Systems and methods for constructing epileptogenic networks and calculating node criticality

Abstract

Systems and methods for constructing an epileptogenic network using non-invasive sensor data or invasive sensor data of a subject are provided. The non-invasive sensor data or invasive sensor data include epileptic sensor activity. Epileptic epoch are generated by cropping the non-invasive sensor data or the invasive sensor data based on epileptic event timing and one or more nodes are defined. For one defined node, a epileptogenic network is constructed that includes one node and zero edges. For two or more defined nodes, node activity can be determined by transforming either source spatialtemporal cluster(s) (for non-invasive sensor data) or the epileptic epoch (for invasives sensor data). A source-target influence value can be estimated for each node pair, and an epileptogenic network is constructed including n nodes and e edges based on the node pairs in the two or more nodes and the estimated source-target influence value for each node pair.

Description

SYSTEMS AND METHODS FOR CONSTRUCTING EPILEPTOGENIC NETWORKS AND CALCULATING NODE CRITICALITYCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application is based on, claims priority to, and incorporates herein by reference in its entirety U.S. Serial No. 63 / 588,709 filed October 7, 2023, and entitled “System For And Method Of Identifying Critical Nodes In The Epileptogenic Network."STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0002] N / AFIELD

[0003] The present disclosure relates generally to neuroimaging, and, in particular, to neuroimaging for constructing epileptogenic networks and calculating node criticality using non- invasive sensor data and / or invasive sensor data.BACKGROUND

[0004] Epilepsy is a neurological disorder characterized by recurrent, unprovoked seizures, which are sudden bursts of abnormal electrical activity in the brain. These seizures can vary widely in their intensity, duration, and manifestation, ranging from brief lapses of attention or muscle twitches to more severe convulsions and loss of consciousness. A variety of non-invasive and invasive imaging modalities are used to locate where seizures originate in a patient’s brain, each location known as seizure onset zone (SOZ). SOZs are surgical targets for procedures such as resection, ablation, and device implantation, and while targeting SOZs for some patients can result in a reduction or elimination of seizures, many patients experience little or no improvement in surgical outcomes. Targeting SOZs alone is often insufficient due to the dynamic nature of the epileptic seizures, where bursts of abnormal electrical activity (which originate in SOZs) propagate through brain regions and certain regions significantly influence the neurodynamic more than others. Understanding the dynamics of epileptogenic networksallows for the identification of important surgical targets, and thereby improves the chances for seizure freedom. Thus, there is a need for constructing epileptogenic networks to identify surgical targets beyond SOZs to achieve better surgical outcomes.SUMMARY

[0005] In accordance with an embodiment, a computer implemented method for constructing an epileptogenic network using non-invasive sensor data where the epileptogenic network comprises n nodes and e edges includes receiving non-invasive sensor data of a subject comprising epileptic sensor activity, receiving epileptic event timing, cropping the non-invasive sensor data of the subject based on the epileptic event timing to generate an epileptic epoch, estimating source activity of the epileptic epoch to generate an estimated epileptic epoch source activity, clustering the estimated epileptic epoch source activity to generate at least one source spatiotemporal cluster, defining one or more nodes by defining a point in each of the at least one source spatiotemporal cluster. When the one or more nodes comprises one node, the method further includes constructing an epileptogenic network comprising one node and zero edges. When the one or more nodes comprising two or more nodes, the method further includes for each node transforming each at least one source spatialtemporal cluster to a node activity, for each node pair in the two or more nodes, estimating a source-target influence value, and constructing an epileptogenic network comprising two or more nodes and e edges based on the node pairs in the two or more nodes and the estimated source-target influence value for each node pair, where e is a number between 0 to n(n-l).

[0006] In accordance with another embodiment, a computer implemented method for constructing an epileptogenic network using invasive sensor data where the epileptogenic network comprises n nodes and e edges includes receiving invasive sensor data of a subject comprising epileptic sensor activity, receiving epileptic event timing, cropping the invasive sensor data of the subject based on the epileptic event timing to generate an epileptic epoch, defining one or more nodes. When the one or more nodes comprises one node, the method further includes constructing a epileptogenic network comprising one node and zero edges. When the one or more nodes comprising two or more nodes, the method further includes for each node transforming the epileptic epoch to a node activity, for each node pair in the two or more nodes, estimating a source-target influence value, and constructing an epileptogenicnetwork comprising two or more nodes and e edges based on the node pairs in the two or more nodes and the estimated source-target influence value for each node pair, wherein e is a number between 0 to n(n-l).

[0007] In accordance with another embodiment, a computer implemented method for constructing a multi-event epileptogenic network (MEEN) includes receiving sensor data for a subject for each of a plurality of epileptic events, the sensor data for each epileptic event comprising epileptic sensor activity, integrating the sensor data for each epileptic event to form a set of integrated multi-event sensor data and defining one or more nodes. When the one or more nodes comprises one node, the method further includes constructing a multi-event epileptogenic network comprising one node and zero edges. When the one or more nodes comprising two or more nodes, the method further includes estimating a source-target influence value for each node pair in the two or more nodes, and constructing a multi-event epileptogenic network comprising two or more nodes and e edges based on the node pairs in the two or more nodes and the estimated source-target influence value for each node pair, wherein e is a number between 0 to n(n-l).BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The disclosure is best understood from the following detailed description when read in connection with the drawings disclosed herein, with similar elements having the same reference numbers. When a plurality of similar elements is present, a single reference number may be assigned to the plurality of similar elements with a small letter designation referring to at least one specific similar element. When referring to the similar elements collectively or to a nonspecific similar element, the small letter designation may be dropped. The various features of the drawings may not be drawn to scale and may be arbitrarily expanded or reduced for clarity. Included in the drawings are the following figures:

[0009] FIG. 1 is a block diagram of a system for constructing an epileptogenic network and calculating node criticality in accordance with an embodiment;

[0010] FIGs. 2A and 2B illustrate a method for constructing an epileptogenic network using non- invasive sensor data in accordance with an embodiment;

[0011] FIGs. 3A and 3B illustrate a method for constructing an epileptogenic network using invasive sensor data in accordance with an embodiment;

[0012] FIG. 4 illustrates a method for constructing a multi-event epileptogenic network (MEEN) from single-event epileptogenic networks (SEENs) in accordance with an embodiment;

[0013] FIG. 5 illustrates an example single-event epileptogenic network (SEEN) in accordance with an embodiment;

[0014] FIG. 6A illustrates two example single-event epileptogenic networks in accordance with an embodiment;

[0015] FIG. 6B illustrates an example multi-event epileptogenic network (MEEN) constructed from the single-event epileptogenic networks of FIG. 5 A in accordance with an embodiment; and

[0016] FIG. 7 is a block diagram of an example computer system in accordance with an embodiment.DETAILED DESCRIPTION

[0017] Unless otherwise defined, all terms used herein have the same meaning as commonly understood by a person having ordinary skill in the art to which the disclosure pertains. All patents, patent applications, publications, and other references mentioned herein and / or listed in the Application Data Sheet are hereby incorporated by reference in their entirety. In case of conflict, the specification will control. When a range of values is provided, the range includes the end values. The terms "include," "includes," "including," "for example," "example of," "examples of," and "such as" and other like terms are not intended to be limiting. The materials, methods, components, features, embodiments, examples, and drawings disclosed herein are not intended to be limiting.

[0018] To facilitate understanding of the present disclosure, a number of terms are defined herein.

[0019] “Anatomical data” means data that represents the anatomy of at least part of a patient’s or subject’s brain or a template brain. An example of a modality that generates anatomical data includes, but is not limited to, magnetic resonance imaging (MRI).

[0020] “Edge” means a directional connection between 2 nodes, a first node and a second node, in which the first node affects the activity in the second node after a specific period of time. For each node pair there are up to 2 edges, meaning for each node pair there are 0, 1, or 2 edges. For example, for a node pair consisting of a first node and a second node, there may be (i) no edges or (ii) an edge from the first node to the second node and / or (iii) an edge from the second node to the first node.

[0021] “Epileptic event” means abnormal electrical activity in the brain that is characteristic of epilepsy. Epileptic events can be categorized into 2 main types: interictal and ictal events.

[0022] “Epileptic epoch” means cropped sensor data that includes all or part of the duration of an epileptic event.

[0023] “Epileptic event timing” means the time of the onset of an epileptic event.

[0024] “Estimated epileptic epoch source activity” means an estimated activity on the cortical and / or sub-cortical surface or volume that causes the sensor activity in an epileptic epoch (which is based on the epileptic timing of a single epileptic event).

[0025] “Estimated non-epileptic epoch source activity” means an estimated activity on the cortical and / or sub-cortical surface or volume that causes the sensor activity in at least 1 nonepileptic epoch (which is not based on the epileptic timing of a single epileptic event).

[0026] “Ictal event” means the onset of abnormal electrical activity in the brain that leads to a clinical manifestation of a seizure.

[0027] “ Influence value” means a value that represents the degree to which the activity in a first node (also referred to as the “source node”) affects the activity in a second node (also referred to as the “target node”) after a certain period of time.

[0028] “Influence likelihood” means the probabilistic measure that quantifies the significance of the directional effect that a first node (also referred to as the “source node”) exerts over a second node (also referred to as the “target node”).

[0029] ‘ ‘Interictal event” means the onset of abnormal electrical epileptic activity in the brain that does not lead to a clinical manifestation of a seizure.

[0030] ‘ ‘Invasive sensor data” means data from at least 1 sensor located on and / or in a subject patient’s brain that probes neuroactivity over time. Examples of modalities that probe invasive sensor data include, but are not limited to, depth electrodes (including but not limited to stereoelectroencephalography (sEEG)), subdural electrodes (including but not limited to electrocorticography (ECoG) grids and strips), epidural electrodes, foramen ovale electrodes, and microelectrodes.

[0031] “Multi-event epileptogenic network” or “MEEN” means an epileptogenic network comprising 2 or more single-event epileptogenic networks (SEENs).

[0032] ‘ ‘Node” means (i) for invasive sensor data, a sensor that was active during an epileptic event or (ii) for non-invasive sensor data, a point in a patch of cortical or sub-cortical surface or volume that was active during an epileptic event.

[0033] “Node activity” means (i) for invasive sensor data, the data recorded during an epileptic event by the sensor that represents a node or (ii) for non-invasive sensor data, the activity of a patch of cortical or sub-cortical surface or volume during an epileptic event.

[0034] “Node criticality” means a measure of the degree that a node in an epileptogenic network affects the network.

[0035] “Non-epileptic epoch” means cropped sensor data that does not include all or part of an epileptic event.

[0036] ‘ ‘Non-epileptic source activity” means brain activity that is not associated with epileptic seizures or abnormal neural discharges but is detected during neuroimaging studies.

[0037] ‘ ‘Non-invasive sensor data” means data from at least one sensor located outside of a subject patient’s brain that probes neuroactivity over time. Examples of modalities that probe non-invasive sensor data include, but are not limited to, magnetoencephalography (MEG) and non-invasive electroencephalography (EEG).

[0038] “Single-event epileptogenic network” or “SEEN” means an epileptogenic network comprising (i) 1 node that was active during a single epileptic event and 0 edges (no edges) or (ii) 2 or more nodes, represented as n nodes, that were active during a single epileptic event and connected by 0 to n(n-l) edges. For example, a SEEN comprising 3 nodes could be connected by 0, 1, 2, 3, 4, 5, or 6 edges.

[0039] “ Source activity” means the value of at least one vertex on the cortical surface for a particular time point or the value of at least one voxel in the three-dimensional space for a particular time point. Source activity is calculated by executing a surface-based brain activity modeling algorithm (for vertices) or a volume-based brain activity modeling algorithm (for voxels) as part of the source localization process for non-invasive sensors.

[0040] “ Source spatialtemporal cluster” means a group of spatially contiguous brain regions, vertices, or voxels that show statistically significant activity or connectivity related to a specific condition, task, or event.

[0041] “ Source-target influence value” means, as between a source node and a target node, the degree to which the source node’s activity affects the target node’s activity after a specific lag in time.

[0042] “Source-target influence likelihood value” means, as between a source node and a target node, the likelihood that the source node has a direct causal influence on the target node after a specific lag in time.

[0043] “Surface-based brain activity modeling algorithm” means an algorithm that models brain activity on the cortical surface, often represented as a mesh derived from anatomical MRI scans (like the pial or white matter surfaces). The activity is constrained to this two-dimensional surface, which is essentially the folded sheet of the cortex where most brain activity occurs. Examples of surface-based brain activity modeling algorithms include, but are not limited to, dynamic Statistical Parametric Mapping (dSPM), Minimum Norm Estimate (MNE), and Standardized Low Resolution Electromagnetic Tomography (sLORETA), Exact Low Resolution Electromagnetic Tomography (eLORETA), and Sequential Weighted Averaging Regularization Method (SWARM).

[0044] ‘ ‘Volume-based brain activity modeling algorithm” mean an algorithm that models brain activity in three-dimensional space, not limited to the surface of the cortex but distributed across the entire brain volume. This includes both cortical and subcortical regions. Examples of volume-based brain activity modeling algorithms include, but are not limited to, Linearly Constrained Minimum Variance (LCMV), Synthetic Aperture Magnetometry (SAM), and Single or Multiple Dipole Fit.

[0045] The current state of the art in identifying surgical targets for epilepsy treatment is limited to seizure onset zones (SOZs). The present disclosure is directed to systems and methods for constructing epileptogenic networks and calculating node criticality to identify surgical targets (nodes) beyond SOZs. In some embodiments, the disclosed systems and methods can include constructing both single-event epileptogenic networks (SEENs) and multi-event epileptogenic networks (MEENs) as well as for calculating node criticality. Constructing epileptogenic networks and calculating node criticality can be useful for a variety of tasks including, but not limited to, surgical planning, surgical simulation, and surgical execution.

[0046] FIG. l is a block diagram of a system for constructing an epileptogenic network and calculating node criticality in accordance with an embodiment. The system 100 can include anepileptogenic network construction module 102, a node criticality module 104, one or more inputs (e.g., non-invasive sensor data 106, invasive sensor data 108, anatomical data 110, and / or epileptic event timing 112), a pre-processing module 114, an output 116 of the epileptogenic network construction module 102 (e.g., an epileptogenic network 118 and node criticality 120), data storage (or memory) 122, and a display 124. The system 200 may be configured to construct or generate a epileptogenic network using the epileptogenic network construction module 102, and to determine (or calculate) node criticality for the constructed epileptogenic network using, for example, the node criticality module 104 of the epileptogenic network construction module 102. In some embodiments, the constructed epileptogenic network can be a single-event epileptogenic network (SEEN). In some embodiments, the constructed epileptogenic network can be a multi-event epileptogenic network (MEEN). In some embodiments, the epileptogenic network construction module 102 can be configured to construct an epileptogenic network (e g., a SEEN or MEEN) based on non-invasive sensor data 106, epileptic event timing 112, and optionally, anatomical data 110. In some embodiments, the epileptogenic network construction module 102 can be configured to construct an epileptogenic network (e.g., a SEEN or MEEN) based on invasive sensor data 108, epileptic event timing 112, and optionally, anatomical data 110.

[0047] In some embodiments, as mentioned, an input to the epileptogenic network construction module 102 can include, for example, non-invasive sensor data 106. Non-invasive sensor data refers to information collected from sensors that do not require insertion into the body or any form of invasive procedure to gather data. These sensors operate externally, typically placed on the skin or near the surface of the body, and they measure physiological, environmental, or mechanical signals without penetrating bodily tissues or causing significant discomfort to the subject. Examples of non-invasive sensor data include, but are not limited to, electroencephalography (EEG), magnetoencephalography (MEG), and near-infrared spectroscopy (NIRS). In some embodiments, the non-invasive sensor data includes epileptic sensor activity. In some embodiment, the non-invasive sensor data can also optionally include non-epileptic sensor data.

[0048] Non-invasive EEG, also known as scalp EEG, is a method that records the electrical activity of the brain using electrodes placed on the scalp. Electrodes detect the small electrical impulses produced by neuronal activity. These signals are amplified and recorded for analysis.EEG is widely used in clinical settings to diagnose and monitor neurological conditions such as epilepsy, sleep disorders, and brain injuries. It is also used in research to study brain function, cognitive processes, and neurofeedback. High-density EEG (HD EEG) is an advanced form of EEG that involves the use of a large number of closely spaced electrodes placed on the scalp to record electrical activity from the brain. Typically, HD EEG systems use 64, 128 or even more sensors (electrodes) compared to the standard 19-32 electrodes used in EEG. HD EEG allows for more detailed mapping of brain activity including source localization.

[0049] MEG measures the magnetic fields produced by neuronal electrical activity in the brain. Superconducting quantum interference devices (SQUIDs) or optically pumped magnetometers detect the magnetic fields generated by neural currents. These signals are analyzed to map brain activity. MEG is used in research and clinical settings to study brain function, particularly in areas such as language processing, sensory processing, and epilepsy and are also used for pre- surgical mapping.

[0050] NIRS is a technique that uses near-infrared light to measure the concentration of oxygenated and deoxygenated hemoglobin in tissues, providing information about blood oxygenation and hemodynamics. Near-infrared light is emitted into the tissue, and detectors measure the light that is either absorbed or scattered. The data is used to calculate changes in blood oxygenation and blood flow. NIRS is used in brain research to monitor cerebral oxygenation and hemodynamics, in sports science to assess muscle oxygenation during exercise, and in clinical settings to evaluate tissue viability and monitor brain function in neonates and during surgery.

[0051] In some embodiments, any type of non-invasive sensor data may be used in system 100 with a preference for MEG or HD EEG. The non-invasive sensor data 106 may be acquired using the appropriate systems for EEG, MEG, and NIRS, as known in the art. In some embodiments, the non-invasive sensor data 106 may be retrieved or received from data storage (or memory, or database) such as, for example, data storage 122 of system 100, data storage of the system used to acquire the non-invasive sensor data, or data storage of other computer systems (e.g., storage device 616 of the example computer system 600 shown in FIG. 6).

[0052] In some embodiments, an input to the epileptogenic network construction module 102 can include, for example, invasive sensor data 108. Invasive sensor data refers to information collected from sensors that require insertion into the body or direct contact with internal tissuesor organs. These sensors often provide more precise and localized measurements compared to non-invasive sensors but involve surgical or minimally invasive procedures to place them within the body. Invasive sensors are typically used when external measurements are insufficient to obtain the required data, especially in clinical and research settings. Examples of invasive sensor data include, but are not limited to, stereo-electroencephalography (sEEG), epidural electrocorticography (ECoG), and microelectrode arrays.

[0053] sEEG involves the insertion of depth sensors (electrodes) into specific regions of the brain to record electrical activity from within the brain tissue. Sensors are surgically implanted through small holes drilled into the skull. They can record electrical signals from deep brain structures, providing detailed information about neuronal activity and are primarily used in the diagnosis and treatment of epilepsy, particularly for identifying the precise location of seizure foci before surgical intervention. sEEG is used in research to study brain function and connectivity.

[0054] ECoG involves placing electrodes on the surface of the brain, beneath the dura mater (the outermost layer of the meninges). Sensors are surgically implanted during a craniotomy (a surgical procedure that opens the skull) to record electrical activity directly from the cerebral cortex. ECoG is typically used to localize seizure foci in epilepsy patients, map functional areas of the brain before surgical resections, and to research brain-machine interfaces.

[0055] Microelectrode arrays consist of multiple tiny sensors that can record from or stimulate individual neurons or small groups of neurons. These arrays are surgically implanted into specific brain regions to record high-resolution neural activity or deliver electrical stimulation. Microelectrode arrays are used in neurophysiological research to study neural coding and connectivity, brain-machine interfaces for controlling prosthetic devices, and neuroprosthetics to restore lost sensory or motor functions.

[0056] In some embodiments, any type of invasive sensor data may be used with a preference for sEEG or ECoG. The invasive sensor data 108 may be acquired using the appropriate systems for sEEG, ECoG, and microelectrode arrays, as known in the art. In some embodiments, the invasive sensor data 108 may be retrieved or received from data storage (or memory, or database) such as, for example, data storage 122 of system 100, data storage of the system used to acquire the invasive sensor data, or data storage of other computer systems (e.g., storage device 616 of the example computer system 600 shown in FIG. 6).

[0057] In some embodiments, an input to the epileptogenic network construction module 102 can optionally (as indicated by the dashed lines in FIG. 1) include, for example, anatomical data 110. Anatomical data allows for the registration of non-invasive and invasive sensor data to the anatomical space of at least part of the subject patient’s brain. For non-invasive sensor data, anatomical data can be used to estimate the exact location of neuronal activity based on the data acquired by the sensors. Anatomical data can be used to identify the exact location of the invasive sensors and the brain region(s) responsible for generating neuronal activity. In some embodiments, the anatomical data can be determined using magnetic resonance imaging (MRI). In some embodiments, the MR images used to determine anatomical data can be acquired using an MRI system using various sequences (instructions and protocols), and in some embodiments, for example, the T1 Magnetization Prepared Rapid Gradient Echo (MPRAGE) sequence can be used for neuroimaging. However, a person having ordinary skill in the art to which this invention pertains will appreciate that other modalities may provide anatomical data sufficient for constructing an epileptogenic network. In some embodiments, the optional anatomical data 110 may be retrieved or received from data storage (or memory, or database) such as, for example, data storage 122 of system 100, data storage of the system (e.g., an MRI system) used to acquire the anatomical data, or data storage of other computer systems (e.g., storage device 616 of the example computer system 600 shown in FIG. 6).

[0058] In some embodiments, an input to the epileptogenic network construction module 102 can also include, for example, epileptic event timing 112. Epileptic event timing exists for both ictal events and interictal events and is commonly determined through visual inspection of the captured time series data (from non-invasive and invasive sensors). This process can also be automated through automatic detection algorithms. For ictal events, the epileptic timing is the seizure onset timing which is the precise moment in time when a seizure begins. For interictal events, the epileptic timing is the onset timing of the electrical discharges known as interictal epileptiform discharges (lEDs). In some embodiments, the epileptic timing data 112 may be retrieved or received from data storage (or memory, database) such as, for example, data storage 122 of system 100, data storage of the system used to determine the epileptic timing data, or data storage of other computer systems (e.g., storage device 616 of the example computer system 600 shown in FIG. 6).

[0059] In some embodiments, one or more of the inputs, for example, the non-invasive sensor data 106, the invasive sensor data 108, and the anatomical data 110 can be pre-processed before being provided to the epileptogenic network construction module 102 using the optional preprocessing module 114.

[0060] In some embodiments, the pre-processing module 114 can be configured to perform preprocessing on the non-invasive sensor data 106. In some embodiments, preprocessing of non- invasive sensor data can include, for example, (i) filtering to remove unwanted frequency components and improve signal quality; (ii) removing and / or correcting artifacts to detect and remove and / or correct non-brain signals, such as eye blinks, muscle activity, and heartbeat; (iii) re-referencing to adjust the reference electrode to a better configuration for improved signal quality and interpretation; (iv) down-sampling to decrease data size and computational load while retaining the information necessary for analysis; (v) detecting and interpolating bad channels to identify and correct noisy or non-functioning sensors; and (vi) correcting baseline to adjust the signal to correct for baseline drifts that occur over time. An example of a tool that automates all or part of preprocessing non-invasive sensor data 106 is the MaxFilter (Megin Oy, Espoo, Finland) used for pre-processing MEG data.

[0061] In some embodiments, the pre-processing moule 114 can be configured to perform preprocessing on the invasive sensor data 108. In some embodiments, preprocessing of invasive sensor data can be similar to the preprocessing of non-invasive sensor data except invasive sensors typically capture higher-frequency signals (FIFOs), whereas non-invasive sensors tend to capture more low- to mid-frequency ranges; non-invasive sensor data is more prone to external artifacts while invasive sensor data is subject to more localized artifacts like cardiac signals; and non-invasive sensors have more flexibility in reference selection, while invasive sensors usually employ bipolar referencing between adjacent electrodes.

[0062] In some embodiments, the pre-processing module 114 can be configured to perform preprocessing on the optional anatomical data 110. In some embodiments, the optional preprocessing of anatomical data can reconstruct cortical surfaces and segment sub-cortical regions. The unprocessed data files from an anatomical imaging device (e.g., an MRI system) are usually in raw format such as Digital Imaging and Communications in Medicine (DICOM) format or packaged format such as Neuroimaging Informatics Technology Initiative (NIfTI) format. Common tools for preprocessing anatomical data include FreeSurfer (Laboratories forComputational Neuroimaging, Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Charlestown, Massachusetts, USA), NeuroQuant (Cortechs Labs, Inc., San Diego, California, USA), iPlan (BrainLab AG, Munich, Germany), and BESA MRI (BESA GmbH, Grafelfing, Germany). Reconstructed cortical surfaces can be outputted in the form of three-dimensional meshes and segmented sub-cortical regions can be outputted in the form of volumetric segmentation maps. In addition, for source localization for non-invasive sensors, a head modeling method may be required to solve the forward problem (biophysical model) of source localization that involves determining how neuronal activity in the brain generates measurable electrical or magnetic fields outside of the head. Common head modeling methods for solving the forward problem include Boundary Element Method (BEM), Finite Element Method (FEM), and spherical modeling.

[0063] The epileptogenic network construction module 102 can be configured to receive one or more of the inputs 106, 108, 110, 112 (directly, or optionally from pre-processing module 114) and to generate outputs 118 including, for example, a constructed (or generated) epileptogenic network 118 (e.g., a SEEN or a MEEN) and, optionally, node criticality 120 calculated (or determined) using a node criticality module 104 based at least on the constructed epileptogenic network. In some embodiments, the epileptogenic network construction module 102 can receive non-invasive sensor data 106 and epileptic event timing 112 (and optionally, anatomical data 110) and construct an epileptogenic network (e.g., a single-event epileptogenic network or a multi-event epileptogenic network) and, optionally, node criticality, using methods as described below with respect to FIGs. 2A and 2B. In some embodiments, the epileptogenic network construction module 102 can receive invasive sensor data 108 and epileptic event timing 112 (and optionally, anatomical data 110) and construct an epileptogenic network (e.g., a singleevent epileptogenic network or a multi-event epileptogenic network) and, optionally, node criticality, using methods as described below with respect to FIGs. 3 A and 3B. The outputs 118 of the epileptogenic network construction module 102, for example, an epileptogenic network 118 and node criticality 120 can be stored in, for example, data storage 122 of the system 100. In some embodiments, the epileptogenic network 118 and the node criticality 120 (e.g., the critical nodes) can be displayed to a user or operator on a display 124.

[0064] In some embodiments, the epileptogenic network construction module 102 (including the node criticality module 104) and the pre-processing module 114 may be implemented on one ormore processors (or processor devices) of a computer system such as, for example, any general purpose computer system or device, such as a personal computer, workstation, cellular telephone, smartphone, laptop, tablet, or the like. As such, the computer system may include any suitable hardware and components designed or capable of carrying out a variety of processing and control tasks, including, for example, steps for implementing the pre-processing module 114, implementing the epileptogenic network construction module 102 (including the node criticality module 104), receiving one or more inputs including, for example, non-invasive sensor data 106, invasive sensor data 108, anatomical data 110, and epileptic event timing 112. For example, the computer system may include a programmable processor or combination of programmable processors, such as central processing units (CPUs), graphic processing units (GPUs), and the like. In some implementations, the one or more processors of the computer system may be configured to execute instructions stored in a non-transitory computer readable-media. In this regard, the computer system may be any device or system designed to integrate a variety of software, hardware, capabilities and functionalities. Alternatively, and by way of particular configurations and programming, the computer system may be a special purpose system or device. For example, such special-purpose system or device may include one or more dedicated processing units or modules that may be configured (e.g., hardwired, or pre-programmed) to carry pout steps, in accordance with aspects of the present disclosure.

[0065] FIGs. 2A and 2B illustrate a method for constructing an epileptogenic network using non- invasive sensor data in accordance with an embodiment. The process illustrated in FIGs. 2A and 2B is described below as being carried out by the system 100 as illustrated in FIG. 1. However, in some examples, the process of FIGs. 2A and 2B may be implemented by another system. Although the blocks of the process are illustrated in a particular order, in some embodiments, one or more blocks may be executed in a different order than illustrated in FIGs. 2A and 2B or may be bypassed.

[0066] In FIG. 2A, at block 202, non-invasive sensor data 106 for a subject (e.g., of the head and brain of the subject) can be received. As mentioned, the non-invasive sensor data can be, for example, electroencephalography (EEG), magnetoencephalography (MEG), and near-infrared spectroscopy (NIRS). In some embodiments, the non-invasive sensor data includes epileptic sensor activity. In some embodiment, the non-invasive sensor data can also optionally include non-epileptic sensor data. In some embodiments, the non-invasive sensor data can be received(or retrieved) from data storage (or memory) such as, for example, data storage 122 of system 100, data storage of the system used to acquire the non-invasive sensor data, or data storage of other computer systems (e.g., storage device 616 of the example computer system 600 shown in FIG. 6). At optional block 204 (illustrated with dashed lines in FIG. 2A), in some embodiments, anatomical data 110 for the subject may be received. As mentioned, the anatomical data can be determined using, for example, magnetic resonance imaging (MRI) or other modalities. In some embodiments, the optional anatomical data 110 may be retrieved or received from data storage (or memory, or database) such as, for example, data storage 122 of system 100, data storage of the system (e.g., an MRI system) used to acquire the anatomical data, or data storage of other computer systems (e.g., storage device 616 of the example computer system 600 shown in FIG. 6). At block 206, in some embodiments, epileptic event timing 112 may be received. As mentioned, epileptic event timing exists for both ictal events and interictal events and is commonly determined through visual inspection of the captured time series data (from non- invasive and invasive sensors). In some embodiments, this process can also be automated through automatic detection algorithms. For ictal events, the epileptic timing is the seizure onset timing which is the precise moment in time when a seizure begins. For interictal events, the epileptic timing is the onset timing of the electrical discharges known as interictal epileptiform discharges (lEDs). In some embodiments, the epileptic timing 112 may be retrieved or received from data storage (or memory, database) such as, for example, data storage 122 of system 100, data storage of the system used to determine the epileptic timing data, or data storage of other computer systems (e.g., storage device 616 of the example computer system 600 shown in FIG. 6).

[0067] At block 208, optional pre-processing (illustrated with dashed lines in FIG. 2A) may be performed on, for example, the non-invasive sensor data and / or the anatomical data (e.g., using a pre-processing module 114). As mentioned, in some embodiments, preprocessing of non- invasive sensor data can include, for example, (i) filtering to remove unwanted frequency components and improve signal quality; (ii) removing and / or correcting artifacts to detect and remove and / or correct non-brain signals, such as eye blinks, muscle activity, and heartbeat; (iii) re-referencing to adjust the reference electrode to a better configuration for improved signal quality and interpretation; (iv) down-sampling to decrease data size and computational load while retaining the information necessary for analysis; (v) detecting and interpolating badchannels to identify and correct noisy or non-functioning sensors; and (vi) correcting baseline to adjust the signal to correct for baseline drifts that occur over time. In some embodiments, as mentioned, the preprocessing of anatomical data can include, for example, reconstructing cortical surfaces and segmenting sub-cortical regions.

[0068] In some embodiments, the non-invasive sensor data 106 (unprocessed or pre-processed) and, in some embodiments, the optional anatomical data 110 (unprocessed or preprocessed) and epileptic event timing 112 can be provided as input to, for example, an epileptogenic network construction module 102. At block 210, the non-invasive sensor data 106 can be cropped to generate epileptic epoch. In addition, in some embodiments, the non-invasive sensor data may also be cropped to generate non-epileptic epoch. Cropping refers to the process of selecting and retaining a specific segment of the data based on defined time intervals. It can involve, for example, specifying a time window (e.g., from 2 seconds before an event to 4 seconds after) and discarding data outside of that window. Cropping is commonly used to extract epochs around an event of interest (e.g., seizure onset). In some embodiments, cropping can be done manually. In some embodiments, cropping can be done by an automated procedure (e.g., using the epileptogenic network construction module 102). Epileptic epoch is the segment of data that results after the cropping non-invasive sensor data, and epileptic epoch comprises at least one epileptic event. The optional non-epileptic epoch is the segment of data that results after cropping non-invasive sensor data, and non-epileptic epoch does not comprise at least one epileptic event (i.e., there are no epileptic events).

[0069] At block 212, source activity of epileptic epoch can be estimated to generate estimated epileptic epoch source activity. In some embodiments, at optional block 214 (illustrated with dashed lines in FIG. 2A), source activity of non-epileptic epoch can be estimated to generate estimated non-epileptic epoch source activity. In some embodiments, the estimation of the source activity of epileptic epoch (block 212) and the estimation of source activity of non-epileptic epoch (block 214) can be performed using the epileptogenic network construction module 102. In some embodiments, estimating the epileptic epoch source activity or non-epileptic source activity can include, for example: (i) non-invasive sensor data-anatomical data co-regi strati on, (ii) forward modeling, (iii) noise covariance estimation, and (iv) source activity reconstruction.

[0070] Non-invasive sensor data-anatomical data co-regi strati on can include, for example, aligning the sensor positions with an MRI generated from a patient’s (or subject’s) brain ortemplate brain. The patient’s (or subject’s) head shape can be digitized and aligned to the patient’s brain or the template brain. Forward modeling can include, for example, computing a lead field matrix, which describes how each potential source impacts the non-invasive sensors. This can involve solving Maxwell's equations, often using the Finite Element Method (FEM), spherical modeling, and preferably Boundary Element Method (BEM). Noise covariance estimation can include estimating a noise covariance matrix by concatenating non-epileptic epoch sensor data or, in some embodiments, using any subset of the sensor data. Alternatively, and preferably, estimation of the noise covariance matrix can be calculated using measurements of an empty room. Source activity reconstruction can include using a surface-based brain activity modeling algorithm or a volume-based brain activity modeling algorithm to reconstruct source activity during an epileptic epoch.

[0071] At optional block 216 (illustrated with dashed lines in FIG. 2A), in some embodiments, the estimated epileptic epoch source activity (block 212) can be normalized using, for example, the non-epileptic source activity. In some embodiments, normalizing the estimated epileptic epoch source activity with non-epileptic source activity can include, for example, for each time point in the epileptic epoch, subtracting the average (over time) of the non-epileptic source activity from the epileptic epoch source activity to calculate baseline-corrected values. In addition, in some embodiments, the baseline-corrected values can be optionally divided by the standard deviation of the non-epileptic source activity to calculate z-scores. The process of normalization can be on a vertex-by-vertex basis or a voxel-by-voxel basis. Normalization controls for noise and drift over time, can make it easier to identify significant changes during the epileptic event, and can allow for the removal of bias resulting from changes in signal strength across regions or time. In some embodiments, normalization of the estimated epileptic epoch source activity can be performed using the epileptogenic network construction module 102.

[0072] At block 218, the estimated epileptic epoch source activity can be clustered to generate a source spatialtemporal cluster. In some embodiments, for non-invasive sensor data 106, source spatialtemporal clustering can include, for example, grouping adjacent brain regions or cortical sources that show similar activation patterns. The goal is to identify spatially and temporally contiguous regions that exhibit coherent activity, which can help highlight functionally or anatomically meaningful clusters. In some embodiments, spatialtemporal clustering can be basedon significance and size to cluster vertices or voxels that exceed a certain significance level (threshold) or source value and / or a minimum size. A common approach can be to use normalized source activity (z-score) and set the threshold to 2 standard deviations. In addition, cluster-based statistical methods can often use permutation testing to assess the significance of spatial clusters to control for false positives. There are a variety of methodologies for clustering including, for example, hierarchical clustering based on similarity above a chosen level; k-means clustering based on activity patterns, graph-based clustering where vertices are nodes, and edges represent adjacency or similarity; and region-growing methods starting from a seed vertex and iteratively include neighboring vertices that exhibit similar activity until a cluster is formed. The result of surface-based source spatialtemporal clustering (for non-invasive sensor data) on the cortical surface is at least one cortical patch that is active during an epileptic event. The result of volume-based source spatialtemporal clustering (for non-invasive sensor data) on the three- dimensional volume is at least one cluster of voxels that is active during an epileptic event. In some embodiments, the source spatialtemporal clustering can be performed using the epileptogenic network construction module 102. At block 220, the process continues to block 222 in FIG. 2B.

[0073] As mentioned, in some embodiments, the epileptogenic network constructed from the non-invasive sensor data can be, for example, a single-event epileptogenic network (SEEN) or a multi-event epileptogenic network (MEEN). Blocks 202-218 have been described above with respect to a SEEN. When more than one epileptic event is recorded (e.g., by non-invasive sensors), a SEEN can be constructed from each event (e g., using the process of FIGs. 2A and 2B), resulting in a plurality of networks (SEENs). In some embodiments, the plurality of SEENs can be used to construct a multi-event epileptogenic network (MEEN), for example, as described below with respect to FIG. 4. In other embodiments, the process of FIGs. 2A and 2B can be used for constructing a MEEN directly without constructing a separate SEEN for each epileptic event. In such embodiments, the non-invasive sensor data or source data from the plurality of (different) SEENs (e.g., the non-invasive sensor data from the plurality of different events) can be integrated at one of a number of different points in the reconstruction pipeline, in particular, one of blocks 202 and 208-218 (indicated by brackets 221 in FIG. 2A) and the integrated data used for the remainder of the process.

[0074] In some embodiments, the integration can be performed on non-invasive sensor data 106 (block 202) comprising epileptic sensor activity from at least two epileptic event timings. The integration of the non-invasive sensor data can use, but is not limited to, the average, the median, the maximum, or the principal component analysis. The process in FIG. 2A can then continue to blocks 204-218 using the integrated sensor data. In some embodiments, the integration can be performed using the epileptogenic network construction module 102.

[0075] In some embodiments, the integration can be performed on preprocessed non-invasive sensor data (block 208) from at least two epileptic event timings after preprocessing individually each non-invasive sensor data from each of the at least two epileptic event timings. The integration of the preprocessed non-invasive sensor data can use, but is not limited to, the average, the median, the maximum, or the principal component analysis. The process in FIG. 2A can than continue to blocks 210-218 using the integrated preprocessed sensor data. In some embodiments, the integration can be performed using the epileptogenic network construction module 102.

[0076] In some embodiments, the integration can be performed on epileptic epochs (block 212) from at least two events after cropping individually each non-invasive sensor data for each event. The integration of the epileptic epochs can use, but is not limited to, the average, the median, the maximum, or the principal component analysis. The process in FIG. 2A can then continue to blocks 212-218 using the integrated epileptic epochs. In some embodiments, the integration can be performed using the epileptogenic network construction module 102.

[0077] In some embodiments, the integration can be performed on estimated source activity (blocks 212 and 214) from at least two epileptic event timings generated from different epileptic event timings. The integration of estimated source activities can use, but is not limited to, the average, the median, the maximum, or the principal component analysis on each vertex or voxel in the estimated source activities. The process in FIG. 2A can then continue to blocks 216-218 using the integrated estimated source activities. In some embodiments, the integration can be performed using the epileptogenic network construction module 102.

[0078] In some embodiments, the integration can be performed on normalized source activities (block 216) generated from at least two epileptic event timings. The integration of normalized source activities can use, but is not limited to, the average, the median, the maximum, or the principal component analysis on each vertex or voxel in the normalized source activities. Theprocess in FIG. 2A can then continue to block 218 using the integrated normalized source activities. In some embodiments, the integration can be performed using the epileptogenic network construction module 102.

[0079] In some embodiments, the integration can be performed on clustered epileptic epoch source activities generated from at least two epileptic event timings. The integration of clustered source activities can use, but is not limited to, the average, the median, the maximum, or the principal component analysis on each vertex or voxel in the clustered source activities. The process in FIG. 2A can then continue at block 220 to block 222 in FIG. 2B. In some embodiments, the integration can be performed using the epileptogenic network construction module 102.

[0080] At block 224 in FIG. 2B, one or more nodes can be defined. In some embodiments, for each source spatialtemporal cluster (or integrated source spatialtemporal cluster for constructing a MEEN) for non-invasive data (block 218), a point can be selected to represent each cluster as a node in the epileptogenic network. There are several methods that can be used for selecting the point for each node including, for example, (i) selecting the vertex (for surface-based clustering) or voxel (for volume-based clustering) with the highest activity (i.e., the highest significance when the values are normalized to z-scores) and (ii) selecting the center of the cluster. In some embodiments, defining node(s) can be performed using the epileptogenic network construction module 102. At block 226, if the number of nodes (n) is equal to one, then the process continues with block 228 and an epileptogenic network can be constructed that comprises one node and zero edges. As mentioned, in some embodiments, the epileptogenic network can be a SEEN or a MEEN. Construction of a epileptogenic network is discussed further below with respect to block 242. In some embodiments, the construction of the epileptogenic network can be performed using the epileptogenic network construction module 102. At block 230, the epileptogenic network can be stored in data storage (or memory) 122 and / or displayed on a display 124 (e.g., a display 618 of a computer system 600 shown in FIG. 6).

[0081] At block 226, if the number of nodes (n) is greater than or equal to two then edges can be defined and the process can continue at block 232. An edge is a causality relation between 2 nodes. For example, if between 2 nodes, node A and node B, node A’s activity influences node B’s activity after a specific lag in time, then an edge exists between node A and node B (in this case, an A-to-B edge because node A’s activity influences node B’s activity). In someembodiments, defining edges can be performed using the epileptogenic network construction module 102. At block 232, for each node, the source spatialtemporal cluster can be transformed to node activity. A source spatialtemporal cluster represents a group of spatially contiguous vertices (for surface-based clustering) or voxels (for volume-based clustering) that show statistically significant patterns. To transform the source spatialtemporal cluster activity into a node activity, the activity of all the vertices or voxels must be estimated. Then, using a statistic operator such as, but not limited to, weighted average, median, or principal component analysis (PC A) on the different activities will result in the node’s activity. In the case of non-invasive sensor data and surface-based clusters, there can be additional approaches to estimate the node’s activity. The orientation of sources for standard source estimates varies across the cortical surface, which can lead to cancellation (zeroing out). To extract the cluster’s time course from multiple vertices, a variety of techniques can be used including, but not limited to: Max, which uses the maximum absolute value across vertices at each time point within each label; Flip, which finds the dominant direction of source space normal vector orientations, applies a sign-flip to time series at vertices whose orientation is more than 90-degrees different from the dominant direction, and then averages across vertices at each time point within each label; and PCA, which applies singular value decomposition to the time courses and uses the first right-singular vector as the representative label time course. This signal can be scaled to match the average (per- vertex) power and sign-flipped. This sign-flip ensures that extracting time courses does not result in 180-degrees directi on / phase changes. In some embodiments, the transformation of source spatialtemporal clusters to node activity can be performed using the epileptogenic network construction module 102.

[0082] At block 234, for each node pair, source-target influence values can be estimated. As between 2 nodes (in a node pair), node A and node B, if node A’s activity influences node B’s activity (referred to as an A-to-B edge), node A is the source node, node B is the target node, and the source-target influence value represents how much node A’s activity affects node B’s activity after a specific lag in time. Examples of algorithms that estimate the source-target influence value include but are not limited to Granger causality, transfer entropy, conditional mutual information, and partial directed coherence. In some embodiments, estimating source-target influence values can be performed using the epileptogenic network construction module 102.

[0083] In an example, for each node pair, where the total number of node pairs is [n(n-l)] / 2 and each node pair comprises a node A and a node B, an influence value of node A’s activity on node B’s activity can be estimated to generate a source-target influence value, where node A is the source node and node B is the target node. The source-target influence value represents the weight of an edge between node A and node B. In addition, an influence value of node B’s activity on node A’s activity can be estimated to generate a source-target influence value, where node B is the source node and node A is the target node. The source-target influence value represents the weight of an edge between node B and node A.

[0084] At optional block 236 (illustrated with dashed lines in FIG. 2B), for each node pair, source-target influence likelihood values can be estimated. The source-target influence likelihood value of a causal relationship refers to a quantitative measure of how probable it is that one time series (source) has a direct causal influence on another (target). The likelihood can be calculated based on a model that formalizes the supposed causal link between the source activity and the target activity. If the likelihood is high, it suggests that the model, which posits a causal relationship, explains the data well. This indicates strong evidence for causality. If the likelihood is low, the model of causality does not explain the data well, meaning there is weak or no evidence for the causal relationship. The likelihood value is typically calculated through statistical methods like the Likelihood Ratio Test (LRT) or Bayesian inference, but other methods may be used. When using the LRT, the p-value is calculated from the chi-squared distribution using the likelihood ratio statistic and the degrees of freedom. The degrees of freedom for the chi-squared distribution are typically related to the number of lagged terms included. In some embodiments, estimating source-target influence likelihood values can be performed using the epileptogenic network construction module 102.

[0085] In an example, for each node pair, where the total number of node pairs is [n(n-l)] / 2 and each node pair comprises a node A and a node B, a source-target influence likelihood value of node A’s activity on node B’s activity can be estimated to generate a source-target influence likelihood value, where node A is the source node and node B is the target node. In addition, a source-target influence likelihood value of node B’s activity on node A’s activity can be estimated to generate a source-target influence likelihood value, where node B is the source node and node A is the target node.

[0086] At optional block 238 (as illustrated by dashed lines in FIG. 2B), the determined edges can be trimmed based on comparison to a threshold value and / or a likelihood threshold value. In some embodiments, for each edge, the source-target influence value for the edge can be compared to a threshold value and if the source-target influence value is less than the threshold value, the edge can be trimmed. In some embodiments, for each edge, the source-target influence likelihood value for the edge can be compared to a likelihood threshold value and if the sourcetarget influence likelihood value is less than the likelihood threshold value, the edge can be trimmed (removed). For example, if the source-target influence likelihood value is low and the p- value is higher than a specific threshold (for example, but not limited to, 0.05), the edge can be trimmed (removed), because there is no direct significant causality relationship between the two nodes. In some embodiments, trimming edges can be performed using the epileptogenic network construction module 102.

[0087] At block 240, an epileptogenic network can be constructed that comprises n nodes and e edges, where n is greater than or equal to two and e is a number from 0 to n(n-l). As mentioned, in some embodiments, the epileptogenic network can be a SEEN or a MEEN. In some embodiments, the construction of the epileptogenic network can be performed using the epileptogenic network construction module 102. In some embodiments, the epileptogenic network can be represented as a graph. A graph is a mathematical structure representing relationships or connections between entities in a network. It consists of a set of nodes (or vertices) and edges (or links) that connect pairs of nodes. In the epileptogenic network (a graph), nodes represent regions that are active during an epileptic event. As mentioned, nodes are cortical patches (clusters) in the case of non-invasive sensor data. The edges represent the relationships or interactions between the nodes. The epileptogenic network is a directed graph, meaning the edges are directed. A directed edge is a one-way relationship, which means between node A and node B, there are potentially up to 2 edges (before trimming): an A-to-B edge, which represents how node A’s activity affects (influences) node B’s activity, and a B-to-A edge which represents how node B’s activity affects (influences) node A’s activity. In graph theory, an edge with a non-zero value is termed a weighted edge and the value is the weight.

[0088] At block 242, a criticality value for each node can be calculated (or determined). In some embodiments, criticality values can be calculated using a node criticality module 104 of the epileptogenic network construction module 102. Each node in the epileptogenic network has adegree (criticality), representing how much the node affects the rest of the network. In some embodiments, the criticality value can be calculated as, for example, the sum of all the sourcetarget influence values of all the edges originating from the node. In graph theory, this is termed the weighted out-degree centrality. When the weighted directed edges and the degree of the nodes are calculated based on the nodes' activity (based on non-invasive sensor data) from one epileptic event, the result is the single-event epileptogenic network (SEEN). An example of constructing a SEEN and calculating node criticality is discussed below with respect to FIG. 5. As mentioned, when more than one epileptic event is recorded (by invasive or non-invasive sensors), a single-event epileptogenic network (SEEN) can be constructed from each event, resulting in multiple networks (SEENs). These multiple SEENs can be used to construct a multievent epileptogenic network (MEEN). Integration of non-invasive sensor data or source data from a plurality of SEENs (e.g., from a plurality of epileptic events) during the process of FIGs. 2A and 2B for constructing a MEEN is discussed above. An example of constructing a MEEN from a plurality of constructed SEENs and calculating node criticality is discussed below with respect to FIGs. 4 and 6 A and 6B.

[0089] At block 244, the epileptogenic network and / or the node criticality can be stored in data storage (or memory) 122 and / or displayed on a display 124 (e.g., a display 618 of a computer system 600 shown in FIG. 6). In some embodiments, the epileptogenic network and or the node criticality can be displayed as a 2D, 3F, or volumetric display. As mentioned, the constructed epileptogenic network and determined node criticality (i.e., critical nodes) can be used in surgical planning, surgical simulation, and surgical execution to, for example, identify surgical targets (nodes) beyond seizure onset zones. For example, in some embodiments, surgical targets can be selected from nodes in the epileptogenic network based on the determined node criticality. In some embodiments, the selected nodes can be critical nodes that can be defined as, for example, the nodes with the k-highest value, or the nodes with values greater than a certain threshold (e.g., the nodes in the constructed epileptogenic network that affect the rest of the network significantly more than the other nodes). In some embodiments, the selected targets can be used for surgical treatments such as resection, ablation, stimulation, or a combination of treatments.

[0090] As mentioned, in some embodiments, an epileptogenic network can be constructed using invasive sensor data 108. FIGs. 3A and 3B illustrate a method for constructing an epileptogenic network using invasive sensor data in accordance with an embodiment. The process illustrated inFIGs. 3 A and 3B is described below as being carried out by the system 100 as illustrated in FIG. 1. However, in some examples, the process of FIGs. 3 A and 3B may be implemented by another system. Although the blocks of the process are illustrated in a particular order, in some embodiments, one or more blocks may be executed in a different order than illustrated in FIGs. 3A and 3B or may be bypassed.

[0091] At block 302, invasive sensor data 106 for a subject (e.g., of the head and brain of the subject) can be received. As mentioned, the invasive sensor data can be, for example, Examples of invasive sensor data include, but are not limited to, stereo-electroencephalography (sEEG), epidural electrocorticography (ECoG), and microelectrode arrays. In some embodiments, the invasive sensor data includes epileptic sensor activity. In some embodiment, the invasive sensor data can also optionally include non-epileptic sensor data. In some embodiments, the invasive sensor data can be received (or retrieved) from data storage (or memory) such as, for example, data storage 122 of system 100, data storage of the system used to acquire the invasive sensor data, or data storage of other computer systems (e.g., storage device 616 of the example computer system 600 shown in FIG. 6). At optional block 304 (illustrated with dashed lines in FIG. 3 A), in some embodiments, anatomical data 110 for the subject may be received. As mentioned, the anatomical data can be determined using, for example, magnetic resonance imaging (MRI) or other modalities. In some embodiments, the optional anatomical data 110 may be retrieved or received from data storage (or memory, or database) such as, for example, data storage 122 of system 100, data storage of the system (e.g., an MRI system) used to acquire the anatomical data, or data storage of other computer systems (e g., storage device 616 of the example computer system 600 shown in FIG. 6). At block 306, in some embodiments, epileptic event timing 112 may be received. As mentioned, epileptic event timing exists for both ictal events and interictal events and is commonly determined through visual inspection of the captured time series data (from non-invasive and invasive sensors). In some embodiments, this process can also be automated through automatic detection algorithms. For ictal events, the epileptic timing is the seizure onset timing which is the precise moment in time when a seizure begins. For interictal events, the epileptic timing is the onset timing of the electrical discharges known as interictal epileptiform discharges (lEDs). In some embodiments, the epileptic timing 112 may be retrieved or received from data storage (or memory, database) such as, for example, data storage 122 of system 100, data storage of the system used to determine the epileptic timingdata, or data storage of other computer systems (e.g., storage device 616 of the example computer system 600 shown in FIG. 6).

[0092] At block 308, optional pre-processing (illustrated with dashed lines in FIG. 3 A) may be performed on, for example, the invasive sensor data and / or the anatomical data (e.g., using a preprocessing module 114). As mentioned, the preprocessing of invasive sensor data is similar to the preprocessing of non-invasive sensor data (discussed above with respect to FIGs. 1 and 2A) except invasive sensors typically capture higher-frequency signals (HFOs), whereas non-invasive sensors tend to capture more low- to mid-frequency ranges; non-invasive sensor data is more prone to external artifacts while invasive sensor data is subject to more localized artifacts like cardiac signals; and non-invasive sensors have more flexibility in reference selection, while invasive sensors usually employ bipolar referencing between adjacent electrodes. In some embodiments, as mentioned, the preprocessing of anatomical data can include, for example, reconstructing cortical surfaces and segmenting sub-cortical regions.

[0093] In some embodiments, the invasive sensor data 108 (unprocessed or pre-processed), the epileptic event timing 112, and in some embodiments the optional anatomical data 110 (unprocessed or preprocessed) can be provided as input to, for example, an epileptogenic network construction module 102. At block 310, the invasive sensor data 108 can be cropped to generate epileptic epoch. In addition, in some embodiments, the invasive sensor data may also be cropped to generate non-epileptic epoch. Cropping refers to the process of selecting and retaining a specific segment of the data based on defined time intervals. It can involve, for example, specifying a time window (e g., from 2 seconds before an event to 4 seconds after) and discarding data outside of that window. Cropping is commonly used to extract epochs around an event of interest (e.g., seizure onset). In some embodiments, cropping can be done manually. In some embodiments, cropping can be done by an automated procedure (e.g., using the epileptogenic network construction module 102). Epileptic epoch is the segment of data that results after the cropping invasive sensor data, and epileptic epoch comprises at least one epileptic event. The optional non-epileptic epoch is the segment of data that results after cropping invasive sensor data, and non-epileptic epoch does not comprise at least one epileptic event (i.e., there are no epileptic events.

[0094] At optional block 312 (illustrated with dashed lines in FIG. 3A), in some embodiments, the epileptic epoch (block 310) can be normalized using, for example, the non-epileptic epoch(block 310). In some embodiments, the source activity for the epileptic epoch and the nonepileptic epoch is the invasive sensor data from the invasive sensor(s). In some embodiments, normalizing the epileptic epoch with the non-epileptic epoch can include, for example, for each time point in the epileptic epoch, subtracting the average (over time) of the non-epileptic epoch source activity (i.e., the invasive sensor data) from the epileptic epoch source activity (i.e., the invasive sensor data) to calculate baseline-corrected values. In addition, in some embodiments, the baseline-corrected values can be optionally divided by the standard deviation of the non- epileptic source activity to calculate z-scores. The process of normalization can be on a vertex- by-vertex basis or a voxel-by-voxel basis. Normalization controls for noise and drift over time, can make it easier to identify significant changes during the epileptic event, and can allow for the removal of bias resulting from changes in signal strength across regions or time. In some embodiments, normalization of the epileptic epoch can be performed using the epileptogenic network construction module 102. At block 314 the process continues to block 316 in FIG. 3B.

[0095] As mentioned, in some embodiments, the epileptogenic network constructed from the invasive sensor data can be, for example, a single-event epileptogenic network (SEEN) or a multi-event epileptogenic network (MEEN). Blocks 302-312 have been described above with respect to a SEEN. When more than one epileptic event is recorded (e.g., by invasive sensors), a SEEN can be constructed from each event (e.g., using the process of FIGs. 3A and 3B), resulting in a plurality of networks (SEENs). In some embodiments, the plurality of SEENs can be used to construct a multi-event epileptogenic network (MEEN), for example, as described below with respect to FIG. 4. In other embodiments, the process of FIGs. 3 A and 3B can be used for constructing a MEEN directly without constructing a separate SEEN for each epileptic event. In such embodiments, the invasive sensor data from the plurality of (different) SEENs (e.g., the invasive sensor data from the plurality of different events) can be integrated at one of a number of different points in the reconstruction pipeline, in particular, one of blocks 302 and 308-312 (indicated by brackets 315 in FIG. 3 A) and the integrated data used for the remainder of the process.

[0096] In some embodiments, the integration can be performed on invasive sensor data 108 (block 302) comprising epileptic sensor activity from at least two epileptic event timings. The integration of the invasive sensor data can use, but is not limited to, the average, the median, the maximum, or the principal component analysis. The process in FIG. 3A can then continue toblocks 308-312 using the integrated invasive sensor data. In some embodiments, the integration can be performed using the epileptogenic network construction module 102.

[0097] In some embodiments, the integration can be performed on preprocessed invasive sensor data (block 308) from at least two epileptic event timings after preprocessing individually each invasive sensor data for each epileptic event. The integration of the preprocessed invasive sensor data can use, but is not limited to, the average, the median, the maximum, or the principal component analysis. The process of FIG. 3A can then continue to blocks 310-312 using the integrated preprocessed invasive sensor data. In some embodiments, the integration can be performed using the epileptogenic network construction module 102.

[0098] In some embodiments, the integration can be performed on epileptic epochs (block 310) from at least two epileptic event timings after cropping individually each invasive sensor data. The integration of the epileptic epochs can use, but is not limited to, the average, the median, the maximum, or the principal component analysis. The process of FIG. 3A can then continue to block 312 using the integrated epileptic epochs. In some embodiments, the integration can be performed using the epileptogenic network construction module 102.

[0099] In some embodiments, the integration can be performed on normalized epileptic epochs (block 312) from at least two epileptic event timings after normalizing individually each epileptic epoch. The integration of the epochs can use, but is not limited to, the average, the median, the maximum, or the principal component analysis. The process in FIG. 3 A can then continue at block 314 to block 316 in FIG. 3B. In some embodiments, the integration can be performed using the epileptogenic network construction module 102.

[0100] At block 318 in FIG. 3B, one or more nodes can be defined. In some embodiments, for invasive data, defining a node can be accomplished using a variety of means including, but not limited to, a monopolar montage and a bipolar montage. For a monopolar montage, each contact (sensor) can represent a node in the epileptogenic network. For a bipolar montage, the middle point between two adjunct contacts (sensors) can represent a node. In some embodiments, defining node(s) can be performed using the epileptogenic network construction module 102. At block 320, if the number of nodes (n) is equal to one, then the process continues with block 322 and an epileptogenic network can be constructed that comprises one node and zero edges. As mentioned, in some embodiments, the epileptogenic network can be a SEEN or a MEEN.Construction of a epileptogenic network is discussed further below with respect to block 334. Insome embodiments, the construction of the epileptogenic network can be performed using the epileptogenic network construction module 102. At block 324, the epileptogenic network can be stored in data storage (or memory) 122 and / or displayed on a display 124 (e.g., a display 618 of a computer system 600 shown in FIG. 6).

[0101] At block 320, if the number of nodes (n) is greater than or equal to two then edges can be defined and the process can continue at block 326. An edge is a causality relation between 2 nodes. For example, if between 2 nodes, node A and node B, node A’s activity influences node B’s activity after a specific lag in time, then an edge exists between node A and node B (in this case, an A-to-B edge because node A’s activity influences node B’s activity). In some embodiments, defining edges can be performed using the epileptogenic network construction module 102. At block 326, the epileptic epoch (e.g., the sensor data) can be transformed to node activity. The data recorded from invasive sensors (such as electrodes) can be referenced in different ways, known as montages. Two common types are monopolar and bipolar montages. When using monopolar montage (also called referential montage), each invasive sensor can be referenced to a common reference point (e.g., a scalp electrode, an average of several electrodes). By using the monopolar montage, electrical activity can be captured at each electrode relative to the single common reference. In a bipolar montage, each electrode can be referenced to a neighboring electrode rather than to a common point and voltage difference between two adjacent electrodes can be captured. This process is repeated for all pairs in the invasive sensor chain (electrode chain). In some embodiments, the transformation of epileptic epoch to node activity can be performed using the epileptogenic network construction module 102.

[0102] At block 328, for each node pair, source-target influence values can be estimated. As between 2 nodes (in a node pair), node A and node B, if node A’s activity influences node B’s activity (referred to as an A-to-B edge), node A is the source node, node B is the target node, and the source-target influence value represents how much node A’s activity affects node B’s activity after a specific lag in time. Examples of algorithms that estimate the source-target influence value include but are not limited to Granger causality, transfer entropy, conditional mutual information, and partial directed coherence. In some embodiments, estimating source-target influence values can be performed using the epileptogenic network construction module 102.

[0103] In an example, for each node pair, where the total number of node pairs is [n(n-l)] / 2 and each node pair comprises a node A and a node B, an influence value of node A’s activity on node B’s activity can be estimated to generate a source-target influence value, where node A is the source node and node B is the target node. The source-target influence value represents the weight of an edge between node A and node B. In addition, an influence value of node B’s activity on node A’s activity can be estimated to generate a source-target influence value, where node B is the source node and node A is the target node. The source-target influence value represents the weight of an edge between node B and node A.

[0104] At optional block 330 (illustrated with dashed lines in FIG. 3B), for each node pair, source-target influence likelihood values can be estimated. The source-target influence likelihood value of a causal relationship refers to a quantitative measure of how probable it is that one time series (source) has a direct causal influence on another (target). The likelihood can be calculated based on a model that formalizes the supposed causal link between the source activity and the target activity. If the likelihood is high, it suggests that the model, which posits a causal relationship, explains the data well. This indicates strong evidence for causality. If the likelihood is low, the model of causality does not explain the data well, meaning there is weak or no evidence for the causal relationship. The likelihood value is typically calculated through statistical methods like the Likelihood Ratio Test (LRT) or Bayesian inference, but other methods may be used. When using the LRT, the p-value is calculated from the chi-squared distribution using the likelihood ratio statistic and the degrees of freedom. The degrees of freedom for the chi-squared distribution are typically related to the number of lagged terms included. In some embodiments, estimating source-target influence likelihood values can be performed using the epileptogenic network construction module 102.

[0105] In an example, for each node pair, where the total number of node pairs is [n(n-l)] / 2 and each node pair comprises a node A and a node B, a source-target influence likelihood value of node A’s activity on node B’s activity can be estimated to generate a source-target influence likelihood value, where node A is the source node and node B is the target node. In addition, a source-target influence likelihood value of node B’s activity on node A’s activity can be estimated to generate a source-target influence likelihood value, where node B is the source node and node A is the target node.

[0106] At optional block 332 (as illustrated by dashed lines in FIG. 3B), the determined edges can be trimmed based on comparison to a threshold value and / or a likelihood threshold value. In some embodiments, for each edge, the source-target influence value for the edge can be compared to a threshold value and if the source-target influence value is less than the threshold value, the edge can be trimmed. In some embodiments, for each edge, the source-target influence likelihood value for the edge can be compared to a likelihood threshold value and if the sourcetarget influence likelihood value is less than the likelihood threshold value, the edge can be trimmed (removed). For example, if the source-target influence likelihood value is low and the p- value is higher than a specific threshold (for example, but not limited to, 0.05), the edge can be trimmed (removed), because there is no direct significant causality relationship between the two nodes. In some embodiments, trimming edges can be performed using the epileptogenic network construction module 102.

[0107] At block 334, an epileptogenic network can be constructed that comprises n nodes and e edges, where n is greater than or equal to two and e is a number from 0 to n(n-l). As mentioned, in some embodiments, the epileptogenic network can be a SEEN or a MEEN. In some embodiments, the construction of the epileptogenic network can be performed using the epileptogenic network construction module 102. In some embodiments, the epileptogenic network can be represented as a graph. A graph is a mathematical structure representing relationships or connections between entities in a network. It consists of a set of nodes (or vertices) and edges (or links) that connect pairs of nodes. In the epileptogenic network (a graph), nodes represent regions that are active during an epileptic event. As mentioned, nodes are sensors in the case of invasive sensor data. The edges represent the relationships or interactions between the nodes. The epileptogenic network is a directed graph, meaning the edges are directed. A directed edge is a one-way relationship, which means between node A and node B, there are potentially up to 2 edges (before trimming): an A-to-B edge, which represents how node A’s activity affects (influences) node B’s activity, and a B-to-A edge which represents how node B’s activity affects (influences) node A’s activity. In graph theory, an edge with a non-zero value is termed a weighted edge and the value is the weight.

[0108] At block 336, a criticality value for each node can be calculated (or determined). In some embodiments, criticality values can be calculated using a node criticality module 104 of the epileptogenic network construction module 102. Each node in the epileptogenic network has adegree (criticality), representing how much the node affects the rest of the network. In some embodiments, the criticality value can be calculated as, for example, the sum of all the sourcetarget influence values of all the edges originating from the node. In graph theory, this is termed the weighted out-degree centrality. When the weighted directed edges and the degree of the nodes are calculated based on the nodes' activity (based on invasive sensor data) from one epileptic event, the result is the single-event epileptogenic network (SEEN). An example of constructing a SEEN and calculating node criticality is discussed below with respect to FIG. 5. As mentioned, when more than one epileptic event is recorded (by invasive sensors), a singleevent epileptogenic network (SEEN) can be constructed from each event, resulting in multiple networks (SEENs). These multiple SEENs can be used to construct a multi-event epileptogenic network (MEEN). Integration of invasive sensor data or source data from a plurality of SEENs (e.g., from a plurality of epileptic events) during the process of FIGs. 3A and 3B for constructing a MEEN is discussed above. An example of constructing a MEEN from a plurality of constructed SEENs and calculating node criticality is discussed below with respect to FIGs. 4 and 6 A and 6B.

[0109] At block 338, the epileptogenic network and / or the node criticality values can be stored in data storage (or memory) 122 and / or displayed on a display 124 (e.g., a display 618 of a computer system 600 shown in FIG. 6). As mentioned, the constructed epileptogenic network and determined node criticality can be used in surgical planning, surgical simulation, and surgical execution to, for example, identify surgical targets (e.g., critical nodes) beyond seizure onset zones. For example, in some embodiments, surgical targets can be selected from nodes in the epileptogenic network based on the determined node criticality. In some embodiments, the selected nodes can be critical nodes that can be defined as, for example, the nodes with the k- highest criticality value, or the nodes with criticality values greater than a certain threshold (e.g., the nodes in the constructed epileptogenic network that affect the rest of the network significantly more than the other nodes). In some embodiments, the selected targets can be used for surgical treatments such as resection, ablation, stimulation, or a combination of treatments.

[0110] As mentioned, when more than one epileptic event is recorded by the sensors (e.g., either non-invasives sensors or invasive sensors), in some embodiments a SEEN can be reconstructed from each event (e.g., using the process of FIGs. 2A-2B or FIGs. 3A-3B resulting in a plurality of SEENs that can then be used to construct a MEEN. FIG. 4 illustrates a method forconstructing a multi-event epileptogenic network (MEEN) from single-event epileptogenic networks (SEENs) in accordance with an embodiment. At block 402, a closeness value and a minimum sample value for a clustering process. Closeness is the maximum distance between two nodes to be clustered into the same cluster. Minimum-samples is the minimum number of nodes in a cluster required to become a compiled node in the MEEN. In some embodiments, the closeness value and the minimum samples value can be selected manually or automatically. At block 404, a clustering process can be performed to cluster the nodes from the plurality of SEENs to generate compiled nodes for the MEEN. In some embodiments, various algorithms can be used for clustering SEENs into a MEEN, such as but not limited to, the density-based spatial clustering of applications with noise (DBSCAN) algorithm that takes into account both closeness and minimum-samples. At block 406, a criticality value can be calculated for each compiled node in the MEEN. In some embodiments, criticality values can be calculated using a node criticality module 104 of the epileptogenic network construction module 102. Each node in the epileptogenic network has a degree (criticality), representing how much the node affects the rest of the network. In some embodiments, the criticality value of a compiled node can be calculated as, for example, the sum of the edges (e.g., the source target influence values) originating from a compiled node. In some embodiments, in the MEEN, the edges between each node pair can be aggregated to represent all edges as a single edge between each node pair. One example method of aggregating is averaging the edge values for the source-target influence value and source-target influence likelihood value. In other embodiments, the criticality value for each compiled node can be determined as maximum value of the edges (e g., the source target influence values) originating from a compiled node. At block 408. the MEEN and / or the node criticality values can be stored in data storage (or memory) 122 and / or displayed on a display 124 (e.g., a display 618 of a computer system 600 shown in FIG. 6).[OHl] The following examples set forth, in detail, ways in which the present disclosure was evaluated and ways in which the present disclosure may be used or implemented, and will enable one of ordinary skill in the art to more readily understand the principles thereof. The following examples are presented by way of illustration and are not meant to be limiting in any way.EXAMPLE 1

[0112] FIG. 5 illustrates an example single-event epileptogenic network (SEEN) in accordance with an embodiment. FIG. 4 depicts a SEEN with 3 nodes (A, B, and C), and 4 edges (A-to-B, A-to-C, C-to-A, and C-to-B). Each arrow represents an edge with two properties (x,y) where x is the source-target influence value and the y is source-target influence likelihood value. For each edge, the node from which the arrow originates is the source node, and the node to which the arrow points is the target node. To calculate the likelihood threshold value using the Likelihood Ratio Test (LRT), the p-value can be calculated from the chi-squared distribution. In this scenario, the threshold can be based on the common concept of p-value <= 0.05, setting the likelihood threshold to be 1.3 (-loglO(p-value)). Any edge with a source-target influence likelihood value smaller than 1.3 will be trimmed (removed), which results in the C-to-B edge being trimmed. To calculate the criticality value for each node, all the source-target influence values for all the non-trimmed edges originating from the node (i.e., all the non-trimmed edges where the node is the source node) can be summed:• node A: 2 (A-to-B edge) + 3 (A-to-C edge) = 5• node B: 0 (no edges where node B is the source node) = 0• node C: 4 (C-to-A edge) = 4.

[0113] In this example, node A is the most critical node and node B is the least critical node.EXAMPLE 2

[0114] FIG. 6A illustrates two example single-event epileptogenic networks in accordance with an embodiment and FIG. 6B illustrates an example multi-event epileptogenic network (MEEN) constructed from the single-event epileptogenic networks of FIG. 5 A in accordance with an embodiment. FIG. 6A depicts 2 SEENs, a first SEEN and a second SEEN, in the same space. The first SEEN comprises node A, node B, node C, and node D, and the second SEEN comprises node X, node Y, and node Z. Each arrow represents an edge with two properties (x,y) where x is the source-target influence value and the y is source-target influence likelihood value. For each edge, the node from which the arrow originates is the source node, and the node to which the arrow points is the target node. The locations of the nodes in FIG. 6A (and compiled nodes FIG. 6B) are relative to the nodes’ 3D coordinates in the patient’s MRI space. For example, node A and node X are close in FIG 5A, representing two close 3D coordinates.

[0115] FIG. 6B depicts a MEEN constructed from the 2 SEENs in FIG. 5A. To construct the MEEN from the 2 SEENs, close nodes can be clustered to form compiled nodes. In this example, node A and node X (both in FIG. 6A) are clustered into (new) compiled node Ml (in FIG. 6B), node B and node Y are clustered into compiled node M2, and node C and node Z are clustered to form compiled node M3. And node D will become a new node, M4. As mentioned, two parameters can affect the clustering step: closeness and minimum-samples. Closeness is the maximum distance between two nodes to be clustered into the same cluster. For example, if closeness value is smaller than the distance between node C and node Z, each node will become its own compiled node in the MEEN. Minimum-samples is the minimum number of nodes in a cluster required to become a compiled node in the MEEN. In this example, the minimumsamples value is set to 1. If the minimum-samples value was instead set to 2, node D would not become a compiled node in the MEEN and would instead be discarded. There are several algorithms that can be used for clustering SEENs into a MEEN, such as but not limited to, the density -based spatial clustering of applications with noise (DBSCAN) algorithm that takes into account both closeness and minimum-samples.

[0116] In the MEEN, the edges between each node pair can be aggregated to represent all edges as a single edge between each node pair. One method of aggregating is averaging the edge values for the source-target influence value and source-target influence likelihood value. For example, for the node pair of node Ml and node M3, the edges from node Ml to node M3 are (3.2, 2.3) and (2.1, 2.4), and the aggregated edge from node Ml to M3 can be calculated by taking the maximum value for each property: (3.2, 2.4).

[0117] One method for calculating the criticality value is to directly calculate the criticality value for all the nodes, based on the edges originating from each node:• node Ml : 4 (Ml-to-M2 edge) + 3.2 (Ml-to-M3 edge) = 7.2• node M2: 0 (no edges where node M2 is the source node) = 0• node M3: 4.1 (M3-to-Ml edge) + 2.2 (M3-to-M2 edge) = 6.3 [the other M3-to-Ml edge (1.5, 0.8) and the other M3-to-M2 edge (2, 1.1) are trimmed because they are less than the likelihood threshold]• node M4: 4.2 (M4-to-M2 edge) = 4.2.

[0118] In this example, node Ml is the most critical node and node M2 is the least critical node.

[0119] FIG. 7 is a block diagram of an example computer system in accordance with an embodiment. Computer system 600 may be used to implement the systems and methods described herein. In some embodiments, the computer system 700 may be a workstation, a notebook computer, a tablet device, a mobile device, a multimedia device, a network server, a mainframe, one or more controllers, one or more microcontrollers, or any other general-purpose or application-specific computing device. The computer system 700 may operate autonomously or semi-autonomously, or may read executable software instructions from the memory or storage device 716 or a computer-readable medium (e.g., a hard drive, a CD-ROM, flash memory), or may receive instructions via the input device 720 from a user, or any other source logically connected to a computer or device, such as another networked computer or server. Thus, in some embodiments, the computer system 700 can also include any suitable device for reading computer-readable storage media.

[0120] Data, such as data acquired with an imaging system (e.g., a magnetic resonance imaging (MRI) system) may be provided to the computer system 700 from a data storage device 716, and these data are received in a processing unit 702. In some embodiment, the processing unit 702 includes one or more processors. For example, the processing unit 702 may include one or more of a digital signal processor (DSP) 704, a microprocessor unit (MPU) 706, and a graphics processing unit (GPU) 708. The processing unit 702 also includes a data acquisition unit 710 that is configured to electronically receive data to be processed. The DSP 704, MPU 706, GPU 708, and data acquisition unit 710 are all coupled to a communication bus 712. The communication bus 712 may be, for example, a group of wires, or a hardware used for switching data between the peripherals or between any components in the processing unit 702.

[0121] The processing unit 702 may also include a communication port 714 in electronic communication with other devices, which may include a storage device 716, a display 718, and one or more input devices 720. Examples of an input device 720 include, but are not limited to, a keyboard, a mouse, and a touch screen through which a user can provide an input. The storage device 716 may be configured to store data, which may include data such as, for example, non- invasive sensor data, invasive sensor data, anatomical data, epileptic event timing, epileptogenic network(s), node criticality values, etc. whether these data are provided to, or processed by, the processing unit 702. The display 718 may be used to display images and other information, such as magnetic resonance images, patient health data, and so on.

[0122] The processing unit 702 can also be in electronic communication with a network 722 to transmit and receive data and other information. The communication port 714 can also be coupled to the processing unit 702 through a switched central resource, for example the communication bus 712. The processing unit can also include temporary storage 724 and a display controller 726. The temporary storage 724 is configured to store temporary information. For example, the temporary storage 724 can be a random access memory.

[0123] Computer-executable instructions for constructing an epileptogenic network (e.g., SEEN or MEEN) using non-invasive sensor data or invasive sensor data and calculating node criticality according to the above-described methods may be stored on a form of computer readable media. Computer readable media includes volatile and nonvolatile, removable, and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer readable media includes, but is not limited to, random access memory (RAM), read-only memory (ROM), electrically erasable programmable ROM (EEPROM), flash memory or other memory technology, compact disk ROM (CD-ROM), digital volatile disks (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired instructions and which may be accessed by a system (e.g., a computer), including by internet or other computer network form of access.

[0124] The present technology has been described in terms of one or more preferred embodiments, and it should be appreciated that many equivalents, alternatives, variations, and modifications, aside from those expressly stated, are possible and within the scope of the invention.

Claims

CLAIMS1. A computer implemented method for constructing an epileptogenic network using non- invasive sensor data where the epileptogenic network comprises n nodes and e edges, the method comprising: receiving non-invasive sensor data of a subject comprising epileptic sensor activity; receiving epileptic event timing; cropping the non-invasive sensor data of the subject based on the epileptic event timing to generate an epileptic epoch; estimating source activity of the epileptic epoch to generate an estimated epileptic epoch source activity; clustering the estimated epileptic epoch source activity to generate at least one source spatialtemporal cluster; defining one or more nodes by defining a point in each of the at least one source spatialtemporal cluster; when the one or more nodes comprises one node, constructing an epileptogenic network comprising one node and zero edges; when the one or more nodes comprising two or more nodes, for each node transforming each at least one source spatialtemporal cluster to a node activity; for each node pair in the two or more nodes, estimating a source-target influence value; and constructing an epileptogenic network comprising two or more nodes and e edges based on the node pairs in the two or more nodes and the estimated source-target influence value for each node pair, wherein e is a number between 0 to n(n-l).

2. The computer implemented method according to claim 1, wherein the epileptogenic network with one node and zero edges and the epileptogenic network with two or more nodes and e edges are single-event epileptogenic networks (SEENs).

3. The computer implemented method according to claim 1, wherein the non-invasive sensor data of the subject comprises non-invasive sensor data for a plurality of epileptic events and wherein constructing an epileptogenic network comprising two or more nodes and e edgesfurther comprises constructing a single-event epileptogenic network (SEEN) for each epileptic event of the plurality of epileptic events.

4. The computer implemented method according to claim 3, wherein constructing an epileptogenic network comprising n nodes and e edges further comprises constructing a multievent epileptogenic network by clustering nodes from each of the single-event epileptogenic networks (SEENs) to generate compiled nodes.

5. The computer implemented method according to claim 1, wherein the non-invasive sensor data of the subject further comprises non-epileptic sensor activity.

6. The computer implemented method according to claim 1, further comprising receiving anatomical data for the subject.

7. The computer implemented method according to claim 6, further comprising preprocessing one or more of the non-invasive sensor data and the anatomical data.

8. The computer implemented method according to claim 1, further comprising: cropping the non-invasive sensor data of the subject based on a timing with no associated epileptic event to generate at least one non-epileptic epoch; estimating source activity of the at least one non-epileptic epoch to generate at least one estimated non-epileptic epoch source activity; and normalizing the estimated epileptic epoch source activity using the at least one estimated non-epileptic epoch source activity.

9. The computer implemented method according to claim 1, further comprising: for each node pair in the two or more nodes, estimating a source-target influence likelihood value; removing each edge where the source-target influence value is less than a predetermined threshold value; andremoving each edge where the source target influence likelihood value is less than a likelihood threshold value.

10. The computer implemented method according to claim 1, further comprising determining a criticality value for each node in the epileptogenic network comprising two or more nodes and e edges.

11. The computer implemented method according to claim 1, wherein the non-invasive sensor data is at least one of magnetoencephalography (MEG) data and non-invasive electroencephalography (EEG) data.

12. The computer implemented method according to claim 6, wherein the anatomical data is magnetic resonance imaging (MRI) data.

13. The computer implemented method according to claim 1, wherein the epileptogenic network with one node and zero edges and the epileptogenic network with two or more nodes and e edges are configured for use with at least one of surgical planning, surgical simulation, and surgical execution.

14. A computer implemented method for constructing an epileptogenic network using invasive sensor data where the epileptogenic network comprises n nodes and e edges, the method comprising: receiving invasive sensor data of a subject comprising epileptic sensor activity; receiving epileptic event timing; cropping the invasive sensor data of the subject based on the epileptic event timing to generate an epileptic epoch; defining one or more nodes; when the one or more nodes comprises one node, constructing a epileptogenic network comprising one node and zero edges; when the one or more nodes comprising two or more nodes, for each node transforming the epileptic epoch to a node activity;for each node pair in the two or more nodes, estimating a source-target influence value; and constructing an epileptogenic network comprising two or more nodes and e edges based on the node pairs in the two or more nodes and the estimated source-target influence value for each node pair, wherein e is a number between 0 to n(n-l).

15. The computer implemented method according to claim 14, wherein the epileptogenic network with one node and zero edges and the epileptogenic network with two or more nodes and e edges are single-event epileptogenic networks (SEENs).

16. The computer implemented method according to claim 14, wherein the invasive sensor data of the subject comprises invasive sensor data for a plurality of epileptic events and wherein constructing an epileptogenic network comprising two or more nodes and e edges further comprises constructing a single-event epileptogenic network (SEEN) for each epileptic event of the plurality of epileptic events.

17. The computer implemented method according to claim 16, wherein constructing an epileptogenic network comprising n nodes and e edges further comprises constructing a multievent epileptogenic network by clustering nodes from each of the single-event epileptogenic networks (SEENs) to generate compiled nodes.

18. The computer-implemented method according to claim 14, wherein the invasive sensor data of the subject further comprises non-epileptic sensor activity.

19. The computer implemented method according to claim 14, further comprising receiving anatomical data for the subject.

20. The computer implemented method according to claim 19, further comprising preprocessing one or more of the invasive sensor data and the anatomical data.

21. The computer implemented method according to claim 14, further comprising: cropping the invasive sensor data of the subject based on a timing with no associated epileptic event to generate at least one non-epileptic epoch; and normalizing the epileptic epoch using the at least one estimated non-epileptic epoch.

22. The computer implemented method according to claim 14, further comprising: for each node pair in the two or more nodes, estimating a source-target influence likelihood value; removing each edge where the source-target influence value is less than a predetermined threshold value; and removing each edge where the source target influence likelihood value is less than a likelihood threshold value.

23. The computer implemented method according to claim 14, further comprising determining a criticality value for each node in the epileptogenic network comprising two or more nodes and e edges.

24. The computer implemented network according to claim 14, wherein the invasive sensor data is acquired using at least one of depth electrodes, subdural electrodes, epidural electrodes, foramen ovale electrodes, and microelectrodes.

25. The computer implemented network according to claim 14, wherein the invasive sensor data is at least one of stereoelectroencephalography (sEEG) data and electrocorticography (ECoG) data.

26. The computer implemented method according to claim 14, wherein the anatomical data is MRI data.

27. The computer implemented method according to claim 14, wherein the epileptogenic network with one node and zero edges and the epileptogenic network with two or more nodesand e edges are configured for use with at least one of surgical planning, surgical simulation, and surgical execution.

28. A computer implemented method for constructing a multi-event epileptogenic network (MEEN), the method comprising: receiving sensor data for a subject for each of a plurality of epileptic events, the sensor data for each epileptic event comprising epileptic sensor activity; integrating the sensor data for each epileptic event to form a set of integrated multi-event sensor data; defining one or more nodes; when the one or more nodes comprises one node, constructing a multi-event epileptogenic network comprising one node and zero edges; when the one or more nodes comprising two or more nodes; estimating a source-target influence value for each node pair in the two or more nodes; and constructing a multi-event epileptogenic network comprising two or more nodes and e edges based on the node pairs in the two or more nodes and the estimated source-target influence value for each node pair, wherein e is a number between 0 to n(n-l).

29. The computer implemented method according to claim 28, wherein the sensor data for a subject for each of a plurality of epileptic events comprises non-invasive sensor data and wherein integrating the sensor data for each epileptic event to form the set of integrated multi-event sensor data comprises integrating the non-invasive sensor data of each of the plurality of epileptic events.

30. The computer implemented method according to claim 28, wherein the sensor data for a subject for each of a plurality of epileptic events comprises non-invasive sensor data, the method further comprising: pre-processing the non-invasive sensor data for each of the plurality of epileptic events; andwherein integrating the sensor data for each epileptic event to form the set of integrated multi -event sensor data comprises integrating the preprocessed non-invasive sensor data of each of the plurality of epileptic events.

31. The computer implemented method according to claim 28, wherein the sensor data for a subject for each of a plurality of epileptic events comprises non-invasive sensor data, the method further comprising: cropping the non-invasive sensor data for each of the plurality of epileptic events to form an estimated epileptic epoch for each epileptic event; and wherein integrating the sensor data for each epileptic event to form the set of integrated multi-event sensor data comprises integrating the estimated epileptic epoch of each of the plurality of epileptic events.

32. The computer implemented method according to claim 28, wherein the sensor data for a subject for each of a plurality of epileptic events comprises non-invasive sensor data, the method further comprising: cropping the non-invasive sensor data for each of the plurality of epileptic events to form an estimated epileptic epoch for each epileptic event; and estimating source activity of the epileptic epoch of each of the plurality of epileptic events to generate an estimated epileptic epoch source activity for each of the plurality of epileptic events; wherein integrating the sensor data for each epileptic event to form the set of integrated multi-event sensor data comprises integrating the estimated epileptic epoch source activity of each of the plurality of epileptic events.

33. The computer implemented method according to claim 28, wherein the sensor data for a subject for each of a plurality of epileptic events comprises non-invasive sensor data, the method further comprising: cropping the non-invasive sensor data for each of the plurality of epileptic events to form an estimated epileptic epoch for each epileptic event;estimating source activity of the epileptic epoch of each of the plurality of epileptic events to generate an estimated epileptic epoch source activity for each of the plurality of epileptic events; and clustering the estimated epileptic epoch source activity for each of the plurality of epileptic events to generate at least one source spatialtemporal cluster for each of the plurality of epileptic events; wherein integrating the sensor data for each epileptic event to form the set of integrated multi-event sensor data comprises integrating the at least one source spatialtemporal cluster of each of the plurality of epileptic events.

34. The computer implemented method according to claim 28, wherein the sensor data for a subject for each of a plurality of epileptic events comprises invasive sensor data and wherein integrating the sensor data for each epileptic event to form the set of integrated multi-event sensor data comprises integrating the invasive sensor data of each of the plurality of epileptic events.

35. The computer implemented method according to claim 28, wherein the sensor data for a subject for each of a plurality of epileptic events comprises invasive sensor data, the method further comprising: pre-processing the invasive sensor data for each of the plurality of epileptic events; and wherein integrating the sensor data for each epileptic event to form the set of integrated multi-event sensor data comprises integrating the preprocessed invasive sensor data of each of the plurality of epileptic events.

36. The computer implemented method according to claim 28, wherein the sensor data for a subject for each of a plurality of epileptic events comprises invasive sensor data, the method further comprising: cropping the invasive sensor data for each of the plurality of epileptic events to form an estimated epileptic epoch for each epileptic event; andwherein integrating the sensor data for each epileptic event to form the set of integrated multi-event sensor data comprises integrating the estimated epileptic epoch of each of the plurality of epileptic events.

37. The computer implemented method according to claim 28, wherein the multi-event epileptogenic network with one node and zero edges and the multi-event epileptogenic network with two or more nodes and e edges are configured for use with at least one of surgical planning, surgical simulation, and surgical execution.

Images