, Flores BH, Menon V, Glover GH, Solvason HB, Kenna H, et al. (2007): Resting-state functional connectivity in major depression: abnormally increased contributions from subgenual cingulate cortex and thalamus. Biol Psychiatry. 62:429-437. 60. Davey CG, Breakspear M, Pujol J, Harrison BJ (2017): A Brain Model of Disturbed Self-Appraisal in Depression. American Journal of Psychiatry.appi. ajp. 2017.16080883. 61. Ilmoniemi RJ, Virtanen J, Ruohonen J, Karhu J, Aronen HJ, Katila T (1997): Neuronal responses to magnetic stimulation reveal cortical reactivity and connectivity. Neuroreport. 8:3537-3540. 62. Fox P, Ingham R, George MS, Mayberg H, Ingham J, Roby J, et al. (1997): Imaging human intra-cerebral connectivity by PET during TMS. Neuroreport. 8:2787-2791. 63. Paus T, Jech R, Thompson CJ, Comeau R, Peters T, Evans AC (1997): Transcranial magnetic stimulation during positron emission tomography: a new method for studying connectivity of the human cerebral cortex. Journal of Neuroscience. 17:3178-3184. 64. Eldaief MC, Halko MA, Buckner RL, Pascual-Leone A (2011): Transcranial magnetic stimulation modulates the brain's intrinsic activity in a frequency-dependent manner. Proceedings of the National Academy of Sciences of the United States of America. 108:21229-21234. 65. Cocchi L, Sale MV, L LG, Bell PT, Nguyen VT, Zalesky A, et al. (2016): A hierarchy of timescales explains distinct effects of local inhibition of primary visual cortex and frontal eye fields. Elife. 5. 66. Watanabe T, Hanajima R, Shirota Y, Ohminami S, Tsutsumi R, Terao Y, et al. (2013): Bidirectional effects on interhemispheric resting-state functional connectivity induced by excitatory and inhibitory repetitive transcranial magnetic stimulation. Hum Brain Mapp. 67. Hill CA, Suzuki S, Polania R, Moisa M, O'Doherty JP, Ruff CC (2017): A causal account of the brain network computations underlying strategic social behavior. Nature neuroscience. 20:1142-1149. 68. Bortoletto M, Veniero D, Thut G, Miniussi C (2015): The contribution of TMS-EEG coregistration in the exploration of the human cortical connectome. Neuroscience and biobehavioral reviews. 49:114-124. 69. Murray JD, Bernacchia A, Freedman DJ, Romo R, Wallis JD, Cai X, et al. (2014): A hierarchy of intrinsic timescales across primate cortex. Nature neuroscience. 17:1661-1663. 70. Honey CJ, Thesen T, Donner TH, Silbert LJ, Carlson CE, Devinsky O, et al. (2012): Slow cortical dynamics and the accumulation of information over long timescales. Neuron. 76:423-434. 71. Hasson U, Yang E, Vallines I, Heeger DJ, Rubin N (2008): A hierarchy of temporal receptive windows in human cortex. The Journal of neuroscience : the official journal of the Society for Neuroscience. 28:2539-2550. 72. Gollo LL, Zalesky A, Hutchison RM, van den Heuvel M, Breakspear M (2015): Dwelling quietly in the rich club: brain network determinants of slow cortical fluctuations. Philosophical transactions of the Royal Society of London Series B, Biological sciences. 370. 73. Nummenmaa A, McNab JA, Savadjiev P, Okada Y, Hamalainen MS, Wang R, et al. (2014): Targeting of white matter tracts with transcranial magnetic stimulation. Brain Stimul. 7:80-84. 74. Avissar M, Powell F, Ilieva I, Respino M, Gunning FM, Liston C, et al. (2017): Functional connectivity of the left DLPFC to striatum predicts treatment response of depression to TMS. Brain Stimulation. 75. Fornito A, Zalesky A, Breakspear M (2015): The connectomics of brain disorders. Nature reviews Neuroscience. 16:159-172. 76. Stokes MG, Chambers CD, Gould IC, Henderson TR, Janko NE, Allen NB, et al. (2005): Simple metric for scaling motor threshold based on scalp-cortex distance: application to studies using transcranial magnetic stimulation. Journal of neurophysiology. 94:4520-4527.
Cocchi & Zalesky
23
77. Walsh V, Ellison A, Battelli L, Cowey A (1998): Task–specific impairments and enhancements induced by magnetic stimulation of human visual area V5. Proceedings of the Royal Society of London B: Biological Sciences. 265:537-543. 78. Hilgetag CC, Théoret H, Pascual-Leone A (2001): Enhanced visual spatial attention ipsilateral to rTMS-induced'virtual lesions' of human parietal cortex. Nature neuroscience. 4:953-957. 79. Oliveri M, Rossini P, Traversa R, Cicinelli P, Filippi M, Pasqualetti P, et al. (1999): Left frontal transcranial magnetic stimulation reduces contralesional extinction in patients with unilateral right brain damage. Brain. 122:1731-1739. 80. Pascual-Leone A, Walsh V (2002): Transcranial magnetic stimulation. Brain mapping: The methods.225-290. 81. Sack AT, Kadosh RC, Schuhmann T, Moerel M, Walsh V, Goebel R (2009): Optimizing functional accuracy of TMS in cognitive studies: a comparison of methods. Journal of Cognitive Neuroscience. 21:207-221. 82. Cuthbert BN, Insel TR (2013): Toward the future of psychiatric diagnosis: the seven pillars of RDoC. BMC medicine. 11:126. 83. Insel T, Cuthbert B, Garvey M, Heinssen R, Pine DS, Quinn K, et al. (2010): Research domain criteria (RDoC): toward a new classification framework for research on mental disorders. Am Psychiatric Assoc. 84. Krzanowski W (2000): Principles of multivariate analysis. OUP Oxford. 85. Reynolds DA, Quatieri TF, Dunn RB (2000): Speaker verification using adapted Gaussian mixture models. Digital signal processing. 10:19-41. 86. Kaufman L, Rousseeuw PJ (2009): Finding groups in data: an introduction to cluster analysis. John Wiley & Sons. 87. McCutcheon AL (1987): Latent class analysis. Sage. 88. Borsboom D, Rhemtulla M, Cramer A, Van der Maas H, Scheffer M, Dolan C (2016): Kinds versus continua: a review of psychometric approaches to uncover the structure of psychiatric constructs. Psychological medicine. 46:1567-1579. 89. Tibshirani R, Walther G, Hastie T (2001): Estimating the number of clusters in a data set via the gap statistic. Journal of the Royal Statistical Society: Series B (Statistical Methodology). 63:411-423. 90. Clementz BA, Sweeney JA, Hamm JP, Ivleva EI, Ethridge LE, Pearlson GD, et al. (2015): Identification of distinct psychosis biotypes using brain-based biomarkers. American Journal of Psychiatry. 173:373-384. 91. Ivleva EI, Clementz BA, Dutcher AM, Arnold SJ, Jeon-Slaughter H, Aslan S, et al. (2017): Brain structure biomarkers in the psychosis biotypes: findings from the bipolarschizophrenia network for intermediate phenotypes. Biological psychiatry. 82:26-39. 92. Barch DM (2017): Biotypes: Promise and Pitfalls. Biol Psychiatry. 82:2-3. 93. Kaczkurkin A, Moore T, Calkins M, Ciric R, Detre J, Elliott M, et al. (2017): Common and dissociable regional cerebral blood flow differences associate with dimensions of psychopathology across categorical diagnoses. Molecular psychiatry. 94. Datta A, Bikson M, Fregni F (2010): Transcranial direct current stimulation in patients with skull defects and skull plates: high-resolution computational FEM study of factors altering cortical current flow. NeuroImage. 52:1268-1278. 95. Hartwigsen G, Bergmann TO, Herz DM, Angstmann S, Karabanov A, Raffin E, et al. (2015): Modeling the effects of noninvasive transcranial brain stimulation at the biophysical, network, and cognitive level. Progress in brain research. 222:261-287. 96. Thielscher A, Opitz A, Windhoff M (2011): Impact of the gyral geometry on the electric field induced by transcranial magnetic stimulation. NeuroImage. 54:234-243. 97. Deco G, Jirsa VK, Robinson PA, Breakspear M, Friston K (2008): The dynamic brain: from spiking neurons to neural masses and cortical fields. PLoS computational biology. 4:e1000092. 98. Fung P, Robinson P (2013): Neural field theory of calcium dependent plasticity with applications to transcranial magnetic stimulation. Journal of theoretical biology. 324:72-83.
Cocchi & Zalesky
24
99. Wilson MT, Fung PK, Robinson PA, Shemmell J, Reynolds JN (2016): Calcium dependent plasticity applied to repetitive transcranial magnetic stimulation with a neural field model. J Comput Neurosci. 41:107-125. 100. Fung PK, Haber AL, Robinson PA (2013): Neural field theory of plasticity in the cerebral cortex. J Theor Biol. 318:44-57. 101. Heinemann V, Douillard J, Ducreux M, Peeters M (2013): Targeted therapy in metastatic colorectal cancer–an example of personalised medicine in action. Cancer treatment reviews. 39:592-601. 102. Ladapo JA, Budoff M, Sharp D, Zapien M, Huang L, Maniet B, et al. (2017): Clinical Utility of a Precision Medicine Test Evaluating Outpatients with Suspected Obstructive Coronary Artery Disease. The American journal of medicine. 130:482. e411-482. e417. 103. Whirl‐Carrillo M, McDonagh E, Hebert J, Gong L, Sangkuhl K, Thorn C, et al. (2012): Pharmacogenomics knowledge for personalized medicine. Clinical Pharmacology & Therapeutics. 92:414-417.
Cocchi & Zalesky
25
