Virtual brain twin framework

Abstract

In the virtual brain twin framework, the propagation network is segmented into nodes that represent grey matter areas in the brain. Each node has its own time-resolved dynamics, which are influenced by both local and global connections with other nodes in the network. The distance-based approach is used to allocate weights rendering short range connectivity, while long-range tract-based connectivity is allocated using a weight rendering method. Differentiated conduction velocities are also assigned to long-range and short-range connectivities, which allows for more accurate simulation of neural activity and its propagation within the brain.

Description

[0001] Description

[0002] Title of Invention: Virtual brain twin framework

[0003] Technical field

[0004] [1] The technical field of the invention is brain function simulation, and more specifically patient based brain modelling by customization with measured neuroimaging data, adapted for different use cases and particularly for brain disease study with the purpose of treatment.

[0005] Background art

[0006] [2] US2015206051 A1 discloses a virtual brain model combining two networks of nodes: a local network, representing short-range connections between neighbouring neurons represented by local nodes, and a global network representing long-range connections between large groups of neurons, each group being reduced to one global node. Nodes are distributed over a standard brain atlas according to a typical and conventional approach. The brain regions are used to carry out a computer-implemented method using computation in the local network and computation in the global network, thus providing a simulation of the propagation of an activity over the whole brain.

[0007] [3] In the prior art, brain regions differ from one subject to another one. Atlases are used to define the areas globally, but do not provide subject-specific locations of the areas. Tests and measures are necessary to define the subject specific areas in a sufficiently accurate way. Otherwise, a model that is not customized to each patient cannot provide any usable outputs.

[0008] [4] Some references are used in this description:

[0009] [5] Dale et al. 1999 : Dale AM, Fischl B, Sereno Ml. Cortical surface-based analysis. I. Segmentation and surface reconstruction. Neuroimage. 1999 Feb;9(2): 179-94. doi: 10.1006 / nimg.1998.0395. PMID: 9931268.

[0010] [6] - Jenkinson et al. 2012 : Mark Jenkinson, Christian F. Beckmann, Timothy E.J. Behrens, Mark W. Woolrich, Stephen M. Smith, FSL, NeuroImage, Volume 62, Issue 2, 2012, Pages 782-790, ISSN 1053-8119, https: / / doi.Org / 10.1016 / j.neuroimage.2011.09.015.

[0011] [7] - Tournier et al. 2019 : Tournier JD, Smith R, Raffelt D, Tabbara R, Dhollander T, Pietsch M, Christiaens D, Jeurissen B, Yeh CH, Connelly A. MRtrix3: A fast, flexible and open software framework for medical image processing and visualisation. Neuroimage. 2019 Nov 15:202:116137. doi: 10.1016 / j.neuroimage.2019.116137. Epub 2019 Aug 29. PMID: 31473352. [8] - Keren et al. 2016 : Keren, H., & Marom, S. (2016). Long-range synchrony and emergence of neural reentry. Scientific Reports, 6(1 ), 1-10. https: / / doi.Org / 10.1038 / srep36837

[0012] [9] - Shamim et al. 2022 : Shamim D, Nwabueze O, Uysal U. Beyond Resection: Neuromodulation and Minimally Invasive Epilepsy Surgery. Noro Psikiyatr Ars. 2022 Dec 16;59(Suppl 1):S81-S90. doi: 10.29399 / npa.28181 . PMID: 36578991 ; PMCID: PMC9767135.

[0013]

[0010] - Osorio et al. 2009 : Osorio I, Frei MG. Seizure abatement with single de pulses: is phase resetting at play? Int J Neural Syst. 2009 Jun; 19(3): 149-56. doi:

[0014] 10.1142 / S0129065709001926. PMID: 19575505.

[0015]

[0011] - Motamedi et al. 2002 : Motamedi GK, Lesser RP, Miglioretti DL, Mizuno- Matsumoto Y, Gordon B, Webber WR, Jackson DC, Sepkuty JP, Crone NE. Optimizing parameters for terminating cortical afterdischarges with pulse stimulation. Epilepsia. 2002 Aug;43(8):836-46. doi: 10.1046 / j.1528-1157.2002.24901.x. Erratum in: Epilepsia 2002 Nov;43(11 ):1441. PMID: 12181002.

[0016] Technical problem

[0017]

[0012] A purpose of the invention is to propose a novel virtual brain twin framework, which can provide high resolution and outperform existing computational frameworks without the burden of accurate customization of brain regions.

[0018]

[0013] A first object of the invention is a method for modelling a primate brain for simulating the propagation of a neural activity represented by a time-resolved dynamics, characterized in that the method uses a propagation network made of nodes connected using long-range and short-range weighted connections, segmenting the grey matter of the brain, network in which an activity of each node results from its own time-resolved dynamics, superimposed with local and global contributions from other nodes and conveyed through a propagation network, the method comprises the steps of:

[0019] - reconstructing a cortical surface of the primate brain and obtaining a segmentation of a subcortical grey matter of said primate brain, resulting in a meshed surface discretised into mesh elements,

[0020] - modelling each mesh element as a node of the propagation network,

[0021] - estimating single tracts ensuring white matter connectivity, using previously recorded brain images,

[0022] - determining global contribution to each node from the other nodes using intersections of estimated single tracts with the meshed surface, allowing to define long-range weights rendering long range tract-based connectivity between said node and other nodes, - determining local contribution to each node from the other nodes by defining short- range weights rendering short range connectivity between said node and other nodes and

[0023] - assigning differentiated conduction velocities to long range and short range connectivities.

[0024]

[0014] In the context of the invention, a “rendering long-range tract-based or short-range connectivity between a node and other nodes” can mean a numeric coefficient applicable to a value propagated between the two nodes of any pair of nodes, whether said value depicts a biological phenomenon or an abstract representation, like a state variable.

[0025]

[0015] The invention utilizes measured neuroimaging data to create a customizable model that can be adapted for different use cases. This approach gets rid of traditional brain atlases which define brain regions globally but do not provide specific locations of these regions for each individual patient.

[0026]

[0016] According to the invention, the propagation network renders two connectivities: the neighbour connectivity where nodes are interfering according to surface conductivity, and the global connectivity, where all nodes are interfering according to long-distance tract-based connectivity.

[0027]

[0017] Hence, the propagation network of the invention is a cumulation of local coupling and global coupling. The structure of the entire model can be represented by a sum of three terms: the dynamics that is linked to a simulated disease or an expected behaviour of a brain cell, a term for global coupling and a term for local coupling.

[0028]

[0018] This propagation network represents a functional improvement over traditional brain atlases which define brain regions globally but do not provide specific locations of these regions for each individual patient.

[0029]

[0019] According to a particular embodiment of the modelling method, the propagation network is defined according to the geometry of the cortex and subcortex and the single tracts are constructed by an image processing pipeline based on specialised software toolboxes adapted to:

[0030] - construct the meshed surface as a triangulated surface, wherein single mesh elements represent areas of 0.5 mm2 to 1.2 mm2, preferably 0.8 mm2, and wherein subcortical nuclei are represented by volumetric grids,

[0031] - process diffusion weighted MR images to estimate white matter connectivity.

[0032]

[0020] According to a particular embodiment of the modelling method, the weights rendering long range tract-based connectivity are inputted in a global connectivity matrix.

[0033]

[0021] According to a particular embodiment of the modelling method, a short-range weight depends on the cortical thickness at the node.

[0022] According to a particular embodiment of the modelling method, a short-range weight decreases exponentially with increasing distance between nodes.

[0034]

[0023] According to a particular embodiment of the modelling method, the differentiated conduction velocities consist in one average long range conduction velocity for long range connectivities and one average short range conduction velocity for short range connectivities.

[0035]

[0024] These values are specific to each individual patient, as they take into account the unique structure and anatomy of its brain.

[0036]

[0025] The role of differentiated conduction velocities in the model is to provide a more accurate representation of how neural signals travel through different types of connections within the brain.

[0037]

[0026] According to a particular embodiment of the invention, the method further comprises the step of implementing an activity threshold for at least one connectivity selected in the group consisting of the short range and the long range connectivities, and neutralizing the propagation according to the selected connectivity when the node activity is under the threshold.

[0038]

[0027] The method of modelling a primate brain using this framework is advantageous for its ability to provide a more detailed and customized representation of the brain's connectivity, which allows for a more realistic simulation of neural activity and its propagation in individual patients. More specifically, this approach provides an improvement by enabling long-range connections at the level of a node, rather than aggregating information at the level of a region.

[0039]

[0028] This framework also enables support for phenomena such as travelling waves and can be used to guide brain stimulation and new methods to stop epileptic seizures.

[0040]

[0029] Then, another object of the invention is a method for modelling an epileptic brain, characterized in that an epileptogenic model is introduced as the time-resolved dynamics in the modelling method.

[0041]

[0030] The epileptogenic model is based on the idea that an epileptic brain has a specific time-resolved dynamics, which differs from that of a healthy brain. This means that the activity patterns in the brain during an epileptic seizure are different from those during normal brain activity.

[0042]

[0031] The virtual brain twin framework uses the epileptogenic model to simulate the propagation of neural activity in the brain, which is useful for understanding how different areas of the brain work together to cause seizures and how they can be influenced by countermeasures. By simulating the propagation of neural activity, the virtual brain twin framework provides an accurate representation of how the brain responds to different types of countermeasures, including electrical and surgical therapies.

[0032] For instance, the epileptogenic model can be used to test brain stimulation for stopping seizures. The model simulates the effects of different stimulation parameters on neural activity in patients with epilepsy, allowing researchers to identify the most effective stimulation parameters for each patient. This can help to improve the accuracy and effectiveness of brain stimulation therapies for patients with epilepsy.

[0043]

[0033] Hence, the method allows for testing countermeasures on an epileptic brain via stimulation to suppress re-entry and stop seizures.

[0044]

[0034] Alternatively, another object of the invention is a method for testing another type of countermeasure on an epileptic brain, characterized in that the method includes a step of testing the effect of the destruction of at least one long-range or short-range weighted connection on the propagation when implementing the method.

[0045]

[0035] This method for testing countermeasures on an epileptic brain involves simulating the destruction of at least one long-range or short-range weighted connection to evaluate the effectiveness of different treatments. The simulation allows for the evaluation of the removal of any specific connections on the propagation of neural activity in the brain. This provides valuable information on the potential effects of these treatments and helps to guide the development of more effective therapies for epilepsy patients.

[0046]

[0036] In a nutshell, the use of the virtual brain twin framework of the invention, capable of high resolution and accurate customization, enables a more realistic simulation of neural activity and its propagation, allowing for a better understanding of the impact of different countermeasures on a diseased brain. This allows for the study of how different areas of the brain work together to cause seizures and how they can be influenced by stimulation or surgery.

[0047]

[0037] Another object of the invention is a computer program product for modelling a primate brain, characterized in that it comprises instructions which, when the program is executed by a computer, cause the computer to perform at least one of the above-described methods.

[0048]

[0038] Another object of the invention is a data processing system, characterized in that it comprises a processor configured to perform at least one of the above-described methods.

[0049]

[0039] Another object of the invention is a computer-readable storage medium, characterized in that it comprises instructions which, when executed by a computer, cause the computer to perform at least one of the above-described methods.

[0050]

[0040] The storage medium comprising the computer program refers to a physical device or medium used to store digital information, such as software programs, data files, and other types of digital content. In the context of this disclosure application, the storage medium would likely refer to a computer-readable storage medium that contains instructions for performing the methods disclosed in the disclosure. These instructions would be executed by a processor or computer system to perform the various steps outlined in the method for modelling a primate brain and other related methods, such as modelling an epileptic brain and testing countermeasures. The storage medium may be a CD-ROM, DVD, USB drive, hard drive, or any other type of digital storage device that can store and retrieve data. It is functional to note that the storage medium does not refer to the physical computer system itself, but rather the medium used to store the program that runs on the computer system.

[0051]

[0041] The virtual brain twin framework proposed in this disclosure has numerous applications, including guiding brain stimulation and stopping epileptic seizures. The ability to customize the model using measured neuroimaging data for different use cases and brain diseases makes it a valuable tool for researchers and clinicians alike.

[0052] Brief description of drawings

[0053]

[0042] The invention will be better understood with the non-limiting examples disclosed hereafter. Drawings illustrate these examples:

[0054] ■ [Fig- 1] figure 1 illustrates the construction of a high-resolution connectome, by calculating intersections between fibre tracts and the cortical surface,

[0055] - [Fig. 2] figure 2 describes a high-resolution connectivity matrix,

[0056] - [Fig. 3] figure 3 shows a simulated epileptogenic zone in the brain,

[0057] ■ [Fig. 4] figure 4 shows snapshots of the space-time evolution of the simulated epileptic seizure,

[0058] - [Fig. 5] figure 5 shows an example of surgery-induced white matter lesions and their influence on tract density,

[0059] - [Fig. 6] figure 6 depicts time-space plots of simulated seizures without and with surgery- induced lesions,

[0060] ■ [Fig. 7] figure 7 shows electrical field estimation after virtual implantation of contacts for electrical stimulation,

[0061] - [Fig. 8] figure 8 illustrates terminating seizure activity using electrical stimulation,

[0062] - [Fig. 9] figure 9 provides an example of the volumetric grid structure in the subcortex, with zoomed in representation,

[0063] - [Fig. 10] figure 10 is another example focusing on the temporal lobe for epilepsy simulation,

[0064] - [Fig. 11] figure 11 is a snapshot of neural activity simulated on the temporal lobe section of the brain.

[0065]

[0043] First, the method for reconstructing a virtual brain is described. To construct the virtual brain model, data from magnetic resonance (MR) images are used in order to accurately render cortical grey matter as well as white matter tracts ensuring connectivity.

[0066]

[0044] In the model, neurons with their cell bodies in the cortex can be approximated by a 2D surface, by ignoring any kind of layered structure. For this reason, neuronal populations produce dipole moments predominantly aligned normal to the cortical surface. However, for subcortical grey matter, in which the cell bodies are distributed in the volume occupied by the grey matter, a 2D surface approximation isn't suitable. Hence, because 2D surface approximation would not work well, high-resolution simulation of subcortical regions is necessary. Thus, a volumetric grid, leading to the mentioned "volumetric dipole fields", is preferable. Hence, in the model, subcortical nuclei exhibit dipole orientations distributed throughout the 3D volume. Their activity is modeled as volumetric dipole fields rather than surface-normal dipoles.

[0067]

[0045] For the construction, an image processing pipeline based on the software toolboxes Freesurfer (Dale et al. 1999), FSL (Jenkinson et al. 2012) and MRtrix3 (Tournier et al. 2019) is implemented.

[0068]

[0046] T1-weighted structural MR images are processed with Freesurfer to reconstruct the cortical surface and obtain a subcortical grey matter segmentation. On the resulting triangulated surface mesh, single vertices represent on average an area of 0.8 mm2. Subcortical nuclei are represented by volumetric grids.

[0069]

[0047] White matter connectivity is determined using diffusion weighted MR images that are processed using FSL and MRtrix3.

[0070]

[0048] Figure 1 shows the triangulated surface mesh resulting from the processing of MR images. The surface mesh describing the brain cortex 1 comprises elements of varying size as it adapts to the topography of the brain gyri and sulci. The intersections of single tracts with the cortical mesh or subcortical grids are computed and appear in the inset 2 of figure 1 as thick lines in black.

[0071]

[0049] The inset 2 on the right hand side of figure 1 shows how some of the elements of the surfacic mesh intersect with a large number of tracts.

[0072]

[0050] The white matter connectivity within any pair of mesh elements is represented in the global connectivity matrix 3 shown in figure 2. Mesh elements are indexed according to their location on one of the left 4 or right 5 brain hemispheres, or on the cerebellum 6 or in the sub-cortex 7.

[0073]

[0051] Each ( / , j) element of the matrix represents the connection strength between mesh elements / and j. The connection strength is displayed using a density of points, where a higher density indicates stronger connections.

[0074]

[0052] The global connectivity matrix (3) in figure 2 shows non-zero values of ( / , j) elements of the matrix where / and j indices designate mesh elements belonging to different hemispheres, illustrating inter-hemispheric connectivity occurring through white matter tracts.

[0075]

[0053] In contrast, the matrix 8 on the right-hand side of figure 2 represents local connectivity within grey matter, which occurs in short ranges only, between neurons of the same brain area, or between neurons of neighbouring brain areas. Elements of the local connectivity matrix 8 designating pairs of mesh elements belonging to different hemispheres are essentially set to zero.

[0076]

[0054] The weights used to determine the value of the elements of the local connectivity matrix are estimated by a distance based approach. Connection strength between neighbouring mesh elements decreases exponentially with increasing distance.

[0077]

[0055] Moreover, in order to compare surface structures with volumetric structures (i.e. subcortical nuclei), each cortical vertex is assigned a volume, according to the cortical thickness. This volume is used to weigh the local connectivity matrix.

[0078]

[0056] Once the geometric structure and connectivity of the brain model have been fully determined, each mesh element can be assigned to a node of a propagation network.

[0079]

[0057] The activity of a given node can therefore be described as resulting from its own time-resolved dynamics, superimposed with local and global contributions from other nodes transmitted by the propagation network.

[0080]

[0058] The global differential equation that governs the brain simulations is the same for 2D surface and for 3D volume. The global connectivity matrix of Fig. 2, computed from the white matter fibres, reflects the volumetric grids by having some rows and columns of it belonging to the grid points and others to the surface. It is the same for the local connectivity matrix if Fig. 2, where connection strength is a function of distance between points on the surface or in the grid. Just the distance metric changes between the two, being geodesic distance for the surface and Euclidean distance for the grid.

[0081]

[0059] More precisely, the dynamics of any node / can be described using state variables x, and z,. The time evolution of these state variables is given by the so-called 2D Epileptor equations:

[0082]

[0060] Where the ygcand yicare global and local coupling scaling terms, di is the distance along white matter fibres, Vj is the vertex volume of node j, and gij is the geodesic distance along the cortical surface. H and L are the Heaviside step function and Laplace kernel, respectively, defined as: if > 0 otherwise

[0083]

[0061] The global and local coupling terms ensure that state variables time derivatives are functions of state variables values of other nodes.

[0062] Figure 3 shows a detail of a modelled brain in which the epileptogenic zone (EZ) is estimated to be in the left anterior temporal lobe 9, delimited by a continuous contour line. The other regions of the brain are in the healthy zone (HZ), delimited by a dashed contour line. Insets 10, 11 and 12 highlight the location of the EZ in the brain using different viewing angles.

[0084]

[0063] The dynamics of the nodes of the EZ are governed by the 2D Epileptor equations described above, with parameters that put the system in an excitable regime. In the absence of perturbations, the system would settle on the stable focus point in the phase space containing the values of state variables x and z.

[0085]

[0064] Within the left anterior temporal lobe 9 is defined an onset zone 13. When a slight perturbation is applied in the onset zone 13, it can cause the system to generate an excitation which can travel through the network and excite other parts of the left anterior temporal lobe 9.

[0086]

[0065] By integrating both local and global coupling at the node level, the dynamical 2D Epileptor model allows modelling travelling waves as well as re-entry dynamics in seizures. Reentry happens when a previously activated part of the system goes through its refractory period and is activated again by incoming excitation. This phenomenon has previously been observed in in-vitro studies (Keren et al. 2016).

[0087]

[0066] This capability of the model to describe long-range dynamics such as re-entry is illustrated by snapshots 14 of a seizure simulation given in figure 4. Each snapshot shows the cortical patch from the mesial side. The seizure starts from the onset zone 15 on the most anterior part of the temporal pole and a wave of activity 16, represented by crossed-hatched areas across all snapshots, passes to the posterior parts of the patch. Re-entry 17 due to delays in long range connections excites anterior parts of the cortical patch again, sustaining the seizure.

[0088]

[0067] Applications of the Virtual Brain Twin Framework also include simulating different intervention approaches.

[0089]

[0068] Beside epilepsy surgery, less invasive methods such as laser ablation or radiofrequency thermocoagulation have been developed which target the epileptogenic tissue with more precision and introduce lesions in the brain (Shamim et al. 2022).

[0090]

[0069] The effect of introducing lesions by either method is visible on figure 5 showing tract density before 18 and after 19 surgery. The area displaying tract density pre-surgery 18 is reduced when observed post-surgery 19.

[0091]

[0070] Introduction of lesions was implemented the high-resolution Virtual Brain model by placing 3 virtual lesions into the white matter of the left anterior temporal lobe, in order to reproduce the conditions shown in figure 5.

[0092]

[0071] Any tracts passing through these virtual lesions have been removed from the connectome, thus weakening connectivity in the network.

[0072] The effect of introducing these lesions is illustrated in figure 6.

[0093]

[0073] The upper half of figure 6 shows a time-space plot 20 describing a seizure event in a virtual brain where no lesions were introduced. Each row of the plot is a vertex index that is also a node of the propagation network. Columns are time intervals discretizing a time span of 1000 ms. The periodic cross-hatched regions of the time-space plot renders electrical activity in the form of oscillations recorded in each of the nodes. These electrical oscillations are shifted in time as they propagate from one node to another, but they all expand across the time window of the simulation, showing sustained electrical seizure activity in all of the nodes.

[0094]

[0074] The lower half of figure 6 is a time-space plot 21 from a simulation of the same conditions leading to seizure, but in the case where lesions have been introduced in the virtual brain. In this case, electrical activity only occurs at the beginning of the time window and terminates for times beyond 300ms.

[0095]

[0075] Another clinical practice intervention aimed at treating epileptic seizures consists in implementing closed-loop electrical stimulation devices into the estimated epileptogenic zone.

[0096]

[0076] These devices have an electrode implemented into the estimated epileptogenic zone and sense ongoing neural activity (Shamim et al. 2022). As soon as the measured signal surpasses a set threshold, a stimulus is delivered to the brain area aiming to stop the developing seizure.

[0097]

[0077] Currently used clinical closed-loop stimulation devices administer a burst of high frequency pulses. However, it has also been found that a single phase-dependent pulse could terminate seizure activity in-vivo (Osorio et al. 2009) and in after-discharges clinically (Motamedi et al. 2002).

[0098]

[0078] To test this method, virtual electrodes were implemented using the Virtual Brain Twin Framework. Figure 7 shows the normal component of the electrical field exerted by the electrodes onto the neural tissue. The hatched regions 22, 23 and 24 show where the normal component of the electric field is highest.

[0099]

[0079] Figure 8 illustrates the effect of a single phase-dependent pulse on the propagation of the seizure. The upper panel is a time-space plot 25 showing electrical activity throughout all nodes of the network. One of the electrodes measures the amplitude of the electrical signal (26) shown in the middle panel.

[0100]

[0080] From this signal, the instantaneous phase 27 is computed and shown in the lower panel. After sensing seizure activity for 500ms, three phase-dependent pulses 28 are applied to the neural field, effectively terminating ongoing seizure activity.

[0101]

[0081] Figure 9 shows a modelled volumetric grid structure of the subcortex. The rounded points 30 represent the subcortical grid of the Putamen. The white lines 31 indicate the local connectivity edges of a selected grid point, based on Euclidean distance, which build the local connectivity matrix.

[0102]

[0082] Figure 10 focuses on the temporal lobe for epilepsy simulation. The grid 32 represents the Amygdala 33, a volumetric subcortical structure. The dark "white matter” fibres 33 show how there can be connectivity between the volumetric grid and the cortical surface.

[0103]

[0083] A snap shot of neural activity simulated on that section of the brain is visible on figure 11 , where one can see how neural activity (white colour areas 34) spreads across the brain, i.e., on the cortical surface and in the subcortical grid.

[0104]

[0084] The invention is not limited to disclosed embodiments.

[0105] List of references

[0106] 1 : ... brain cortex

[0107] 2: ... inset

[0108] 3: ... global connectivity matrix

[0109] 4: ... left brain hemisphere

[0110] 5: ... right brain hemisphere

[0111] 6: ... cerebellum

[0112] 7: ... or in the sub-cortex

[0113] 8: ... matrix

[0114] 9: ... left anterior temporal lobe

[0115] 10, 11 , 12: ... locations of the epileptic zone in the brain using different viewing angles

[0116] 13: ... onset zone

[0117] 14: ... snapshots 13 of a seizure simulation

[0118] 15: ... onset zone

[0119] 16: ... wave of activity

[0120] 17: ... re-entry

[0121] 18: ... tract density before surgery

[0122] 19: ... tract density after surgery

[0123] 20: ... time-space plot

[0124] 21 : ... time-space plot

[0125] 22, 23, 24: ... hatched regions

[0126] 25: ... time-space plot

[0127] 28: ... phase-dependent pulses

[0128] 30: ... subcortical grid points of the Putamen

[0129] 31 : ... local connectivity edges of a grid point

[0130] 32: ... grid

[0131] 33: ... white matter fibres

[0132] 34: ... areas of neural activity

Claims

1. Claims1. Computer-implemented method for modelling a primate brain for simulating the propagation of a neural activity represented by a time-resolved dynamics, characterized in that the method uses a propagation network made of nodes connected using long-range and short- range weighted connections, segmenting the grey matter of the brain, network in which an activity of each node results from its own time-resolved dynamics, superimposed with local and global contributions from other nodes, the method comprises the steps of:- reconstructing a cortical surface of the primate brain and obtaining a segmentation of a subcortical grey matter of said primate brain, resulting in a meshed surface discretised into mesh elements,- modelling each mesh element as a node of the propagation network,- estimating single tracts ensuring white matter connectivity, using previously recorded brain images,- determining global contribution to each node from the other nodes using intersections of estimated single tracts with the meshed surface, allowing to define long range weights rendering long range tract-based connectivity between said node and other nodes,- determining local contribution to each node from the other nodes by defining short range weights rendering short range connectivity between said node and other nodes and- assigning differentiated conduction velocities to long range and short range connectivities.

2. Method according to claim 1 , wherein the single tracts are constructed by an image processing pipeline based on specialised software toolboxes adapted to:- construct the meshed surface as a triangulated surface, wherein single mesh elements represent areas of 0.5 mm2 to 1.2 mm2, preferably 0.8 mm2, and wherein subcortical nuclei are represented by volumetric grids,- process diffusion weighted MR images to estimate white matter connectivity.

3. Method according to any one of claims 1 and 2, wherein the long range weights rendering long range tract-based connectivity are inputted in a global connectivity matrix.

4. Method according to any one of claims 1 , 2 and 3, wherein a short range weight depends on the cortical thickness at the node.

5. Method according to any one of claims 1 , 2, 3 and 4, wherein a short range weight decreases exponentially with increasing distance between nodes.

6. Method according to any one of claims 1 , 2, 3, 4 and 5, wherein the differentiated conduction velocities consist in one average long range conduction velocity for long range connectivities and one average short range conduction velocity for short range connectivities.

7. Method according to any one of claims 1 , 2, 3, 4, 5 and 6, wherein the method further comprises the step of implementing an activity threshold for at least one connectivity selected in the group consisting of the short range and the long range connectivities and neutralizing the propagation according to the selected connectivity when the node activity is under the threshold.

8. Method for modelling an epileptic brain, characterized in that an epileptogenic model is introduced as the time-resolved dynamics in the modelling method according to any one of claims 1 , 2, 3, 4, 5, 6 and 7.

9. Method for testing a countermeasure on an epileptic brain, characterized in that the method includes a step of testing the effect of setting at least one short range or long range weight to zero when implementing the method according to claim 8.

10. Computer program product for modelling of a primate brain, characterized in that it comprises instructions which, when the program is executed by a computer, cause the computer to perform the method according to any one of claims 1 , 2, 3, 4, 5, 6, 7, 8, and 9.11 . Data processing system, characterized in that it comprises a processor configured to perform the method according to any one of claims 1 , 2, 3, 4, 5, 6, 7, 8, 9 and 10.

12. Computer-readable storage medium, characterized in that it comprises instructions which, when executed by a computer, cause the computer to perform the method according to any one of claims 1 , 2, 3, 4, 5, 6, 7, 8, 9, 10 and 11 .

Images