Method for focused transcranial electrical current stimulation
Patent Information
- Application Number
- IN202417032103
- Authority / Receiving Office
- IN · IN
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2021-09-24
- Filing Date
- 2024-04-23
- Publication Date
- 2026-09-30
AI Technical Summary
Existing electrode placement algorithms for Transcranial Electrical Stimulation (TES) fail to effectively utilize the nonlinear nature of neurons, particularly the thresholding phenomenon, leading to inefficient focal stimulation and potential unnecessary neural activation in non-target regions due to overly restrictive field minimization in the cancel region.
A method that optimizes electrode placement by using a loss function that only penalizes electric fields above a specified threshold in the cancel region, allowing for non-zero fields below the stimulation threshold, thereby enabling more focused neural activation while maintaining fields in the cancel region below the stimulation threshold.
This approach provides a 20% reduction in stimulated area compared to prior art methods like Directionally Constrained Maximization (DCM), achieving more focused neural stimulation while ensuring safe and effective current delivery.
Abstract
Description
Background of the Invention
[0003] Transcranial Electrical Stimulation (TES) is a general term used for either stimulatingor modulating the neural activity of the brain using electrical currents delivered fromelectrodes placed at the scalp. TES is a promising therapeutic tool for manyneurological diseases and disorders, including, but not limited to, clinical depression,chronic pain, Parkinson's, and many more.
[0004] Traditionally, TES is performed using two large electrodes (a cathode and an anode)to induce electric current into the brain. This arrangement of electrodes creates adiffused field in the brain causing a widespread stimulation / modulation, even whenonly a focal stimulation is desired. In recent years, advances have been usingmultielectrode TES arrangements and current patterns that create fields which onlyhave significant amplitude in a region of interest. For example, the motor orsomatosensory cortex. Such multielectrode arrangements can improve focality,which can improve the clinical efficacy of TES-based therapies and help developmore precise brain-machine interfaces.
[0005] The problem of designing such electrode montages and the injected currentamplitudes is a problem of electrode placement. Fundamentally, there are twoprimary challenges that make the problem of electrode placement hard. First, lawsof physics dictate that the currents generated by the electrodes at the scalp diffuseas they travel through the layers of head (i.e., the scalp, skull and cerebrospinalfluid), making it hard to constrain the area of stimulation. Second, to avoid pain ortissue damage in the scalp, the current at each electrode, and the total current thatis injected into the scalp, cannot be too high, making it hard to create highamplitude electric fields in the brain.
[0006] Existing works formulate the problem of electrode placement as asearch / optimization problem. They divide the brain into two regions: a region ofinterest where it is desirable that the field be high, referred to as the "focus region,and the remaining region of "non-interest" where it is desirable that field is low,referred to as the "cancel region".
[0007] The optimization problem is set up in such a way that its solution is an electrodemontage that produces the smallest electric field (i.e., as close to zero as possible) inthe cancel region, given a desired amplitude of electric field in the focus region.These algorithms succeed in producing focused fields, and, consequently, focusedstimulation, in simplified models (ranging from spherical head models to real-headmodels). While the existing electrode placement algorithms are successful increating focused electric fields, they do not explicitly take advantage of the nonlinear nature of neurons, especially the thresholding phenomenon of neurons,wherein neurons only fire when the stimulating field is above a certain threshold.For focused stimulation, one does not require electric fields to be minimized in thecancel region. Instead, it is sufficient that it be below the stimulation threshold(which depends on the stimulating waveform but is fixed for a given waveform).
[0008] Thus, there is a need for a method of obtaining the desired electrical field in thefocus region, while keeping the field in the cancel region below the stimulationthreshold, such as to not cause the neurons in the cancel region to fire.Summary of the Invention
[0009] This Summary is provided to introduce a selection of concepts in a simplified formthat are further described below in the Detailed Description. This Summary is notintended to identify key features or essential features of the claimed subject matter,nor is it intended as an aid in determining the scope of the claimed subject matter.
[0010] Disclosed herein is a method for optimizing electrode placement that directlyexploits the thresholding phenomenon of neurons. The method employs a lossfunction which only becomes non-zero when the electric field is above a userspecified threshold (Etol) in the cancel region, thereby allowing for fields which canhave significant non-zero current in the cancel region, but still provide more focusedneural activation.
[0011] The improvement in performance provided by the invention is quantified bycomparison with a prior art method referred to as the Directionally ConstrainedMaximization (DCM) algorithm, in simulated (spherical head) models. DCM has beenshown to subsume most of the existing TES electrode placement algorithmsproposed in the prior art, and therefore is a good candidate for comparison.
[0012] The disclosed method performs strictly better than DCM in terms of the area ofstimulation (approximated by area of the electric field above a certain pre-definedthreshold), providing as much as a 20% reduction in stimulated area when theallowed injected current at each electrode is moderately high (see Fig. 2).
[0013] The disclosed method specifies an optimization framework that keeps the fieldsoutside of the focus region below a specified threshold (Etol, which is the"tolerable" level of the field). The disclosed method is first specified, with an explicitexplanation of why prior art algorithms minimize the fields outside the region offocus, instead of simply keeping the fields them below a threshold.Brief Description of the Drawings
[0014] The patent or application file contains at least one drawing executed in color. Copiesof this patent or patent application publication with color drawing(s) will beprovided by the Office upon request and payment of the necessary fee.
[0015] In the drawings, like reference characters generally refer to the same partsthroughout the different views. In the following description, various embodimentsof the present invention are described with reference to the following drawings, inwhich:
[0016] FIG. 1 shows a comparison between the currently disclosed method and the prior artDCM method, showing differences in the current density and the activated regionsgenerated by the respective methods.Detailed Description
[0017] The present invention is directed a method for arranging the electrodes andapplying power such as to maximize the field in the focus regions while maintainingthe field in the cancel region below the stimulation threshold. The method exploitsthe thresholding phenomenon of neurons by designing multi-electrodearrangements which allow the electric fields outside the target region to be nonzero, but still below the stimulation threshold.
[0018] Assume that (symbol) locations are chosen on the scalp, where electrodes are to bereplaced. Let I = Equation be an n-dimensional vector, where representsthe current injected at the jthlocation. If no electrode is placed at the jth electrodelocation, then the corresponding = 0. Therefore, the vector I completelydescribes an electrode montage, and the problem of electrode placement can bereduced to finding the "optimal" I. A frequent assumption made in electrodeplacement algorithms is that the physics of the electrode placement problem isgoverned by the quasi-static Maxwell equations. Therefore, given the conductivitiesand geometries of all the layers of the head, namely the scalp, the skull, thecerebrospinal fluid (CFS) and the white and grey matter of the brain, and the current(injected by the electrodes) at each of the location, the field in the brain can becalculated using the Laplace equation. The solution of the Laplace equation can befound by discretizing the head model and solving it using a numerical technique suchas Finite Element Method (FEM). This discretization yields the following linearsystem:Equation (1)where: Equation is the number of voxels into which the head model is discretized;is the electric field induced in the head in x, y and z directions (i.e., 3M); andT is the transfer matrix / function that maps the injected currents to the inducedelectric field EIE1.
[0019] T only depends on the conductivity and geometry of the head model. Using theabove formulation, the problem of electrode placement can be conceptualized astwo sub-problems: (a) Let the focus region be denoted by F and the cancel region bedenoted by C. Then DCM seeks to maximize the current / electric field intensity alongthe direction Equation in the focus region, while minimizing the fieldmagnitude in the cancel region through the following optimization:Equation (2)
[0020] The first term Equation is a discrete approximation of the integral of the electricfield along the direction in the focus region Equation Here, D (symbol) Equationis the vector denoting the direction along which the electric field for eachvoxel in F should be projected. ΓF 3 x 3|F| (where |F| is the number of voxelspresent in the focus region) is the matrix that results due to the integrationoperation, and its elements are just volumes of each voxel in the focus region. Equation is the matrix constructed from sub-sampling the rows of the transfer matrixT, which correspond to the voxels lying in the focus region F. Therefore, TI is theelectric field EIfor each voxel in the focus region. The optimization problem seeks tomaximize the integral of electric field along the direction d in the focus region, whichis equivalent to maximizing the average intensity of the electric field in the focusregion along d.
[0021] The quadratic term EquationI represents the energy of the electric field in thecancel region c. The particular form of the quadratic term again stems fromdiscretizing the integral for calculating the energy of the field in the cancel region:Equation
[0022] Here, TEquatin (where is the number of voxels present in the cancel region)is the matrix constructed from sub-sampling the rows of the transfer matrix TT, whichcorrespond to the voxels lying in the cancel region C, and ΓcC (symbol) 3|CC| x 3|CC| is thematrix that results due to the integration operation, and is a diagonal matrix withthe diagonal elements being just volumes of each voxel in the cancel region.Therefore, the constraint IEquation bounds the electric energy below acertain threshold.
[0023] The constraint Equation restricts the total current being injected from theelectrode montage to and the constraint EquationISAFE restricts the maximumcurrent injected per electrode to Equation, to ensure safe stimulation. The constraint EquationI = 0 is present for satisfying Kirchhoff's law (i.e., the amount of current goinginto the head should be equal to the amount of current going out). For the sake ofbrevity, herein, Equationis denoted by aF and IEquationI is denoted by EquationAc(which also defines Ac).
[0024] As previously discussed, it is not obvious from the formulation described in Eq. (2)that DCM minimizes the electric field in the cancel region. Therefore, to prove thatDCM minimizes the electric field in the cancel region, the following optimizationformulation is defined:Equation (3)
[0025] Solving Eq. (3) is equivalent to solving Eq. (2). As can be seen from Eq. (3), the idealoptimal I for DCM would be the one which satisfies Equation.Therefore, DCM and by extension all the algorithms subsumed under it overpenalizes the electric field in the cancel region and do not try to take advantage ofthe thresholding phenomenon of neurons.
[0026] The disclosed method provides an optimization framework that explicitly takesadvantage of the thresholding phenomenon of neurons. The optimizationframework is given by the equation:IEquation (4)where: Equation is a vector consisting of all of the diagonal elements of ΓC,max(.) is applied element-wise to each of the vectors; and Equation, whose elements specifies the level of electric field thatis tolerable in each voxel of the cancel region.
[0027] A common choice for EquationEis the vector of all ones,and ETOL is some global level of electric field that is tolerable. For the sake of brevity,we will denote the loss function in Eq. (4) is denoted as:Equation (5)
[0028] The solution (i.e., the electrode montage) produced by the DCM is equivalent to thesolution derived by the disclosed method (derived with the same constraints onelectrode currents on the scalp and the electric field created in the focus region)when EquationE is very low zero or near zero. If ETOL is chosen to be very low, thedifference between neural activation between DCM and the disclosed method isfunctionally equivalent. The (symbol) term in Eq. (2) indicates the desired field in the focusregion, as does the term in Eq. (4). The (symbol) term does not dictate the tolerablestrength of the field in the cancel region (Equation).
[0029] The only difference between Eq. (4) and Eq. (3) is in their loss functions EquationIfor the disclosed method and an l2 norm constraint Equation in case of DCM-typeapproaches. The role of the constraints in Eq. (4) is the same as it was in DCM.
[0030] LH(I) can be thought of as a discretized version of the following integral:Equation
[0031] LH(I) will have a non-zero value only if any component of Equationwhere >,< represent elementwise inequalities. Hence, LH(I) only penalizesthe electric fields which violate their upper or lower thresholds in the cancel region,thereby allowing fields having non-zero (albeit below pre-defined thresholds)amplitudes in the cancel region. In contrast, DCM-type approaches minimize theelectric field and bring the fields as close to 0 as possible. Because of this, DCM-typeapproaches are overly-restrictive in that they disallow fields which have significantnon-zero electric field (but below the stimulation threshold of neurons outside ofthe target region, even if they provide better focus. In contrast, the disclosedmethod does not penalize such electric fields, and can actually find such fields whichcan lead to potentially more focused neural stimulation.
[0032] This is illustrated in FIGS. 1A-1D. FIG. 1A shows a current density generated by DCMat the upper hemisphere of the shell at the depth of 1.3 cm projected onto the x,y yplane, while FIG. 1B shows the current density generated by the disclosed method atthe upper hemisphere of the shell at the depth of 1.3 cm projected onto the xx,y yplane. FIG. 1C shows the corresponding region of the electric field shown in FIG. 1Athat leads to neural stimulation for DCM (approximated by area above 80% of themaximum field), while FIG. 1D shows the corresponding region of the electric fieldshown in FIG. 1B that leads to neural stimulation for the disclosed method(approximated by area above 80% of the maximum field).
[0033] A new approach for electrode placement has been described that explicitly exploitsthe thresholding phenomenon of neurons to achieve focal neural stimulation. A Incomparison with DCM, which subsumes a large class of electrode placementalgorithms, the invention provides significant gains in terms of focality.
[0034] The neural response of the brain to transcranial fields is approximated by athresholding function. While it is commonly believed that, to a first order, the neuralresponse can be approximated by a thresholding function (e.g., for fixed waveformshape, and changing amplitude), actual neural responses are quite complex. Forpurposes of explanation of the invention, only one neuron-type is assumed to bepresent. This best approximates gyri in the cortex where cortical columns have afairly regular structure with similar cell-types. There is known to be significantvariability in the thresholds of neurons (across types, and even within the sametype), which can make choosing the threshold for the disclosed method difficult, as ahigh threshold might lead to spurious stimulation in the cancel region, and a lowthreshold might lose the advantages offered by the disclosed method. Additionally,even though electric fields might be low enough to not cause neural stimulation,they still might modulate the activity of neurons, so care might be needed whileusing the disclosed method. However, even using the simplest of non-linear neuralresponse (i.e. thresholding), significant gains was realized compared to existingalgorithms. This shows the potential of incorporating knowledge of neuron-typesand neural dynamics while designing TES electrode placement algorithms. Neuronsare highly non-linear systems, which makes them hard to analyze but at the sametime allows creative strategies to harness this non-linearity to potentially improveperformance.
[0035] As would be realized by one of skill in the art, the disclosed systems and methodsdescribed herein can be implemented by a system comprising a processor andmemory, storing software that, when executed by the processor, performs thefunctions comprising the method. For example, the optimization method can beimplemented by software executing on a processor.
[0036] As would further be realized by one of skill in the art, many variations onimplementations discussed herein which fall within the scope of the invention arepossible. Specifically, many variations of the parameters, the components used, andtheir arrangement could be used to obtain similar results. The invention is notmeant to be limited to the particular exemplary model disclosed herein. Moreover,it is to be understood that the features of the various embodiments describedherein were not mutually exclusive and can exist in various combinations andpermutations, even if such combinations or permutations were not made expressherein, without departing from the spirit and scope of the invention. Accordingly,the method and apparatus disclosed herein are not to be taken as limitations on theinvention but as an illustration thereof. The scope of the invention is defined by theclaims (and their equivalents) which follow.
Claims
1. A method for optimizing electrode placement for transcranial electrical current stimulation of a brain comprising: placing a plurality of electrodes at a plurality of locations on a scalp; discretizing the brain in to a plurality of voxels; mapping, for each voxel, an electric field induced in the voxel by current applied to each of the plurality of electrodes; identifying a focus region comprising a first set of voxels and a cancel region comprising a second set of voxels; optimizing placement of the plurality of electrodes to maximize the induced electric field in the focus region while maintaining the electric field induced in the cancel region below a predetermined threshold.
2. The method of claim 1 wherein the placement of the plurality of electrodes is represented by an N-dimensional vector wherein each dimension of the vector represents a current applied at a single electrode location; wherein N is equal to the number of electrodes in the plurality of electrodes.
3. The method of claim 2 wherein the N-dimensional vector is optimized by a loss function that penalizes electric fields in voxels in the cancel region that violate a predetermined upper threshold.
4. The method of claim 2 wherein loss function further penalizes electric fields in the cancel region that violate a predetermined lower threshold.
5. The method of claim 1 wherein the predetermined threshold is non-zero.
6. The method of claim 1 wherein the electric field induced in the cancel region is not minimized.
7. The method of claim 1 wherein the predetermined threshold is a maximum electric field in the cancel region that is below the stimulation threshold of neurons in the cancel region.
8. The method of claim 7 wherein the stimulation threshold is dependent on the stimulation waveform.
9. The method of claim 1 wherein the predetermined threshold is user-specified.
10. The method of claim 1 wherein a transfer function maps the current injected at each electrode location to the electric field induced in each voxel and further wherein the transfer function is dependent on the conductivity and geometry of a head model containing the brain.
