WO2019232125A1

WO2019232125A1 - Method and computing device for generating a brain stimulation treatment atlas

Info

Publication number: WO2019232125A1
Application number: PCT/US2019/034511
Authority: WO
Inventors: Tommi RAIJ; Aapo NUMMENMAA
Original assignee: Rehabilitation Institute Of Chicago; The General Hospital Corporation
Priority date: 2018-05-30
Filing date: 2019-05-30
Publication date: 2019-12-05

Abstract

A method for generating a transcranial magnetic stimulation (TMS) treatment atlas includes computing (or measuring) an electric field (E-field) intensity induced by TMS of a location of a cortex of each of a plurality of patients, resulting in a plurality of E-field values; determining a treatment efficacy for each of the plurality of patients, resulting in a plurality of efficacy values; calculating a correlation between the plurality of E-field values and the plurality of efficacy values, resulting in a correlation value for each cortical location; and generating a TMS treatment atlas based on the plurality of correlation values. This atlas on the cortical surface can be extended to include large-scale networks throughout the brain.

Description

METHOD AND COMPUTING DEVICE FOR GENERATING A BRAIN

STIMUUATION TREATMENT ATUAS

CROSS-REFERENCE TO REUATED APPUICATIONS

[0001] This application claims the priority benefit of U.S. Provisional Application No. 62/677,997, filed May 30, 2018 and incorporated herein by reference in its entirety.

TECHNICAU FIEUD

[0002] The disclosure relates generally to computer-implemented analysis of brain stimulation.

BACKGROUND

[0003] Brain cells may be activated by the delivery of electric current with brain stimulator devices. The electric currents from brain stimulator devices are not uniform throughout the brain, but instead are unevenly distributed, such that some brain areas are exposed to stronger currents than others.

[0004] Brain stimulation is the application of a device that generates electric current in brain tissue. There are various brain stimulation techniques which are used for treating patients. Some of the techniques are non-invasive (not requiring surgery) and others are invasive (requiring surgery). Non-invasive techniques include methods based on electromagnetic induction, such as transcranial magnetic stimulation (“TMS”) and magnetic convulsive therapy (“MCT”). Other non-invasive techniques include methods that place electrodes on the scalp, such as transcranial direct current stimulation (“tDCS"); transcranial alternating current stimulation (“tACS"); and electroconvulsive therapy (“ECT”). Invasive techniques known in the art include epidural, cortical, or deep brain stimulation.

[0005] The human brain shows functional localization, which means that different brain areas perform different functions. For example, the occipital region of the brain performs visual functions while the precentral gyrus performs motor functions.

[0006] Brain stimulation devices are already being used for treating several disorders. The effect of brain stimulation depends on where the stimulation is applied. Different disorders and symptoms are typically associated with abnormalities in different brain regions and networks. Therefore, the clinical efficacy of brain stimulation therapies may depend on the ability to focus the stimulation on the target(s) with best therapeutic efficacy, while avoiding the stimulation of any areas that might worsen any of the symptoms.

[0007] One drawback is that with few exceptions (e.g., deep brain stimulator electrode placement in Parkinson's disease), the optimal location(s) and detrimental area(s) generally is (are) not known. Disorders where the optimal location(s) is (are) not known include, but are not limited to, depression and other mood disorders, different addictions, schizophrenia and other psychotic disorders, personality disorders, phobias, obsessive-compulsive disorder (OCD), posttraumatic stress disorder (PTSD), and other anxiety disorders, acute pain, and chronic pain. Even if there are areas that are routinely targeted, such as the dorsolateral prefrontal cortex (“DLPFC”) in treating depression with TMS, the DLPFC is a large and heterogeneous region, and it is not known where inside or around DLPFC the best location for stimulation might be.

DRAWINGS

[0008] While the appended claims set forth the features of the present techniques with particularity, these techniques may be best understood from the following detailed description taken in conjunction with the accompanying drawings of which:

[0009] FIG. 1A and FIG. 1B illustrate how two different embodiments of treatment atlases would look.

[0010] FIG. 2 illustrates a method for generating a TMS-based treatment atlas according to an embodiment.

[0011] FIG. 3 provides additional details of the method described in conjunction with FIG. 2.

[0012] FIG. 4 illustrates a method for generating a network-level treatment atlas.

[0013] FIG. 5 depicts a diagram of a TMS device.

[0014] FIG. 6 depicts a computing device that may be used to carry out any of the methods described herein.

DESCRIPTION

[0015] This disclosure is generally directed to brain stimulation treatment atlases. The atlases are intended for use together with brain stimulator devices. The purpose of treatment atlases is to optimize the clinical efficacy of brain stimulation therapies by indicating which brain area(s) should be targeted for best treatment outcomes. Specifically, treatment atlases are group-level spatial maps that indicate the most likely therapeutic outcomes associated with stimulation of different brain area(s) or large-scale networks. The idea is based on the effects of brain stimulation being dependent on which brain area(s) is (are) stimulated. The problem that this disclosure addresses is that, for most disorders, the optimal location(s) where stimulation should be applied is (are) not known.

[0016] To address this problem, various embodiments involve creating maps (“brain stimulation treatment atlases”) that reveal the optimal location(s). At the same time, the atlases will reveal any potential locations that have detrimental effects, and areas that have no effects.

[0017] Treatment atlases are computed from data of patients that have already received brain stimulation where the following two items were recorded (or can be estimated): the spatial distribution of brain stimulation (i.e., the stimulation intensity for each brain location), and the treatment outcome in each patient.

[0018] To create an atlas, such data are taken from several patients. First, the TMS- induced electric field (“E-field”) maps showing the distribution of brain stimulation intensities are computed in each individual’s brain (FIG. 2 first column). Second, these maps are spatially aligned between subjects by projecting the individual cortical surfaces to a target template cortical surface (FIG. 2 second column). Third, the individual treatment outcomes (FIG. 2 Therapeutic Efficacy Index, (“TEI”)) are listed. Fourth, the TEI values are combined with information about the brain stimulation intensities (FIG. 2 fourth column show an example where the TEI values were correlated with the stimulation intensity value for each cortical location separately; the fourth column shows the intensity vs. TEI plots for two different cortical locations). Fifth, the values are displayed for all cortical locations (FIG. 2 rightmost column), which is the treatment atlas. This treatment atlas will then be used in new patients to decide where the stimulation is likely to have the best therapeutic effect.

[0019] Possible techniques of combining the information about brain stimulation with information about treatment outcomes include, but are not limited to, computing group-level correlations between the intensity of brain stimulation at each brain location and a number reflecting the treatment outcome, or using pattern recognition / machine learning algorithms. [0020] According to an embodiment, a method for generating a transcranial magnetic stimulation (TMS) treatment atlas includes determining (e.g., measuring or computing) a TMS-induced electric field (“E-field”) distributions on the cortex of each of a plurality of patients, resulting in a plurality of E-field values; determining a treatment efficacy for each of the plurality of patients; calculating a correlation between the plurality of E-field values and the plurality of efficacy values, resulting in a correlation value; and generating a cortical TMS treatment map based on the correlation values at each cortical location.

[0021] In an embodiment, determining a treatment efficacy for each of the plurality of patients comprises determining a percentage of clean drug tests for each of the plurality of patients.

[0022] According to an embodiment, calculating a correlation between the plurality of E- field values and the plurality of efficacy values includes calculating a Spearman’s rho coefficient between the plurality of E-field values and the plurality of efficacy values.

[0023] In an embodiment, the location is one of a plurality of cortex locations and the method further includes repeating the measuring, determining, and calculating steps for each of the plurality of cortex locations, resulting in a plurality of correlation values; and generating the frontal cortex TMS treatment map additionally based on the plurality of correlation values.

[0024] According to an embodiment, generating the cortex TMS treatment map based on the plurality of correlation values includes generating a color coded image of a cortex in which each of a plurality of colors represents a range of correlation values.

[0025] In an embodiment, the method further includes obtaining an image of the cortices of the plurality of patients and computing coregistration matrices between the cortex of each patient and the cortex of a template brain.

[0026] While in the examples below there is an inherent assumption that stronger intensity of brain stimulation results in stronger changes in neuronal activity, this not a necessary assumption. The disclosure also addresses techniques where the atlas is computed using other functions between intensity of brain stimulation and changes in neuronal activity.

[0027] While an atlas may reveal an optimal location where stimulation could be applied, the disclosure also covers the idea of using the entire spatial distribution of the atlas (not only the maxima) for planning stimulation targets (i.e., the stimulation distribution could be made to mimic the entire atlas distribution).

[0028] In the description that follows, reference will be made to colors in the figures. The visual key to colors that may be used in an embodiment is shown in FIG. 1B.

[0029] FIG. 1A and FIG. 1B illustrate how two different embodiments of treatment atlases would look. FIG. 1A shows a surface atlas (“cortical atlas”). This map reveals the areas on the cortical surface that are associated with different experimental outcomes. FIG. 1B shows a network-level treatment atlas (“network atlas”). This map utilizes brain connectivity information to reveal large-scale three-dimensional networks associated with different experimental outcomes. Below is described a method of how these types of atlases are computed according to an embodiment. It is to be understood that the steps do not necessarily have to be carried out in the order described.

[0030] Generating a cortical atlas (to look like, for example, FIG. 1A): First, within each individual patient, the spatial distribution of brain stimulation is weighted with a number reflecting the treatment outcome (i.e., therapeutic efficacy). For example, patients that improved may receive a positive weight, patients where there was no effect receive a zero weight, and patients that became worse receive a negative weight. Thereafter, the individual results are spatially morphed into a standard space, and the results are averaged across subjects. The outcome of this analysis is a spatial map (“atlas”) that indicates which cortical areas are associated with the best treatment outcomes (“positive hotspots”), which areas are associated with no effects, and which areas are associated with the worst outcomes (“negative hotspots”).

[0031] Generating a network atlas (to look like, for example, FIG. 1B): First, the cortical atlas is computed as described above, both at the individual and at the group levels. Then, this information is combined with additional brain connectivity information (e.g., from diffusion MRI tractography as in FIG. 2, or resting-state fMRI, PET, MEG/EEG, and/or intracranial recordings). The individual level spatial distribution maxima are used as seeds in individual level structural and/or functional connectivity analyses, and the individual results are weighted with the treatment outcome, followed by spatial alignment of the brains between individuals and averaging across subjects. [0032] Finally, the group-level atlases are used for target selection in new patients receiving brain stimulation. Provided that the spatial (anatomical) distribution associated with a given brain disorder is somewhat uniform across patients, the treatment efficacy is expected to be optimal when the treatment atlas is used for target selection.

[0033] For computing the group-level treatment atlases, the individual results are converted to a common standard brain space (“template brain,”“template cortex,” or “template”). Further, when the group-level treatment atlas is applied to a new patient, the treatment atlas in the template space needs to be converted into the space of the individual patient's brain. These conversions require computing transfer maps (“coregistration matrices”) between brains. For best accuracy, these coregistration matrices should use non linear morphing techniques, such as those included in the FreeSurfer software package. However, the disclosure is not limited to using any particular coregistration method, and even standard 3D coregistration methods (e.g., 3D MNI) can be used, albeit these would be expected to result in reduced spatial accuracy.

[0034] There are many possible template brains that are suitable for use with the methods described herein. For example, the template brain my be any individual patient in the present sample, a brain of a healthy subject, a patient outside of the present sample, or a group-level average between brains (such as those often used in neuroimaging packages), so long as the template brain has an explicit cortical surface. Accordingly, the template brain may be generated or may be selected from a one or more existing templates.

[0035] Turning to FIG. 2 a method for generating a TMS-based treatment map according to an embodiment will now be described. In the first column, N patients are being stimulated at a location on each of their respective frontal cortices at a somewhat different location for each patient, since spatial variability is needed) with TMS and their E-fields are computed and displayed on an MRI reconstruction on each patient’s own cortical surface. In the next column, using the computing device of FIG. 9 and software that preserves the gyral/sulcal folding architecture between subjects (such as the non-linear tools in the FreeSurfer® package) each individual’s cortical surface is morphed to a template brain surface. After a certain period of brain stimulation, the treatment efficacy for each patient is determined. For example, if the purpose of the TMS treatment is to reduce drug abuse, then each patient submits to multiple drug tests (during and/or after the TMS treatment). The treatment efficacy in each patient is expressed as an index (e.g., a percentage of clean drug tests in the individual patient), shown in the third column of FIG. 2. In the next column, using the computing device of FIG. 6, a correlation value between the efficacy indices of the patients and the E-field strength of the TMS at that location is computed (e.g., by calculating a Spearman’s rho coefficient). This process may then be repeated for multiple locations on the cortex until correlation indices are generated for each of multiple locations on the cortex. The values of the correlation coefficients are then plotted on a template cortex (last column), color coded as a heat map (e.g., negative correlations will be blue, positive correlations between two thresholds are red/yellow) and thresholded (correlations below a certain value are not shown). Clinicians can then refer to this map to determine the areas to focus on during TMS-based treatment (e.g., the“hot” areas) and where to avoid applying TMS (e.g., the“cold” areas).

[0036] Turning to FIG. 3, further details of the method described in conjunction with FIG. 2 will now be provided. In panel A, the dots show the TMS E-field maxima locations and amplitudes for each patient. Panel B shows the corresponding therapeutic efficacy indices (TEIs) for each patient (% clean urine samples in this case). Visual inspection of Panels A and B does not suggest systematic differences in treatment efficacy between brain areas. Panel C shows the corresponding group-level E-field distribution. The color coding reflects the E- field intensity (yellow = highest, peak at black cross). Panel D shows how the TMS-induced E-fields are not pointlike as in panel A (which shows the individual subject maxima) but have a sizable spatial extent, meaning that the E-fields were spatially overlapping across subjects (see also FIG. 2 second column).

[0037] Turning to FIG. 4 a method for generating a network-based treatment map (“network atlas”) will now be described. Referring to panel A, the best and worst areas from the cortical treatment atlas of FIG. 2 may be used as seeds in a resting-state fMRI connectivity analysis. Here, the best seed is in yellow and the worst in blue. There were 5 regions that showed statistically significant differences for these 2 seeds; areas with positive correlations are shown in red and those with negative correlations in dark green. Referring to panel B, structural connectivity data using probabilistic tractography techniques is added. There were strong connections between the best seed (yellow) and the two regions shown in red. Referring to panel C, the connections from the worst seed (blue) showed a clearly different pattern. Referring to panel D, the structural connectivity was analyzed using deterministic tractography techniques. As before, there were strong connections between the best seed (yellow) and the two regions shown in red. Referring to panel E, the connectivity patterns from the worst seed (blue) were clearly different. In FIG. 4, the left column shows the view from left side of the brain with anterior to the left; and the right column shows the view from front of the brain with left hemisphere on the right.

[0038] For computing network atlases, it is not always necessary to have connectivity data from the same patients that received brain stimulation. Instead, it is possible to use surrogate connectivity data from healthy subjects or patients that did not receive brain stimulation. For example, after computation of the cortical atlas as usual (which requires the individual anatomical MRIs from the patients but does not require connectivity MRI data from these patients), the group-level positive and negative hotspots from the cortical atlas may be used as seeds in group-level connectivity analyses (tractography and/or functional) utilizing surrogate data from, e.g., on-line databases that contain tractography and/or functional MRI connectivity data.

[0039] Different brain stimulation parameters often have different effects on neurons. F or example, TMS pulses given at frequencies (about 1 Hz) tend to suppress excitability, whereas high frequencies (about 20 Hz) increase excitability of the targeted cortical area. Correspondingly, TMS treatment for major depressive disorder (“MDD”) given at 1 Hz vs. 15-25 Hz has opposite local effects. Therefore, using TMS at 1 Hz versus 20 Hz will likely require different treatment atlases. More generally, it is likely that treatment atlases are specific to certain stimulation parameters. Similarly, different brain stimulation techniques may have different effects on neurons. Therefore, the atlases may also be specific for a given brain stimulation technique. However, there may also be applications that bridge several techniques. For example, treatment atlases computed from TMS data could inform where the electrodes in invasive stimulation or scalp electrodes in tDCS/tACS should be placed for best efficacy.

[0040] Clinical diagnoses where the treatment atlases may be useful often contain several different symptoms that may each reflect dysfunction of a different brain area and/or network. Therefore, it is expected that there may be several treatment atlases for some clinical diagnoses, each targeting a specific symptom (e.g., in major depressive disorder, stimulation of one area/network could elevate mood, whereas stimulation of another area/network could relieve anxiety).

[0041] The techniques presented herein differ from currently used techniques used for target selection. For example, in treating depression, the TMS coil is typically placed 50-55 mm anterior to the hand representation in the primary motor cortex, or alternatively, placed using external skull landmarks (e.g., at a certain 10-20 EEG electrode location). Since brain anatomy, including cortical folding patterns and relation of each sulcus and gyrus to external landmarks, varies greatly between individuals, such methods result in targeting different brain areas across subjects, and often the stimulation falls outside the intended regions. Taking the individual anatomy (e.g., from MRI) into account could be used to improve the accuracy of targeting the stimulation to occur with the anatomical borders of DLPFC. The present disclosure utilizes a different idea where the group-level treatment atlas is first computed based on outcomes and then used to determine the most likely optimal target in each new patient.

[0042] It would also be possible to use brain functional imaging (e.g., fMRI, PET, SPECT) to explain how TMS to certain brain regions may have therapeutic effects in a disorder, or to use individual level brain imaging to identify TMS targets in DLPFC, or to study correlations between activations and connectivity patterns vs. different types of brain disorders or symptoms, with the purpose of understanding the underlying mechanisms and finding potential targets for brain stimulation. However, such data and analyses can only reveal correlations between brain functions and clinical symptoms, and as such, cannot inform if the stimulation of the identified brain regions (or networks) will result in better treatment outcomes. The atlases in the present disclosure are built on brain stimulation and treatment outcome data, and therefore provide direct causal evidence that stimulation of the identified targets (and/or networks) is maximally effective.

[0043] The various embodiments described herein are applicable to all brain stimulation techniques, such as non-invasive methods based on electromagnetic induction (TMS/rTMS), non-invasive methods that place electrodes on the scalp (e.g., tDCS/tACS), and invasive techniques (epidural, cortical, or deep brain stimulation). tDCS/tACS devices are non- invasive, relatively affordable, and easily portable. [0044] The various embodiments described herein could be used for target selection in conjunction with the use of tDCS/tACS for therapeutic purposes.

[0045] The various embodiments described herein could inform where electrodes should be placed for invasive brain stimulators for MDD and other brain-related disorders.

[0046] The embodiments described herein have advantages over current techniques, in that data where brain stimulation already occurred is utilized, so that the stimulated location(s) in each brain are known. More specifically, the techniques described herein may use stimulation that has already been carried out with navigator devices and individual MRIs, in which the therapeutic effects were recorded.

[0047] The various embodiments described herein may be used to improve the clinical efficacy of all types of brain stimulation. For invasive brain stimulation techniques, cost efficacy is likely to be increased, because the treatment targets cannot be changed after surgical implantation. For TMS/rTMS and tDCS/tACS, treatment effects will be stronger (more patients will respond, and/or patients that respond may experience better results).

[0048] Further description of various embodiments, including an actual study where a treatment atlas was developed for treating cocaine addiction, is in the following sections:

[0049] Drug use disorders are an epidemic influencing over a billion individuals worldwide with major medical, social, legal, and financial consequences. Among them, cocaine use disorder (CUD) is a leading cause of morbidity and emergency room visits. Currently, there are no pharmacological or brain stimulation approaches approved to treat patients with CUD. These patients are typically treated with psychotherapies and/or off-label pharmacotherapies, whose efficacy is modest and highly variable. Among CUD patients in rehabilitation programs, 70% relapse within 3 months. The development of new, more effective treatments for addictions, including CUD, is critically needed. Precision/experimental medicine approaches that demonstrate target engagement have been suggested to have the best chances of success.

[0050] Addictive behaviors are correlated with significant changes in brain structure and function that develop as a result of drug use. Specifically, recent studies point to changes in frontal areas supporting fear/reward learning and impulse control with prefrontal cortex hypoactivity, which may behaviorally contribute to drug seeking, dependency, and poor ability to suppress cravings and to resist drug cues. Then, it should be possible to counteract these neuronal and behavioral effects by stimulating the appropriate brain area(s). Supporting this idea, animal studies have found that repeated electrical stimulation of the prefrontal cortex may reduce cocaine seeking behaviour and motivation for its consumption. More recently, optogenetic experiments in the rat provided causal proof for the hypoactivity of the prelimbic (PL) area in chronic cocaine addiction, by showing that stimulation of this area reduces compulsive cocaine seeking and use, whereas inhibition of this area has the opposite effect on drug behaviors. The successes in animal models support the idea of translating these findings to the treatment of CUD patients.

[0051] Among brain stimulation methods potentially suitable for treating addictions in humans, repetitive transcranial magnetic stimulation (rTMS) is a safe and non-invasive technique for inducing strong electric currents to create long-term neuroplastic changes. Most previous rTMS studies of cocaine addiction have targeted locations in the lateral prefrontal cortex (PFC). Since the human lateral (and medial) PFC contains areas that are homologous to the rat PL this opens the possibility of translating the success in rats to human therapies. However, in primates and humans, the lateral PFC can be divided into multiple subregions with different connections and functions. For example, some parts of the dorsolateral aspect drive deep mesolimbic regions that initiate motivated behaviors. Particularly relevant for addiction, the lateral PFC has known roles in reward valuation, motivation, and decision making by performing cognitive integration and response inhibition, and the loss of inhibitory control in drug seeking and use has been attributed to its hypoactivity.

[0052] As evident from the above, the lateral PFC is a large and heterogeneous area. It is unknown where in it stimulation should be targeted for best therapeutic efficacy. To address this problem, a brain stimulation treatment atlas (BSTA) for CUD was developed. To do so, data from CUD patients treated with rTMS (N=20, total of 224 rTMS sessions, or 486,797 TMS pulses) aimed at different PFC targets across patients was analyzed, where individual MRIs and recordings of TMS coil locations, along with treatment outcomes (changes in drug use), were recorded. To build the BSTA for CUD, TMS-induced electric field (E-field) modeling tools according to embodiments described herein were utilized and then a map, i.e., an atlas as described herein, was built. The atlas indicates which areas on the cortical surface were associated with the optimal therapeutic response, and which areas, if any, should be avoided. [0053] Further, since rTMS clinical effects depend not only on the target on the cortex (primary activations under the coil), but also on connectivity-based spread to remote areas (secondary activations), the atlas from the brain surface was extended to large-scale networks throughout the brain (FIG. 4). These connectivity analyses revealed large-scale networks associated with different therapeutic responses. The network-level BSTA provides causal evidence of the relationship between therapeutic effects and large-scale networks throughout the brain, hence illuminating the pathophysiology of addiction, as well as mechanisms of action of rTMS therapy when targeting the optimal atlas area.

[0054] RESULTS

[0055] Results of studies using the techniques described herein (and with appropriate reference to the drawings) will now be described.

[0056] Averaged across all drug urine samples within each patient, the therapeutic efficacy index (TEI, proportion of cocaine-free drug urine tests within each patient) varied from 42% to 100% (mean 86%). These TEI values were used to compute a treatment atlas.

[0057] Navigator results. The targets varied between subjects (FIG. 3 A), but all were within the left PFC, and were kept consistent across sessions within subjects. The exact coil placement was a decision made by the clinicians and not driven by a research hypothesis . As expected, the TMS dl/dt value (reflecting stimulator intensity) varied between subjects (mean 60, range 48-74 A/ps) due to the individual motor thresholds (MTs).

[0058] E-field results. In the study, TMS was aimed at different PFC locations between subjects. FIG. 3, panel A shows the locations and amplitudes of the E-field maxima for each individual patient on the cortical surface of a standard brain. The peak E-field amplitudes at the pial surface varied between subjects (mean 139, range 78-179 V/m) mainly reflecting the individual MTs, dl/dt values, and coil-to-cortex distances. FIG. 3, panel B shows the corresponding TEI scores. Visual inspection of panels A and B did not suggest clear differences in treatment efficacy between the cortical sites. FIG. 3, panel C shows the corresponding group-level TMS-induced E-fields. This is useful in that any estimates are only valid in brain areas that were stimulated at sufficiently strong intensities to lead to long-term effects.

[0059] Correlations between E-fields and treatment outcomes. Next, the relationship between the E-field amplitudes and TEI using correlation tests was studied. FIG. 2, second last column, shows the corresponding results for two example cortical locations. FIG. 2, second last column, upper part, shows the values picked from the location of the group-level maximum (black cross in FIG. 3, panel C). This was not correlated with TEI (Spearman rho = 0.27, below the 95th percentile). FIG. 2, second last column, lower part, shows the values picked at the optimal location of the BSTA (best treatment response in FIG. 2, rightmost column). Here, the E-field amplitude was significantly correlated with TEI (Spearman rho = 0.64, above the 95th percentile).

[0060] Based on FIG. 3, panels A and B, one might think that each stimulated brain location had an N=l , which would undermine any statistics. However, the spatial distributions of the E-fields in each individual are quite wide (FIG. 2, second column from the left), meaning that the E-fields in reality overlap across subjects. FIG. 3, panel D shows areas where the TMS-induced E-fields were strong enough to result in long-term neuroplastic changes in at least 10 subjects. Thus, at least half of the patient sample contributed to all key regions of the atlas.

[0061] It should be noted that a negative correlation between the E-field intensity and TEI values (FIG. 2 rightmost column, which may be shown, for example, as a blue area) is also useful information and represents another possible embodiment of a treatment atlas.

[0062] Taken together, these results suggest that the therapeutic response was not driven by E-field intensity alone, and instead was driven by the efficacy of stimulating the optimal target at the dorsolateral-ventrolateral PFC border.

[0063] Cross-validation. To estimate if using the treatment atlas described herein to guide rTMS therapy could improve outcomes in future patients, a cross-validation analysis was conducted. This analysis computed 20 different versions of the atlas where each patient was removed from the sample one at a time, and then the N=l9 atlases were used to predict the therapeutic outcome in the 20th patient (leave-one-out analysis). Specifically, the cortical location with best therapeutic efficacy was estimated separately for each of the 20 iterations, followed by extracting the E-field and TEI values from the 19 patients. The extracted values were then split at the median, and plotted in a E-field versus TEI scatterplot. Finally, researchers tested if the 20th subject's E-field amplitude at this cortical location could predict his/her therapy outcome, as defined by that the individual was in the correct quadrant in the E-field vs. TEI scatterplot. Significance was estimated with permutation analyses (N= 10,000). The cortical location with peak value was recomputed for each permutation to avoid selection bias. The prediction accuracy for the E-field maximum amplitude alone, regardless of location, was at 60% and non-significant. In contrast, for the atlas optimal location, the prediction accuracy was 75%, which was significant at the 95th percentile. The 5 incorrect predictions comprised of 2 false positives and 3 false negatives, resulting in a sensitivity of 80% and specificity of 70%. This suggests that the treatment atlas has predictive power and may be useful for selecting rTMS targets in new patients.

[0064] Connectivity results. rTMS treatment efficacy depends on long-range connectivity from the rTMS surface target with other (cortical and subcortical) brain regions. Researchers therefore estimated connectivity from the cortical surface atlas best and worst locations (FIG. 2, rightmost column) to addiction related circuitry. Since no connectivity data were available from the CUD patients in this example, researchers used surrogate diffusion MRI and resting- state fMRI data from the Human Connectome project recorded at Massachusetts General Hospital (“MGH”) (N=35). FIG. 4 shows a representation of the results, comparing the connectivity from the best (yellow) versus the worst (blue) location in the treatment atlas in FIG. 2 rightmost column. FIG. 4, panel A shows the resting-state fMRI (rs-fMRI) connectivity results, suggesting that these seeds were functionally connected with 5 regions that significantly differed between the best versus the worst seed. Of these 5 areas, 2 (red) were more strongly correlated with the seed for best than the worst therapeutic outcomes. These areas were the lateral orbitofrontal cortex (OFC) and the posterior region in the posterior inferior temporal sulcus (ITS). The remaining 3 areas were more strongly correlated with the seed for the worst than the best therapeutic outcomes (dark green). FIG. 4, panels B and C show the corresponding probabilistic diffusion MRI tractography analyses; these were in accord with the rs-fMRI results, showing a significantly stronger connection from the cortical seed associated with the best treatment outcome to the 2 areas (red) than from the seed associated with the worst treatment outcome (dark green). FIG. 4, panels D and E show the corresponding results for a deterministic tractography analysis of the same data.

[0065] Discussion

[0066] In the present case, the atlas shows which cortical areas should be stimulated for maximal therapeutic efficacy in CUD. The atlas also provides causal proof of the brain locations that are relevant for CUD, and likely, for addictive disorders at large. The optimal location was at the DLPFC-VLPFC border, which is outside (inferior and more lateral) of the typical targets used in clinical rTMS protocols. The fact that the treatment was effective even in some patients where the maximum was not at the optimal location seems to result from peripheral parts of the TMS-induced E-field distribution reaching the optimal target at a sufficiently strong intensity.

[0067] Given the connectivity patterns, it seems likely that the clinical effects were mediated via enhanced executive control. rTMS of this target may thus lead to a similar increase in ability to resist impulsive behaviors in the presence of drug cues and reduced drug use as PL stimulation in rodents.

[0068] Utility of treatment atlases. BSTAs are a precision medicine approach for non- invasive brain stimulation therapies. While brain stimulation is already being used for many indications, for most disorders it is not known where the optimal targets are situated. Treatment atlases address this problem by directly indicating which targets in the human brain yield optimal clinical outcomes and which areas should be avoided. While the results discussed herein are for cocaine addiction, the various methods are directly applicable to other neuropsychiatric disorders.

[0069] Treatment atlases are primarily aimed at improving clinical efficacy. However, as illustrated in this example, they can also provide causal proof of disease pathophysiology and reveal mechanisms through which therapeutic effects of brain stimulation therapies emerge. Treatment atlases may also have transdiagnostic value. For example, the optimal target at the DLPFC-VLPFC border could be explored as a stimulation site for other disorders where executive control is defective.

[0070] The present results suggest that there may be instances where the cortical targets for best vs. worst clinical outcomes are located near to each other. Moreover, while clinicians often apply rTMS for several indications (e.g., addictions) to targets that have been tested for MDD, the present results suggest that at least for CUD, the optimal location is outside of the usually stimulated areas. This underlines the utility of using TMS navigator devices and individual level anatomy and, more generally, the potential value of building treatment atlases to disorders that are treated with brain stimulation.

[0071] Conclusions [0072] The above-described study introduces brain stimulation treatment atlases, which represent a novel precision medicine approach for rTMS. Specific to CUD, the atlas suggests a new location at the DLPFC-VLPFC border that could be targeted to improve clinical efficacy, implicates links between brain networks and the disorder, and offers a mechanistic explanation why the rTMS treatment of this target was effective. Beyond CUD, the present study paves way for constructing similar atlases for other addictions and neuropsychiatric disorders, with the goals of enhancing the clinical efficacy of rTMS and providing causal proof of the affected brain areas.

[0073] Referring to FIG. 5, a TMS system is shown. Because we know the TMS coil geometry, the TMS intensity (dl/dt) from the brain stimulation device, each subject’s head volume conductor properties, and where the TMS coil is relative to the head when the TMS coil is fired, the TMS-induced E-field inside the skull can be estimated using electromagnetic forward computations.

[0074] FIG. 6 illustrates a basic hardware architecture of a computing device that may be used to implement the various methods described herein. The computing device, generally labelled 600, includes a processor 602 (e.g., a microprocessor, a controller, an application- specific integrated circuit (“ASIC”), a field-programmable gate array (“FPGA”), or a system on chip (“SoC”)). The computing device 600 also includes computer-readable medium 604 (e.g., one or more types of non-transitory memory, such as flash memory, magnetic memory, or dynamic random access memory), a display 606, and one or more communication interfaces 608 (e.g., network communication interfaces (wired or wireless) and user interfaces). The computer-readable medium 604 may be or include a non-transitory computer- readable medium having stored thereon computer executable instructions for carrying out any of the methods described herein.

[0075] It should be noted that the techniques described herein may be applied to any portion of the brain, and not just the cortex. For example, the network-level atlas described herein extends from the cortex (at the brain surface) to other locations that include non-cortex parts. Furthermore, the techniques described herein may be used with other types of electrically-based stimulation techniques, and not just TMS.

[0076] It should be understood that the exemplary embodiments described herein should be considered in a descriptive sense only and not for purposes of limitation. Descriptions of features or aspects within each embodiment should typically be considered as available for other similar features or aspects in other embodiments. It will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from their spirit and scope as set forth in the following claims. For example, the actions described herein can be reordered in ways that will be apparent to those of skill in the art.

Claims

CLAIMS What is claimed is:

1. A method for generating a cortical surface transcranial magnetic stimulation (TMS) treatment atlas, the method comprising:

determining an electric field induced by TMS of a plurality of locations of a cortex of each of a plurality of patients, resulting in a plurality of electric field values;

determining a treatment efficacy for each of the plurality of patients, resulting in a plurality of efficacy values;

calculating a correlation between the plurality of electrical field values and the plurality of efficacy values, resulting in a plurality of correlation values comprising a correlation value for each of a plurality of cortical surface locations; and

generating the cortical surface TMS treatment atlas based on the plurality of correlation values.

2. The method of claim 1, wherein determining a treatment efficacy for each of the plurality of patients comprises determining a percentage of clean drug tests for each of the plurality of patients.

3. The method of claim 1, wherein calculating a correlation between the plurality of electric field values and the plurality of efficacy values comprises calculating a correlation coefficient for the plurality of electric field values and the plurality of efficacy values, separately at each of the plurality of cortical locations.

4. The method of claim 1, wherein generating the cortex TMS treatment atlas based on the plurality of correlation values comprises generating a color coded image of a cortex in which each of a plurality of colors represents a specific correlation value.

5. The method of claim 1, further comprising:

obtaining an image of the cortices of the plurality of patients;

computing coregistration matrices between the cortices of the plurality of patients; wherein generating the cortex TMS treatment atlas based on the plurality of correlation values comprises plotting correlation coefficients on a template image of a cortex.

6. A computing device configured to carry out actions comprising:

measuring an electrical field induced by TMS of a location of a cortex of each of a plurality of patients, resulting in a plurality of electrical field values;

determining a treatment efficacy for each of the plurality of patients for each TMS- induced electrical field, resulting in a plurality of efficacy values;

calculating a correlation between the plurality of electric field values and the plurality of efficacy values, resulting in a plurality of correlation values comprising a correlation value for each of a plurality of cortical surface locations; and

generating a cortex TMS treatment atlas based on the plurality of correlation values.

7. The computing device of claim 6, wherein determining a treatment efficacy for each of the plurality of patients comprises determining a percentage of clean drug tests for each of the plurality of patients.

8. The computing device of claim 6, wherein calculating a correlation between the plurality of electrical field values and the plurality of efficacy values comprises calculating a correlation coefficient for the plurality of electric field values and the plurality of efficacy values.

9. The computing device of claim 6, wherein generating the cortex TMS treatment atlas based on the plurality of correlation values comprises generating a color coded image of a cortex in which each of a plurality of colors represents a specific correlation value.

10. The computing device of claim 6, configured to carry out further actions comprising: obtaining an image of the cortices of the plurality of patients;

computing coregistration matrices between the cortices of the plurality of patients; creating a normalized image of a cortex based on the coregistration matrices, wherein generating a cortex TMS treatment atlas based on the plurality of correlation values comprises plotting correlation coefficients on a template cortex image.

11. A non-transitory computer-readable medium having stored thereon computer- executable instructions for carry out a method for generating a cortex transcranial magnetic stimulation (TMS) treatment atlas, the method comprising:

determining an electric field induced by TMS of a location of a cortex of each of a plurality of patients, resulting in a plurality of electric field values;

generating a cortex TMS treatment atlas based on the correlation values.

12. The computer-readable medium of claim 11 , wherein determining a treatment efficacy for each of the plurality of patients comprises determining a percentage of clean drug tests for each of the plurality of patients.

13. The computer-readable medium of claim 11, wherein calculating a correlation between the plurality of electrical field values and the plurality of efficacy values comprises calculating a correlation coefficient for the plurality of electric field values and the plurality of efficacy values.

14. The computer-readable medium of claim 11, wherein generating the cortex TMS treatment atlas based on the plurality of correlation values comprises generating a color coded image of a cortex in which each of a plurality of colors represents a specific correlation value.

15. The computer-readable medium of claim 11 , further comprising:

obtaining an image of the cortices of the plurality of patients;

computing coregistration matrices between the cortices of the plurality of patients; wherein generating a cortex TMS treatment atlas based on the correlation value comprises plotting the plurality of correlation values on a template image.