A method to forecast multiple seizure types using seizure cycles

Abstract

A method comprising receiving a record of historical epileptic seizure events in a patient, including, for each recorded historical epileptic seizure event, seizure characterisation data characterising duration, symptom profile, and / or severity of the seizure event; dividing the recorded seizure events of the record of historical epileptic seizure events into a plurality of seizure sets based on the seizure characterisation data; for at least one seizure set individually, using the record of historical epileptic seizures to generate a set-specific temporal probability model, wherein the set-specific temporal probability model projects variability over time of likelihood of occurrence of a seizure event fitting the seizure characterisation data for the respective individual seizure set into a future time period.

Description

"A method to forecast multiple seizure types using seizure cycles"Technical Field

[0001] The present disclosure relates to methods and systems for seizure management. Some embodiments relate to forecasting multiple seizure types based on seizure characterisation data.Background

[0002] Worldwide, one in 100 people live with epilepsy. In the US alone, over 150,000 people are diagnosed every year. Over the last decade, researchers have identified patterns in seizures which follow predictable cycles. These cycles allow forecasting “seizure risk,” which is an indication of how likely a person is to have a seizure at a given time.

[0003] Many people with epilepsy experience one or more “types” of seizures, which have similar precursors, symptoms, loss of awareness, and other common features, and some people may experience multiple different types of seizures at different times. Typically, the impact that these seizures have on a person’s wellbeing (in terms of risk of injury, recovery time and general wellbeing) can vary, some are low impact while others may have significant impact. The frequency of the different seizure types may also vary.

[0004] Known existing forecasting solutions do not differentiate between seizure types, but instead produce an equally weighted forecast across all seizure types. This can reduce the relevance of forecasts, as it may increase the amount of time a person spends in a “high risk” state without forecasting the events that are most relevant to the person.

[0005] It is desired to address or ameliorate one or more shortcomings or disadvantages of prior methods of forecasting seizure types, such as the traditional strategy outlined above, or to at least provide a useful alternative thereto.

[0006] Throughout this specification the word "comprise", or variations such as "comprises" or "comprising", will be understood to imply the inclusion of a stated element, integer or step, or group of elements, integers or steps, but not the exclusion of any other element, integer or step, or group of elements, integers or steps.

[0007] Any discussion of documents, acts, materials, devices, articles or the like which has been included in the present specification is not to be taken as an admission that any or all of these matters form part of the prior art base or were common general knowledge in the field relevant to the present disclosure as it existed before the priority date of each of the appended claims.Summary

[0008] Some embodiments relate to a method, the method may comprise: receiving a record of historical epileptic seizure events in a patient, including, for each recorded historical epileptic seizure event, seizure timeline data representing date and time of the seizure, and seizure characterisation data characterising duration, symptom profile, and / or severity of the seizure event; dividing the recorded seizure events of the record of historical epileptic seizure events into a plurality of seizure sets defined by respective membership criteria, the membership criteria comprising, for the respective seizure set, one or more value or value ranges of the seizure characterisation data; for at least one seizure set individually, using the record of historical epileptic seizures to generate a setspecific temporal probability model, wherein the set-specific temporal probability model projects variability over time of likelihood of occurrence of a seizure event fitting the membership criteria for the respective individual seizure set into a future time period.

[0009] In generating the set-specific temporal probability model for each seizure set individually, only the recorded seizure events belonging to the respective individual seizure set may be used to generate the said set-specific temporal probability model.

[0010] In generating the set-specific temporal probability model for each seizure set individually, the recorded seizure events belonging to the respective individual seizureset and recorded seizures belonging to one or more other seizure sets among the plurality of distinct seizure sets may be used to generate the said set-specific temporal probability model.

[0011] The one or more value ranges of the seizure characterisation data forming the membership criteria for the seizure sets may be defined, at least in part, according to user preferences input to an application running on a personal computing device. The one or more value ranges of the seizure characterisation data forming the membership criteria for the seizure sets may be defined, at least in part, according to a clustering algorithm executed on the seizure characterisation data of the record of historical epileptic seizures.

[0012] The seizure characterisation data may be self-reported by the patient via a user interface of an application running on a personal computing device or via a dedicated seizure reporting device in data communication with an application running on a personal computing device. The seizure timeline data may be self-reported by the patient via a user interface of an application running on a personal computing device or via a dedicated seizure reporting device in data communication with an application running on a personal computing device.

[0013] The seizure characterisation data may be automatically generated by one or more of: a wearable monitoring device, an environmental monitoring device, and an implantable monitoring device. The seizure timeline data may be automatically generated by one or more of a wearable monitoring device and an implantable monitoring device.

[0014] The method as described above, may further comprise, at a user interface of an application running on a personal computing device, presenting the temporal probability model for a user-selected one of the plurality of distinct seizure sets to a user on a display screen. The presenting may include, in accordance with user preferences, highlighting time windows in which seizure-set-specific likelihood of seizure event occurrence exceeds a threshold, the threshold being either an absolute value or a defined amount or proportion above an average or other characteristic value.

[0015] The method as described above, may further comprise, weighting the temporal probability models for the plurality of seizure sets according to the one or more value ranges of the seizure characterisation data forming the membership criteria of the respective sets, and summing the weighted temporal probability models to obtain a weighted aggregate temporal probability model representing weighted aggregate likelihood of seizure event occurrence for the plurality of seizure sets collectively for the future period of time.

[0016] The method as described above, may further comprise, at a user interface of an application running on a personal computing device, presenting the weighted aggregate temporal probability model on a display screen.

[0017] The method as described above, may further include, highlighting future time windows in which weighted aggregate likelihood of seizure occurrence for the plurality of seizure sets collectively exceeds a predefined threshold.

[0018] Some embodiments relate to an apparatus, the apparatus may comprise memory hardware and processor hardware, the memory hardware storing processing instructions which, when executed by the processor hardware, cause the apparatus to perform the method as described above.

[0019] Some embodiments relate to a computer program which, when executed by processor hardware of an apparatus comprising memory hardware and processor hardware, may cause the apparatus to perform the method as described above.

[0020] Some embodiments relate to a non-transitory computer-readable medium storing the computer program as described above.

[0021] Embodiments may include a method comprising receiving a record of historical epileptic seizure events in a patient, including, for each recorded historical epileptic seizure event, seizure characterisation data characterising duration, symptom profile, and / or severity of the seizure event; dividing the recorded seizure events of the record ofhistorical epileptic seizure events into a plurality of seizure sets based on the seizure characterisation data; for at least one seizure set individually, using the record of historical epileptic seizures to generate a set- specific temporal probability model, wherein the set-specific temporal probability model projects variability over time of likelihood of occurrence of a seizure event fitting the seizure characterisation data for the respective individual seizure set into a future time period.Brief Description of Drawings

[0022] Embodiments of the present disclosure will now be described by way of nonlimiting example only with reference to the accompanying drawings, in which:

[0023] Figure 1 illustrates a method;

[0024] Figure 2A is a schematic diagram illustrating methods for forecasting multiple seizure types based on user preferences, according to some embodiments;

[0025] Figure 2B is a schematic diagram illustrating methods for evaluating seizure data for generating future seizure activity in a subject having epilepsy, according to some embodiments;

[0026] Figure 2C is a schematic diagram highlighting different data dependencies in different embodiments;

[0027] Figure 3 is a schematic illustration of a seizure management system, according to some embodiments;

[0028] Figure 4 is a schematic illustration of a device configured to generate alerts and / or notifications based on subject-specific temporal probabilistic model(s), according to some embodiments;

[0029] Figure 5 is a schematic illustration of a seizure management system, according to some embodiments;

[0030] Figure 6 is a process flow diagram of a method for projecting variability over time of likelihood of occurrence of a seizure in a subject having epilepsy across a future time period, according to some embodiments;

[0031] Figure 7 is a graphical example of projecting variability over time of likelihood of occurrence of a seizure in a subject having epilepsy across a future time period, according to some embodiments; and

[0032] Figure 8A and 8B are an example forecast visualisation, according to some embodiments.Description of Embodiments

[0033] Embodiments of the present disclosure relate to methods and systems for seizure management. Some embodiments relate to forecasting multiple seizure types based on seizure characterisation data.

[0034] PCT Application No. PCT / AU2018 / 050575 filed 08 June 2018 and titled “Methods and Systems for Forecasting Seizures”, the contents of which are hereby incorporated by reference, describes using historical seizure data to determine or predict a probability or likelihood of seizure in a subject over a plurality of time windows. Described embodiments may include seizure likelihood measures determined according to one or more of the methods disclosed in PCT / AU2018 / 050575 in determining a likelihood of seizure in a subject over a plurality of time windows and using the determined likelihood to determine suitable time(s) for subjecting the patient or subject to an intervention, and / or determining the impact of an intervention on the occurrence of epileptic seizures.

[0035] PCT Application No. PCT / AU2021 / 051118 filed 23 September 2021 and titled “Methods and Systems for Forecasting Epileptic Events”, the contents of which are hereby incorporated by reference, describes using historical seizure data to determine or predict a probability or likelihood of seizure in a subject over a plurality of time windows.Described embodiments may include seizure likelihood measures determined according to one or more of the methods disclosed in PCT / AU2021 / 051118 in determining a likelihood of seizure in a subject over a plurality of time windows and using the determined likelihood to determine suitable time(s) for subjecting the patient or subject to an intervention, and / or determining the impact of an intervention on the occurrence of epileptic seizures.

[0036] Application

[0037] Figure 1 illustrates a method 100. The method 100 may be implemented by computing apparatus. For example, the method 100 may be implemented by executable software running on a user device such as a personal computer or smartphone belonging to a clinician, a patient, or another user with a personal or professional relationship to the patient. The method 100 may be implemented by server-side software operable by a patient or clinician via a client-side interface. The method 100 may be implemented by executable software running on a user device such as a personal computer or smartphone in cooperation with server- side software running on a server to which the user device is connected for exchanging data. The server- side software may be running on a cloudbased server to which the user device is connected for exchanging data.

[0038] At S 101, the method 100 includes receiving a record of historical epileptic seizure events in a patient. The record includes, for each recorded historical epileptic seizure event, seizure timeline data representing date and time of the seizure, and seizure characterisation data characterising duration, symptom profile, and / or severity of the seizure event. The record may be received from a local or remote data store from which the computing apparatus executing the method 100 is configured to obtain patient data, such as a local data storage unit. Alternatively, the record may be received from a local or remote data store from which the computing apparatus executing the method 100 is configured to obtain patient data, such as a data storage server or a cloud-based data storage server. Receiving the record indicates the record being obtained, copied, or otherwise accessed for the purpose of performing the method 100.

[0039] The nature of the seizure timeline data and the seizure characterisation data is dependent upon the implementation scenario and may vary between patients and users. Taking the example of seizure characterisation data, a patient may experience epileptic seizure events that are consistent with respect to symptom profile and perceived severity, but which vary significantly with respect to duration. In this example, the seizure characterisation data may only relate to duration, since the variability in duration across seizure events will enable dividing at S102. Alternatively, a patient or clinician may experience epileptic seizure events which vary in duration, symptom profile, and severity, but may wish to omit one or more of those factors from the seizure characterisation data in order to simplify analysis and understanding of the data, possibly with a view to incorporating the additional factors later.

[0040] Symptom profile may be represented by a list of symptoms experienced by the patient in the pertinent epileptic seizure event. Alternatively, each of a list of potential symptoms may be assigned a value from a range such as 0-5 or 0-10 indicating severity with which that particular symptom was experienced. The assigned value may be assigned to a record of the epileptic seizure event by the patient after occurrence. For example, the epileptic seizure event may be recorded automatically or manually at the time it is occurring, with a user then augmenting the record later with the symptom profile. Alternatively, the epileptic seizure event may be recorded retrospectively by the patient or another user recording timeline data and seizure characterisation data representing time, date, symptom profile, duration, and / or severity of the seizure event after it has occurred.

[0041] Severity may be represented by a single value from within a range such as 0-5 or 0-10. Alternatively, severity may be multi-dimensional, for example with one severity value per symptom, in which case symptom profile data indicating a severity per symptom may itself indicate severity, or may be processed to generate a value of a metric representing severity such as an average or weighted average of the symptom-specific severity values.

[0042] At S102 the historical epileptic seizure events are divided into a plurality of seizure sets. Each seizure set is defined by respective membership criteria. The membership criteria comprise, for the respective seizure set, one or more value or value ranges of the seizure characterisation data. For example, the membership criteria may be non-overlapping so that each seizure event belongs to a maximum of one seizure sets. Alternatively, the membership criteria may be overlapping so that in some instances a single epileptic seizure event may belong to more than one set. The membership criteria may be determined automatically, such as by a clustering algorithm. The membership criteria may be determined manually by a user (which may be a patient, a clinician, or another user). Alternatively, the software implementing the method 100 may enable a user to select from one or more predefined configurations of membership criteria (which may be manually configurable). For example, there may be a membership criteria configuration that is particularly effective for patients who suffer epileptic seizure events that are relatively consistent in terms of symptom profile or severity but vary significantly in terms of duration. Or vice versa. Each seizure set contains a subset of the population of historical seizure events in the received record.

[0043] Epileptic seizure events include seizures and auras.

[0044] The seizure sets are a data artefact representing the record of historical epileptic seizure events received at S101 for the purpose of performing analysis at S103. Thus, they may be stored on short term memory and then re-written after completion of SI 03, or they may be stored on long term (i.e. non-volatile) data storage for further analysis.

[0045] The analysis at S 103 includes, for at least one seizure set individually, using the record of historical epileptic seizures to generate a set-specific temporal probability model, wherein the set-specific temporal probability model projects variability over time of likelihood of occurrence of a seizure event fitting the membership criteria for the respective individual seizure set into a future time period. The data processed at S 103 to generate the set- specific temporal probability model may or may not be limited to the seizure set being analysed. For example, to generate a set-specific temporal probability model for future seizure events having a defined symptom profile (defined bymembership criteria of the pertinent seizure set) the confidence level in the prediction may be enhanced by ignoring seizure events not having the defined symptom profile (i.e. because they add noise to the analysis) or may be enhanced by including seizure events not having the defined symptom profile (perhaps a seizure event having symptom profile A tends to occur 16-24 hours after a seizure event having symptom profile B).

[0046] The function of SI 03 is to provide a probability model over time of likelihood of occurrence of a seizure event fitting the membership criteria of the pertinent set, that is, having seizure characterisation data representing duration, symptom profile, and / or severity that would cause the seizure event to be assigned to the pertinent seizure set. The processing may be repeated for each seizure set or may be isolated to one or more seizure sets according to a selection by the user interacting with the software executing the method 100. Individual seizure sets fitting different membership criteria each have separate cyclical seizure patterns that can be forecast using the set-specific temporal probability models. The function of dividing the historical epileptic seizure events into a plurality of seizure sets and then performing the analysis of S 103 is to generate separate forecasts for each of the seizure sets. Generating separate forecasts may assist in understanding the impact that each of the individual seizure sets fitting different membership criteria has on the patient, for example. The process for generating temporal probability models from historical epileptic seizure event data at S103 is set out in more detail below with reference to Figures 2A onwards. Data dependencies between setspecific seizure event data and temporal probability models are illustrated in Figure 2C, which is discussed in more detail below.

[0047] An intervention may be a passive intervention, such as an event or activity that the subject is exposed to. The intervention may be an active intervention such as the administering to the subject of medication and / or stimulation.

[0048] An intervention may be a trigger, such as an event or an action, that causes a specific subject to be more likely to have seizures, for instance caffeine or alcohol consumption, lack of sleep, flashing lights, loud noises or certain foods. The trigger may be a relatively normal or commonplace activity, such as the subject brushing their teeth.An intervention may comprise one or more actions that are suspected triggers contributing to seizure occurrence in a specific subject.

[0049] An intervention may be a treatment, or seizure treatment, such as administration of medication, administration of electrical stimulation, administration of magnetic stimulation and / or life style / behavioural change. For example, a period of relatively low seizure likelihood may be a suitable time for changing a medication, varying dosage regimes, stopping a medication, or identifying potential triggers. A period of relatively high seizure likelihood may be a suitable time for scheduling seizure monitoring sessions, such as EEG or video-EEG monitoring.

[0050] Some embodiments provide a probabilistic approach to forecasting seizure likelihood in subjects with epilepsy that incorporates prior knowledge about underlying patterns in seizure occurrence with respect to other variables that affect the probability of a subject or patient having a seizure.

[0051] Multiday cycles of cortical excitability have been predominantly investigated in people with epilepsy. As set out in the described embodiments, it has been established that, analogous to circadian rhythms, free-running multiday rhythms are widespread across the major organ systems and that non-electroencephalography (EEG) physiological data (i.e., physiological data that is not derived from a brain of the subject) as well as EEG physiological data correlates with seizure occurrence. For example, suitable non-EEG physiological data may include cardiac data, such as heart rate and / or heart rate variability, sleep data, oxygen saturation data, electrodermal activity data, respiratory data, and / or activity data. It has been established that cardiac output shows multiday rhythms akin to cycles of cortical excitability. Furthermore, there can be differing seizure risk factors (for example, increased heart rate, decreased heart rate variability, sleep deprivation etc.) between people with epilepsy.

[0052] Figure 2A is a schematic diagram illustrating a system 200A for forecasting multiple seizure types based on user preferences. The system of Figure 2A may be configured to perform the method illustrated in Figure 1.

[0053] The system 200A includes user data 102 associated with a subject with epilepsy. User data 102 is recorded over a period of time and may be stored. The user data 102 is one or more data items specific to a user of a system, such as system 200A, which the user may interact with via application software, or some other program, running on a personal computing device such as a smartphone, tablet, laptop computer, desktop computer, smartwatch. The user data 102 may be stored locally and / or stored on a remote server apparatus accessible via the application software or other program. The remote server apparatus may be a cloud-based server. User data 102 may be stored securely so that it is only accessible to the user, who may be authenticated via the application or other program. Optionally, the system 200A may facilitate third party access to user data 102 conditional upon approval by the user. For example, the system 200A may provide a mechanism for a user to grant access to their user data 102 to a clinician. User data 102 may comprise physiological data 102A (EEG and / or non-EEG physiological data), seizure activity data 102B, environmental data 102C, and / or potential trigger data 102D. The seizure activity data 102B comprises seizure timeline data. For example, a time of day and date at which each seizure or seizures occurred during the time period. User data 102 comprises historical epileptic seizure data of a subject with epilepsy.

[0054] The seizure activity data 102B is determined or captured in any suitable manner. For example, the seizure activity data 102B may be determined from data recorded by the patient, for example, keeping a seizure occurrence diary and / or from clinical records. Seizure activity data 102B, such as seizure characterisation data and / or seizure timeline data, may be self-reported by a subject with epilepsy via a user interface of an application running on a personal computing device. For example, seizure activity data 102B, such as seizure characterisation data and / or seizure timeline data, may be self-reported by a subject with epilepsy via a dedicated seizure reporting device in data communication with an application running on a personal computing device.

[0055] The seizure activity data 102B may be derived from physiological signals, which may be captured automatically, and / or may be detected from electrodes, or a wearable device, or an implantable device. For example, seizure activity data 102B may be determined or captured by analysing EEG data indicative of brain activity of thepatient. That is, seizure characterisation data and / or seizure timeline data of seizure activity data 102B may be captured automatically, and / or may be detected from electrodes, or a wearable device, or an implantable device, for example.

[0056] Seizure activity data 102B further comprises seizure characterisation data. Seizure characterisation data characterises a plurality of aspects of each seizure experienced by the patient. Seizure characterisation data may include one or more from among the following characterisations: duration of the seizure, severity of the seizure, symptoms associated with the seizure, awareness of the patient during the seizure, and a seizure type associated with the seizure. Duration of the seizure may include a length of time in seconds or minutes, for example. Severity of the seizure may include a label, such as mild, moderate, and severe, for example. Symptoms associated with the seizure may include, but are not limited to, any one of: cognitive symptoms, emotional or affective symptoms, autonomic symptoms, automatism symptoms, motor symptoms, sensory symptoms, or laterality symptoms, for example. Awareness of the patient during the seizure may include a label, such as aware or impaired awareness, for example. A seizure type associated with the seizure may include a label, such as focal onset, generalised onset, or unknown onset, for example.

[0057] Each seizure type may further include a sub-seizure type label. Focal onset seizure types may be further classified as aware seizures, impaired awareness seizures, motor onset seizures, nonmotor onset seizures, or focal to bilateral tonic-clonic seizures. Focal onset motor onset seizures may be further classified as automatisms, atonic, clonic, epileptic spasms, hyperkinetic, myoclonic, or tonic. Focal onset nonmotor onset seizures may be further classified as autonomic, behaviour arrest, cognitive, emotional, or sensory. Generalised onset seizure types may be further classified as motor seizures or nonmotor (absence) seizures. Generalised onset motor seizures may be further classified as tonic-clonic, clonic, tonic, myoclonic, myoclonic-tonic-clonic, myoclonic-atonic, atonic, or epileptic spasms. Generalised nonmotor seizures may be further classified as typical, atypical, myoclonic, or eyelid myoclonia. Unknown onset seizure types may be further classified as motor seizures, nonmotor seizures or unclassified seizures. Unknownonset motor seizures may be further classified as tonic-clonic or epileptic spasms. Unknown onset nonmotor seizures may be further classified as behaviour arrest.

[0058] Cognitive symptoms may include, but is not limited to, any one of: acalculia, aphasia, attention impairment, deja vu, dissociation, dysphasia, hallucinations, illusions, memory impairment, neglect, forced thinking, or responsiveness impairment. Emotional or affective symptoms may include, but is not limited to, any one of: agitation, anger, anxiety, crying (dacrystic), fear, laughing (gelastic), paranoia, or pleasure. Autonomic symptoms may include, but is not limited to, any one of: asystole, bradycardia, erection, flushing, gastrointestinal, hyper / hypoventilation, nausea, vomiting, pallor, palpitations, piloerection, respiratory changes, or tachycardia. Automatism symptoms may include, but is not limited to, any one of: aggression, eye-blinking, head-nodding, manual, oralfacial, pedaling, pelvic thrusting, perseveration, running (cursive), sexual, undressing, vocalisation / speech, or walking. Motor symptoms may include, but is not limited to, any one of: dysarthria, dystonic, fencer’s posture, incoordination, Jacksonian, paralysis, paresis, or versive. Sensory symptoms may include, but is not limited to, any one of: auditory, gustatory, hot-cold sensations, olfactory, somatosensory, vestibular, or visual. Laterality symptoms may include, but is not limited to, any one of: left, right, or bilateral.

[0059] Environmental data 102C may comprise one or more of temperature, humidity, wind speed, barometric pressure, rainfall. Environmental data 102C may comprise values, ranges of values and / or changes in values or variations. Environmental data 102C may be used to weight probabilistic models generated using physiological data 102A associated with a subject. For example, a correlation has been found between weather conditions and the likelihood of seizures in epilepsy sufferers. Environmental data 102C may be automatically captured from the environment that the specific subject is located. For example, a wearable device of the specific subject may automatically capture environmental data 102C, and the captured data used to determine whether a trigger event has occurred based on the environmental data 102C. For example, the wearable device may automatically record loud noises or flashing lights in the environment and determine that these indicate a trigger event. For example, the wearable device may automaticallyrecord a sudden change in temperature in the environment and determines that this indicates a trigger event.

[0060] The potential trigger data 102D comprises subject- specific potential or suspected individual trigger data. For example, potential trigger data 102D may comprise trigger data indicating a factor that may have potentially caused a seizure. The potential trigger data 102D is determined or captured in any suitable manner. For example, the potential trigger data 102D may be determined from data recorded by the patient. The potential trigger data 102D may be determined from data recorded from the environment such as environmental data 102C. The potential trigger data 102D may be inferred or directly computed from physiological signals, which may be detected from electrodes, or a wearable device, or an implantable device. For example, potential trigger data 102D may be determined or captured by analysing EEG data indicative of brain activity of the patient.

[0061] The patient may record potential trigger data 102D information relating to their diet such as, food consumed, amount of food consumed, and time of consumption, for example. The patient may record potential trigger data 102D information relating to their exercise such as, exercise type, exercise duration, and exercise intensity, for example. The patient may record potential trigger data 102D information relating to their performed activities such as, activity type, activity duration, and activity intensity, for example. The patient may record potential trigger data 102D by logging information via an application, for example.

[0062] The system 200A further illustrates user preferences 104 associated with a subject with epilepsy. The user preferences 104 may be stored locally and / or stored on a remote server apparatus accessible via the application software or other program. The remote server apparatus may be a cloud-based server. User preferences 104 may be associated with a user other than the subject with epilepsy, such as, a clinician, carer or a doctor, for example. User preferences 104 comprises at least one membership criteria indicating at least one value range, or identifier, of the seizure characterisation data of seizure activity data 102B. In this way, the user preferences 104 define or contribute toa definition of the different types of seizure event. Membership criteria may include one or more value ranges, or identifiers, associated with the seizure characterisation data.

[0063] One or more value ranges, or identifiers, associated with the seizure characterisation data forming the membership criteria are defined, at least in part, by user preferences 104 input to an application running on a personal computing device, such as device 400 (Figure 4), for example. This provides the user an element of control over the categorisations of seizure events into bins per type.

[0064] One or more value ranges, or identifiers, associated with the seizure characterisation data forming the membership criteria are defined, at least in part, according to a clustering algorithm executed on the seizure characterisation data of seizure activity data 102B. For example, any one of: density-based clustering algorithm, a distribution-based clustering algorithm, a centroid-based clustering algorithm, or an hierarchical-based clustering algorithm may be executed on the seizure characterisation data to define, at least in part, the membership criteria. Clustering algorithms provide a mechanism to automatically determine boundaries between bins into which seizure events are categorised. Optionally, the value ranges or identifiers generated by the clustering algorithm may be presented to a user (via a graphical user interface or otherwise) as suggestions, enabling the user to implement the suggestions as is, to modify the suggestions and implement the modified version, or to reject the suggestions.

[0065] It is noted that the term ‘value range’ does not necessarily mean that a value range is defined by two values, one at either extreme. A value range may be defined by two values, one at either extreme, or value range may be defined by a single value also implicitly defines a value range.

[0066] At module 106, “split data by seizure types”, user preferences 104 are used to divide the recorded seizures of user data 102 into a plurality of seizure sets defined by respective membership criteria. Module 106 is a function of an apparatus implementing the system 200A, such as a computing apparatus. The functionality of module 106 may be a step in a method, process, or algorithm. The functionality of module 106 may beimplemented at the server-side or at the client-side by the application or other program. User preferences 104 comprises a plurality of options, such as, duration of a seizure, type of seizure, symptoms of the seizure, and severity of the seizure, allowing a user to select or alter seizure forecasts provided to them. For example, recorded seizures of user data 102 may be divided by severity of the seizure into a plurality of seizure sets, being a mild seizure severity set, a moderate seizure severity set, and a severe seizure severity set. More than one membership criteria may be used to divide the recorded seizures of user data 102 into a plurality of seizure sets. For example, membership criteria may include both severity of the seizure and duration of the seizure, where the recorded seizures of user data 102 are divided into a plurality of seizure sets defined by the severity, mild, moderate, and severe, and further divided by duration, such as, less than 20 seconds and more than 20 seconds. That is, the membership criteria may be multi-dimensional insofar as they define boundaries between categorisations based on multiple variables, parameters, or characterisations.

[0067] A portion or subset of the recorded seizure events of user data 102 may be divided into a plurality of seizure sets defined by respective membership criteria. For example, user preferences 104 may indicate that the user wants to forecast seizures with a particular symptom, such as, hyperventilation or palpitations, resulting in a seizure set only including recorded seizures with those symptoms. In other words, the user may apply filtering before the seizure events are processed at functional module 106. Filtering may be temporal, for example seizure events occurring more than 12 months ago may be filtered out.

[0068] The system 200A further includes a plurality of forecast models, forecast model type A 108, forecast model type B 110, and forecast model type C...n 112, where n indicates any number or forecast models dependent on the number of seizure sets defined by respective membership criteria. Each forecast model is associated with a respective seizure set (i.e. a particular type or categorisation of seizure events into which seizure events are split by functional module 106) represented by data having a structure and values such that the model represents a set-specific temporal probability model, wherein the set-specific temporal probability model projects variability over time of likelihood ofoccurrence of a seizure event fitting the membership criteria for the respective individual seizure set into a future time period. The system 200A is configured to generate the forecast models, for example via a forecasting algorithm or equivalent functional module. The forecasting algorithm may be stored and executed on the user device running the application or other program, or the forecasting algorithm may be stored and executed remotely at the server-side using user data transmitted thereto by functional module 106 and hardware such as an I / O interface. For each seizure set defined by respective membership criteria, a set-specific temporal probability model, or forecast model, is generated, for example. For example, where the historical epileptic seizure data of user data 102 is divided into 10 seizure sets defined by respective membership criteria, 10 respective forecast models type A to type J are generated by the forecasting algorithm. Each of the forecast models 108, 110, 112 are generated by a forecasting algorithm using historical epileptic seizure data of user data 102, to be described in detail below. The setspecific temporal probability model projects variability over time of likelihood of occurrence of a seizure fitting the membership criteria for the respective individual seizure set into a future time period.

[0069] Forecast models 108, 110, 112 are generated by the forecasting algorithm based on historical epileptic seizure data of user data 102. That is, forecast model type A 108 may be generated using seizure data associated with any seizure irrespective of membership criteria, for example. For example, forecast model type B 108, defined by seizures having a moderate severity, may be forecast by the forecast algorithm using both seizures with and without a moderate severity. Since it may be the case that in order to predict likelihood of occurrence of a seizure event fitting the criteria for a first seizure set, it is relevant (or necessary or beneficial) to consider historical seizure events fitting the criteria for a second seizure set. In a simplified example, it may be that the historical seizure activity data 102B indicates that a patient is at heightened risk of experiencing a high severity seizure for 72 hours proceeding a moderate severity seizure.

[0070] Optionally, any individual forecast model 108, 110, 112 may be generated by the forecasting algorithm using historical epileptic seizure data of user data 102 associated with respective membership criteria. For example, forecast model type A 108may be generated using only seizure data associated with mild seizures and forecast model type B 110 may be generated using only seizure data associated with seizures of a duration greater than 10 seconds. For example, forecast model type A 108 may be generated using only seizure data associated with severe seizures and forecast model type B 110 may be generated using all historical epileptic seizure data of user data 102.

[0071] Any individual forecast model 108, 110, 112 may be generated by the forecasting algorithm using historical epileptic seizure data of user data 102 not associated with respective membership criteria. For example, forecast model type A 108, defined by membership criteria including severe seizure severity and durations of more than 5 seconds, may be forecast using seizure data not of a severe seizure severity and of a duration less than 5 seconds.

[0072] Any individual forecast model 108, 110, 112 may be generated by the forecasting algorithm using historical epileptic seizure data of user data 102 associated with respective membership criteria and additionally using historical epileptic seizure data of user data 102 not associated with respective membership criteria. For example, forecast model type C 112, defined by membership criteria including seizure durations of more than 6 seconds, may be forecast using seizure data of seizures with a duration of more than 6 seconds and seizure data of seizures with a duration of less than 6 seconds.

[0073] The system 200A may further include an aggregate forecast 114, wherein the aggregate forecast 114 is the generation of a temporal probability model for any seizure events rather than those belonging to a specific set of membership criteria, that is, rather than being limited to specific membership criteria such as in seizure set forecast models 108, 110, 112. That is, the aggregate forecast 114 does not take into account user preferences 104 when generating the aggregate temporal probability model, for example.

[0074] The system 200A may further provide forecast visualisation 116, wherein forecast visualisation 116 is a visual presentation of the generated temporal probability models to a user. Each forecast model 108, 110, 112, 114 may be individually presented to the user. That is, for each forecast model generated by the forecasting algorithm, anindividual visualisation is presented to the user, for example. Two or more forecast models may be combined to generate a derived forecast model to be visually presented to the user. Derived forecasts may include a weighted combination of two or more forecast models previously generated. The two or more forecast models may be combined using statistical methods, such as Bayesian statistics, to generate the derived forecast.

[0075] The user preferences 104 may further include user specified weights for generating derived models. User specified weights allow a user to combine different seizure forecast models by weighting the different model types to their own preferences. Combining different forecast models, relating to different seizure types, as specified weights unique to the user may allow the resulting derived forecast model to account for the different impact the underlying types of each forecast model has on the user’s life. Combining the derived forecast models may be executed by generating the individual forecast models each relating to distinct seizure types (i.e. distinct seizure sets). Alternatively, the raw data describing the seizure events themselves may be combined (so that there are seizure events from plural distinct seizure sets) and the forecast model generated from the seizure event data from the plural distinct seizure sets.

[0076] Referring to Figure 2B, there is shown a schematic diagram illustrating methods 200B for evaluating seizure data for generating future seizure activity in a subject having epilepsy. Figure 2B adopts a convention illustrating data artefacts in boxes with square comers and method, process, or algorithm steps in boxes with rounded corners. Arrows flowing to a box represent data from or generated by the data artefact or method step at the origin of the arrow being an input to the data artefact or method step at the termination of the arrow. Illustrated arrows are not comprehensive but illustrate the key data exchanges for illustrative purposes. Figure 2B is exemplary of the method of Figure 1.

[0077] Figure 2B illustrates a split dataset 202 (a seizure set specific dataset) associated with a subject with epilepsy. The split dataset 202 is representative of each of the plurality of seizure sets resulting from splitting the recorded seizures of user data 102 by respective membership criteria at 106. That is, for each forecast model 108, 110, 112, 114, anindividual split dataset 202 is generated, for example. Split dataset 202 comprises user data 102, which may be a subset or at least a portion of user data 102.

[0078] Split dataset 202 comprises physiological data 202A (EEG and / or non-EEG physiological data), seizure activity data 202B, and / or environmental data 202C. Physiological data 202A comprises seizure data of physiological data 102A for seizures meeting the respective membership criteria of the corresponding individual forecast model 108, 110, 112, 114. Seizure activity data 202B comprises seizure data of seizure activity data 102B for seizures meeting the respective membership criteria of the corresponding individual forecast model 108, 110, 112, 114. Environmental data 202C comprises seizure data of environmental data 102C for seizures meeting the respective membership criteria of the corresponding individual forecast model 108, 110, 112, 114.

[0079] A subject-specific seizure likelihood 206A is determined for a first period of time. Subject- specific seizure likelihood 206A may be determined using a subjectspecific temporal probabilistic model 204 generated according to one or more of the methods disclosed in PCT / AU2018 / 050575 and / or in PCT / AU2021 / 051118. The one or more temporal probabilistic models 204 may be generated according to the method shown in Figure 1 of PCT / AU2018 / 050575, as described between paragraphs

[0092] and

[0132] . The one or more temporal probabilistic models 204 could be generated according to the method shown in Figures 1 and 4 of PCT / AU2021 / 051118, as described between paragraphs

[0036] and

[0040] , and

[0065] and

[0078] .

[0080] The method of estimating the probability or likelihood of a seizure event of a specific type in a subject illustrated in Figure 2B comprises at 208 determining or receiving split dataset 202 associated with epileptic events experienced by the subject over a first time period. The split dataset 202 of the example of Figure 2 includes physiological data 202A associated with each epileptic event and a time at which each epileptic event occurred. A temporal probability model of future epileptic events based on the time of each of the epileptic events may be generated. The temporal probability model 204 represents a probability of future seizure occurrence in each of a plurality of future time windows. A probabilistic model may be generated based on the physiologicaldata 202A associated with each epileptic event. The probabilistic model based on the physiological data 202A and the temporal probability model may be combined to generate a weighted probabilistic model representing a probability of future seizure activity in each of the plurality of future time windows weighted according to time (e.g. where the time window sits in a cycle or rhythm) and according to physiological parameters. The weighted probabilistic model(s) may be used as subject- specific seizure likelihood model(s) and may be configured to output estimate(s) of seizure probability in the subject for one or more of the plurality of future time windows according to values of one or more physiological parameters.

[0081] A method of estimating the probability of a seizure of a specific type in a subject may comprise determining or receiving split dataset 202 associated with epileptic events experienced by the subject over a first time period. The split dataset 202 may comprise a time at which each epileptic event occurred. A combined temporal probability model of future epileptic events may be generated based on the time of each of the epileptic events. The combined temporal probability model may represent a time-weighted probability of a future seizure occurrence in each of a plurality of time windows. The combined temporal probability model may be generated by combining two or more probability density functions, each probability density function derived over a different time period during the first time period. The combined temporal probability model(s) may be used as subject- specific seizure likelihood model(s) and may be configured to output estimate(s) of seizure probability in the subject for one or more of the plurality of future time windows.

[0082] A method of estimating the probability of a seizure of specific type in a subject may comprise determining or receiving split dataset 202 associated with epileptic events experienced by the subject over a first time period. The split dataset 202 may comprise physiological data 202A associated with each epileptic event and environmental data 202C associated with each epileptic event. Such a method comprises generating an environmental probability model of future epileptic events based on the environmental data 202C associated with each epileptic event, the environmental probability model representing a probability of a future seizure occurrence for each of a plurality of valuesof the environmental data 202C. The method comprises generating a probabilistic model based on the physiological data 202A associated with each epileptic event, and combining the probabilistic model based on the physiological data and the environmental probability model to generate a combined probabilistic model representing probability of future seizure activity for each of the plurality of values of the environmental data 202C in one or more future time windows. Optionally weighting may be applied to configure the relative contributions of the environmental probability model and the probabilistic model based on the physiological data. The weighted probabilistic model may be used as subject- specific seizure likelihood model(s) and may be configured to output estimate(s) of seizure probability in the subject for one or more of the plurality of values of the environmental variable for one or more future time windows.

[0083] A method of estimating the probability of a seizure of a specific type in a subject may comprise determining or receiving split dataset 202 associated with a subject experiencing epileptic events over a time period. The split dataset 202 may comprise non-EEG physiological data 202A recorded over the time period, and a time at which epileptic events occurred during the time period. One or more temporal models indicative of a subject-specific cycle may be extracted from the non-EEG physiological data 202A. One or more temporal probabilistic models 204 may be generated based on the respective one or more temporal models, the non-EEG physiological data 202A, and the times at which epileptic event occurred. Each temporal probabilistic model 204 may be representative of a probability of future seizure activity in each of a plurality of time windows. For example, the one or more temporal probabilistic models 204 may be generated by decomposing the non-EEG physiological data 202A into one or more component cycles corresponding to the one or more temporal models to produce one or more respective decomposed non-EEG physiological data, determining a phase of each of the one or more decomposed non-EEG physiological data, and mapping the times of the epileptic events to the phase of each of the one or more decomposed non-EEG physiological data.

[0084] The temporal probabilistic model(s) 204 may be used as subject-specific seizure likelihood or probability model(s) and may be configured to output estimate(s) of seizureprobability in the subject for one or more of the plurality of time windows, 206A, 206B. The estimate(s) of seizure probability in the subject may be a subject-specific first seizure likelihood 206A, or probability, or a measure of seizure activity of the subject across a first time period, wherein the first time period is indicative of a historical time period or past time period. The estimate(s) of seizure probability in the subject may be a subjectspecific second seizure likelihood or probability 206B of seizure activity of the subject across a second time period, wherein the second time period is indicative of a future time period, or, in some embodiments, a historical time period.

[0085] The seizure likelihood may be a retrospective assessment of the seizure risk probability during a time period, being a seizure likelihood measure, or may be a forward looking, forecast or prediction of the seizure risk probability during a time period, being a seizure likelihood forecast. In embodiments where the second seizure likelihood 206B is determined prior to the actual occurrence of the second time period, the second seizure likelihood 206B may be a forecast or prediction of the probability of seizures occurring during the second time period. That is, the second seizure likelihood 206B may be referred to as a seizure likelihood forecast or prediction, for example.

[0086] In the example of Figure 2B, a first seizure likelihood 206A is determined for the first period, being a historical period that has already taken place, and a second seizure likelihood 206B is determined for the second period, being a future period that has yet to take place.

[0087] The seizure likelihood forecast may be beneficial for use in a forecasting algorithm utilising seizure event data to forecast seizure likelihood, as using the seizure likelihood measure affected by interventions administered during the time period would have an adverse effect on the forecasted result. That is, the seizure likelihood forecast may be preferred where interventions or seizure events are known or expected to influence the seizure likelihood. The seizure likelihood measure may be more accurate in cases where the intervention(s) or seizure events are not known or expected to influence the seizure likelihood. That is, the seizure likelihood measure may be preferred where interventions or seizure events are known or expected to not influence the seizurelikelihood. The seizure likelihood measure may be beneficial for taking into account internal body cycles, such as a biological cycle or hormone cycle, for example, through the use of recorded data, such as heart rate data, for example.

[0088] The forecasted second seizure likelihood 206B utilises historical data, such as user data 102, from before the second time period to determine an average seizure likelihood over the second time period. The measured second seizure likelihood 206B may utilise recorded data, such as user data 102, over the second time period to determine an actual average seizure likelihood over the second time period. The first seizure likelihood measure 206A may be any statistical measure of the chance / likelihood / probability of a seizure occurring. The second seizure likelihood 206B may be any statistical measure of the chance / likelihood / probability of a seizure occurring. The second seizure likelihood 206B may be any statistical forecast of the chance / likelihood / probability of a seizure occurring. The second seizure risk probability 206B may be calculated based at least in part on the first seizure likelihood measure 206A.

[0089] A seizure probability may be determined based on seizure likelihood using an inhomogeneous poison point process. Confidence intervals of the inhomogeneous poison point process may then be used to determine statistical significance of interventions, for example. The seizure probability may be determined based on a forecasting algorithm.

[0090] A subject-specific actual seizure rate 208 over the first time period may be derived from user data 102, for example, over the first time period. The actual seizure rate 208 of the specific subject may be determined based on prior observations of seizure occurrences in user data 102. The actual seizure rate 208 may be determined based, at least in part, on data prior to the first time period. The actual seizure rate 208 may be normalised. For example, the actual seizure rate 208 may be normalised on a scale of 0 to 1. The actual seizure rate 208 or the normalised actual seizure rate may be divided by the seizure risk likelihood 206A to generate a subject- specific risk-adjusted actual seizure rate 210 over the first time period.

[0091] The subject-specific risk-adjusted actual seizure rate 210 may then be adjusted by the subject- specific second seizure likelihood 206B to determine a subject- specific second risk-adjusted seizure rate 212 over the second time period. The subject-specific risk-adjusted seizure rate 212 may be an expected seizure rate 212 of the subject over the second time period. In embodiments where the second seizure likelihood 206B is a forecast, the risk-adjusted seizure rate 212 is an expected risk-adjusted seizure rate 212 over the second time period. In embodiments where the second seizure likelihood 206B is a measure, the risk adjusted seizure rate 212 is a measured risk-adjusted seizure rate 212 over the second time period.

[0092] The expected risk-adjusted seizure rate 212 may be used to generate future seizure activity 214. That is, the expected risk-adjusted seizure rate 212 may be projected into the future to forecast seizure activity for the subject, for example.

[0093] An example seizure management system 300 according to an embodiment of the disclosure is illustrated in Figure 3. The system 300 comprises a processing unit 302 comprising a central processing unit (CPU) 310, memory 312, and an input / output (I / O) bus 314 communicatively coupled with one or more of the CPU 310 and memory 312. The system may be configured to perform a method as illustrated in Figures 1, 2A, 2B.

[0094] Memory 312 may be configured to store user data 102, such as physiological data 102A and seizure activity data 102B associated with a particular subject. The physiological data 102A and / or the seizure activity data 102B may be derived from physiological signals such as EEG signals, from patient or clinical records, or may be may be detected from electrodes or an implantable device.

[0095] Memory 312 may further comprise computer executable instructions (for example, a model generator module 312A), which when executed by the CPU 310, cause the processing unit 302 to perform a method for generating one or more subject-specific seizure risk or seizure likelihood model(s) models for determining a probability or likelihood of a future seizure occurrence in each of a plurality of time windows, as described in PCT / AU2018 / 050575 and / or in PCT / AU2021 / 051118, the entire content ofboth of which is incorporated herein. The one or more temporal probabilistic models may be generated according to the method shown in Figure 1 of PCT / AU2018 / 050575, as described between paragraphs

[0092] and

[0132] . The one or more temporal probabilistic models may be generated according to the method shown in Figures 1 and 4 of PCT / AU2021 / 051118, as described between paragraphs

[0036] and

[0040] , and

[0065] and

[0078] ,

[0096] The model generator module 312A may be configured to determine subjectspecific seizure likelihood model(s) iteratively. The processing unit 302 may be configured to generate one or more updated subject- specific seizure likelihood models for a particular subject in response to receipt by the processing unit 302 of new user data 102 associated with the particular subject.

[0097] Memory 312 may comprise computer executable code, (for example, a seizure forecasting module 312B), which when executed by the CPU 310, causes the processing unit 302 to determine an estimate of seizure probability, or seizure likelihood measure in the subject for one or more of the plurality of first time windows using the subjectspecific seizure likelihood model(s) model.

[0098] Memory 312 may comprise computer executable code, (for example, an expected seizure rate determination module 312C), which when executed by the CPU 310, causes the processing unit 302 to determine an expected seizure rate based on an actual seizure rate and the seizure likelihood measure determined by seizure forecasting module 312B.

[0099] Memory 312 may further comprise computer executable code, (for example, an alert generation module 312D), which when executed by the CPU 310, causes the processing unit 302 to generate an alert based on the forecasted seizure cycles in the subject. For example, such alerts may include one or more of: a seizure occurrence probability, a seizure occurrence risk rating, information concerning the cause of risk elevation, a recommendation to take and / or change medication and / or therapeutic typeor amount, a recommendation to administer stimulation and / or to alter one or more parameters of a therapeutic device for delivering stimulation to the subject, for example.

[0100] The system 300 may further comprise a measurement unit 304. The measurement unit 304 may be coupled to one or more devices for recording physiological data, such as non-EEG physiological data, and / or EEG physiological data.

[0101] The non-EEG physiological data 102A may comprise cardiac output recorded over the first time period. The cardiac output may comprise one or more of: (i) heart rate, and (ii) heart rate variability. For example, the non-EEG physiological data 102A may comprise an electrocardiograph (ECG) and / or a photo-plethysmograph (PPG) signal received from the heart of the subject. The non-EEG physiological data 102A may comprise values of one or more variables of sleep recorded over the first time period. The non-EEG physiological data 102A may comprise values of one or more variables of activity recorded over the first time period. For example, the non-EEG physiological data 102A may comprise an actigraphy received from the subject. The non-EEG physiological data 102A may comprise values of one or more variables of oxygen saturation recorded over the first time period. The non-EEG physiological data 102A may comprise values of one or more variables of electrodermal activity recorded over the first time period. The non-EEG physiological data 102A may comprise values of one or more variables of respiratory rate recorded over the first time period.

[0102] Measurement devices which may be coupled to the measurement unit 304 may comprise (but are not limited to) an EEG monitoring device 316, a heart monitor (photo- plethysmograph or ECG) 318, a sweat or electro dermal sensor 320 an accelerometer 322 (or similar motion detector), a temperature sensor 323, and oxygen saturation measurement device 324, such as a pulse oximeter, and a respiratory monitor 325. Other examples of measurement devices which may be coupled to the measurement unit 304 include blood pressure monitors, glucose monitors, cortisol sensor, and gyroscopes etc.

[0103] The measurement unit 304 may comprise one or more amplifiers and / or digital signal processing circuitry for processing signals received from the one or moremeasurement devices 316, 318, 320, 322, 323, 324, 325. Such signal processing circuitry may include, for example, sampling circuits for sampling signals received from the one or more measurement devices 316, 318, 320, 322, 323, 324, 325 as well as filters for filtering such signals in accordance with embodiments described above. The measurement unit may also be configured to extract and process information received from the one or more measurement devices 316, 318, 320, 322, 323, 324, 325. To that end, the measurement unit 304 may include memory to store data received from the one or more measurement devices 316, 318, 320, 322, 323, 324, 325.

[0104] The EEG monitoring device 316 may comprise one or more electrode leads each comprising one or more electrodes. Such leads may be implanted intracranially (intracranial EEG) and / or located subcutaneously and / or located external to the head. Leads which are implanted intracranially may be placed on the surface of the brain and / or implanted within the brain tissue. Leads of the EEG monitoring device 316 may be configured, in use, to record neural activity at a neural structure in a brain of the subject. Where EEG is utilised for determining seizure activity data 102B, the measurement unit 304 may also be used in conjunction with a signal generator (not shown) to measure electrode impedances.

[0105] The measurement unit 304 may be external to or integrated within the processing unit 302. Communication between the measurement unit 304 (and / or the signal generator) on the one hand and the EG bus 314 on the other may be wired or may be via a wireless link, such as over inductive coupling, WiFi, Bluetooth (RTM) or the like. Equally, communication between the measurement unit 304 and the one or more measurement devices 316, 318, 320, 322, 323, 324, 325 may be wired or may be via a wireless link such as those listed above.

[0106] Power may be supplied to some or all elements of the system 300 from at least one power source 324. The at least one power source 324 may comprise a battery such that elements of the system 300 can maintain power independent of mains power when implanted into a subject.

[0107] In some embodiments, some or all functions of the measurement unit 304 may be implemented using the processing unit 302, in which case, the one or more measurement devices 316, 318, 320, 322, 323, 324, 325 may be coupled directly to the I / O bus 314.

[0108] In some embodiments, the I / O bus 314 is configured for wired and / or wireless communications via a communications network, and the processing unit 302 is configured to cooperate with the I / O bus 314 to transmit or otherwise provide one or more models for determining a probability of a future seizure occurrence in each of a plurality of time windows for a particular subject to a remote server (not shown) or device (not shown). In some embodiments, the processing unit 302 is configured to cooperate with the I / O bus 314 to transmit or otherwise provide the estimate of seizure probability in the subject for one or more of the plurality of time windows determined using the model to the remote server (not shown) or device (not shown). The device maybe a smart phone or watch associated with the subject for whom the one or more models have been generated. For example, the one or more models may be made available for the subject to download to a personal device from a website.

[0109] In some embodiments, the processing unit 302 is configured to cooperate with the VO bus 314 to transmit or otherwise provide the one or more possible seizure triggers of the subject and / or an alert based on the possible seizure triggers to a remote server (not shown) or device (not shown).

[0110] In some embodiments, the identified one or more possible seizure triggers may be stored on the device (such as the subject's personal device). In some embodiments, the model generator module 312A, seizure forecasting module 312B, expected seizure rate determination module 312C, and / or alert generation module 312D may be stored in memory of the device and one or more of the described methods may be performed by processor(s) of the device executing computer code of the generator module 312A, seizure forecasting module 312B, expected seizure rate determination module 312C and / or alert generation module 312D.

[0111] The system 300 may further comprise one or more input device 308 and / or one or more output devices 306. The one or more input devices 308 may include, but are not limited to, one or more of a keyboard, mouse, touchpad and touchscreen. Examples of the one or more output devices 306 include displays, touchscreens, light indicators (LEDs), sound generators and haptic generators. Input and / or output devices 308, 306 may be configured to provide feedback (e.g. visual, auditory or haptic feedback) to a subject.

[0112] Feedback provided by the one or more output devices 306 may include information or an alert based on seizure occurrence likelihood in a subject, particularly a subject to which information recorded by the measurement unit 304 relates. Such information or alerts may include one or more of: a seizure occurrence probability, a seizure occurrence risk rating, information concerning the cause of risk elevation (e.g. increased heart rate, decreased heart rate variability, sleep deprivation etc.), a recommendation to take and / or change medication or therapeutic type and / or amount, a recommendation to administer stimulation and / or to alter one or more parameters of a therapeutic device for delivering stimulation to the subject, and / or that the therapeutic device has automatically changed the stimulation regime. Such information may be portrayed graphically and / or through the use of auditory or haptic feedback (as discussed above).

[0113] The one or more input devices 308 may enable the subject to acknowledge information and feedback provided via the one or more output devices 306 as well as input data into the system 300 which may then be used to improve forecasting of future seizures. This information may include data pertaining to the subject symptoms during a seizure. Such information may be used to verify and / or refute predictions previously made by the system 300.

[0114] Other information may include seizure activity data 102B associated with the subject, which may be used to update the seizure activity data 102B stored in memory 312. In some embodiments, receipt of such seizure activity data 102B may instigate execution of the model generator module 312A by the processing unit 302 to generateone or more updated temporal probabilistic models for the subject associated with the new seizure activity data 102B. This may instigate execution of the seizure forecasting module 312B by the processing unit 302 (or processing unit of the device, for example) to generate updated likelihood of seizure in a subject over a plurality of time windows.

[0115] To monitor the time of day for the purposes of forecasting seizure occurrence likelihood based on the circadian profile of a subject, the processing unit 302 may comprise a 12-hour or 24-hour clock operable to measure time.

[0116] The temperature sensor 323 may be provided to monitor temperature for the purpose of forecasting seizure occurrence likelihood based on a temperature based probability model associated with a subject.

[0117] In some embodiments, some or all of the system 300 shown in Figure 3 may be implemented using a smartphone or similar digital device (tablet, watch, computer etc.).

[0118] Figure 4 illustrates an exemplary device 400 upon which one or more units, components or modules of the system 300 may be implemented. The device 400 comprises a display 402. In some embodiments, the display 402 comprises a touchscreen display 402, which can function as both the input device 308 and output device 306 of Figure 3. In the embodiment shown in Figure 4, the touchscreen 402 provides a graphical illustration 404 of the probability of a seizure occurring, together with written information concerning seizure probability 406, a seizure risk rating 408 and information concerning the cause of elevation of risk 410 (as discussed above). In addition, a button 412 is provided on the touchscreen to enable a subject to input data concerning a seizure which has occurred. Auditory warnings may also be playable on the device 400.

[0119] In some embodiments, the one or more measurement devices 316, 318, 320, 322, 323, 324, 325 may be coupled to the device 400 via one or more wired or wireless links, in which case, the measurement unit 304 may be implemented using the device 400. Alternatively, the measurement unit 304 may be separate to the device 400, in whichcase, the measurement unit 304 itself may be wired or wirelessly coupled to the device 400.

[0120] In some embodiments, as depicted in Figure 5 there is shown a communications system 500 comprising a seizure management system 502. The seizure management system may be configured to perform the methods illustrated and discussed in and with reference to Figures 1 to 4. The seizure management system 502 comprises one or more processors 504 and memory 506. The processor(s) 504 may include an integrated electronic circuit that performs the calculations such as a microprocessor, graphic processing unit, for example. In some embodiments, the system 502 may be implemented as a distributed system comprising multiple server systems configured to communicate over a network to provide the functionality of the model generation system 502. One or more of the multiple server systems may be cloud-based server systems.

[0121] Memory 506 may comprise both volatile and non-volatile memory for storing executable program code, or data. Memory 506 comprises program code which when executed by the processor(s) 504, provides the various computational and data management capabilities of the seizure management system 502. The block diagram of Figure 5 illustrates some of the modules which may be stored in memory 506, and which, when executed by the processor(s) 504 of the system 502, cause the system to perform method 600 discussed below with reference to Figure 6. For example, memory 506 may comprise one or more of the generator module 312A, seizure forecasting module 312B, expected seizure rate determination module 312C, and alert generation module 312D.

[0122] The seizure management system 502 may further comprise a network interface 508 to facilitate communications with components of the system 502 across a communications network 510, such as computer device(s) 514, database 512 and / or other servers. The network interface 508 may comprise a combination of network interface hardware and network interface software suitable for establishing, maintaining and facilitating communication over a relevant communication channel.

[0123] The communications system 500 may comprise or otherwise have access to the database 512 for storing data used by the seizure management system 502. The database 512 may be implemented using a relational database or a non-relational database or a combination of a relational database and a NoSQL database. The system 502 may access the database 512 directly or via the communications network 510. The database 512 may comprise user data 102 associated with a plurality of subjects or patients. The user data 102 may be recorded over a period of time and may comprise EEG physiological data 102A, non-EEG physiological data 102A, seizure activity data 102B, and / or environmental data 102C. The database 512 may by updated periodically or aperiodically to update user data 102 associated with the respective subject.

[0124] In some embodiments, the system 502 is configured to generate one or more subject- specific seizure likelihood model(s), such as temporal probabilistic models as described above with reference to Figures 1 to 3, for a particular subject. Once generated, the model(s) may be deployed on a computer device 514 for use by a subject, and may for example, be used to determine seizure likelihood measures such as seizure probability in the subject for one or more time windows. In some embodiments, once the seizure system 502 has estimated seizure likelihood measures for a particular subject using the subject- specific seizure likelihood model(s), the measures may be saved on a computer device 514 for use by a patient.

[0125] The computer device 514 may be remote from the system 502. The computer device 514 may be configured to download the subject- specific seizure likelihood model(s), seizure probability in the subject for one or more time windows (for example, determined using the subject- specific seizure likelihood model(s)), and / or measures, from the seizure management system 502 via communications network 510. For example, the computing device 514 may be an end-user computing device such as a smart watch, a mobile device, a tablet device, a desktop computer, a laptop computer, etc.

[0126] In some embodiments, where updated user data 102 is available to the seizure management system 502, the model generator module 312A of the model generation system 502 may be configured to generate updated subject-specific seizure likelihoodmodel(s) for forecasting future seizure activity in the subject, which may be provided as update(s) to the computing device 514.

[0127] In some embodiments, where updated user data 102 is available to the seizure management system 502, the seizure forecasting module 312B of the seizure management system 502 may be configured to estimate an updated seizure probability in the subject for one or more time windows using the updated subject- specific seizure likelihood model(s), which may be provided as update(s) to the computing device 514.

[0128] Figure 6 is a process flow diagram of a method 600 for projecting variability over time of likelihood of occurrence of a seizure in a subject having epilepsy across a future time period. The method 600 may be executed by one or more processors of the processing unit 302 of the system 300 executing instructions stored in memory 312 of the system 300, such as model generator module 312A, seizure forecasting module 312B, and expected seizure rate determination module 312C, for example. The method 600 may be performed by the seizure management system 502 executing instructions stored in memory 506 of the system 502, such as model generator module 312A, seizure forecasting module 312B, and expected seizure rate determination module 312C, for example.

[0129] At S602, the system determines a first seizure likelihood measure 206A of seizure activity of a subject over or across a first time period. In some embodiments, the first time period is at least the duration of the subject’s longest seizure cycle.

[0130] The system may determine the first seizure likelihood measure 206A using one or more temporal probabilistic models 204 generated according to methods described in PCT / AU2018 / 050575 and / or in PCT / AU2021 / 051118. The one or more temporal probabilistic models 204 may be generated according to the method shown in Figure 1 of PCT / AU2018 / 050575, as described between paragraphs

[0092] and

[0132] . The one or more temporal probabilistic models 204 could be generated according to the method shown in Figures 1 and 4 of PCT / AU2021 / 051118, as described between paragraphs

[0036] and

[0040] , and

[0065] and

[0078] ,

[0131] In some embodiments, the first seizure likelihood measure 206A is based at least in part on subject- specific historical seizure data that relates to seizure events that occurred before the first time period, such as user data 102.

[0132] The first seizure likelihood measure 206A may comprise values or measures across one or more first time windows of the first time period. For example, the first seizure likelihood measure 206A may comprise an hourly probability. For example, the hourly probability may be a probability or likelihood that the subject will have a seizure for each hour of the first time period.

[0133] At S604, the system determines an actual seizure rate 208 of the subject across the first time period. The actual seizure rate 208 may be based on user data 102 associated with the subject experiencing epileptic events. The system may determine the actual seizure rate 208 of the subject over the first time period based at least in part on user data 102 indicating seizure events prior to the first time period.

[0134] The actual seizure rate 208 may comprise measures or values determined across one or more first time windows of the first time period. For example, the actual seizure rate 208 may be a value indicating the number of seizures per hour, that is an hourly seizure rate, for each day, that is, the actual seizure rate 208 may comprise an hourly seizure rate for each day of the duration of the first time period. The actual seizure rate 208 may be a variable indicating the number of seizures per hour, that is an hourly seizure rate, for the first time period. The actual seizure rate 208 may be a variable indicating the number of seizures per week, that is a weekly seizure rate, for the first period. The actual seizure rate 208 may indicate any one of: a seizure rate per hour, a seizure rate per day, a seizure rate per week, or a seizure rate per month, for the first time period, for example.

[0135] The first time period may comprise a first start or beginning time point as being a time before the forecast is generated. For example, the first start time point may be a time a few hours, day, weeks, or months before the forecast is generated. The first time period may comprise a first end time as being a time at or before the forecast is generated and after the first start time point. The duration from the first beginning time point to thefirst end time point may correspond with the longest useful seizure count. The first beginning time point may be a time associated with the earliest seizure data available.

[0136] At S606, the system (for example the system shown in Figures 3 and 5) determines a second seizure likelihood 206B of seizure activity of a subject over or across a second time period. In embodiments where S606 is performed prior to the second time period, the system determines the second seizure likelihood forecast 206B of seizure activity of a subject over the second time period. The second seizure likelihood forecast 206B may be determined during or after the second time period has occurred. In embodiments where S606 is performed after the second time period, the system determines the second seizure likelihood measure 206B of seizure activity of a subject over the second time period. The system may retrospectively determine the second seizure likelihood forecast 206B of seizure activity of a subject the second time period. In some embodiments, the second time period is at least the duration of the subject’s longest forecast seizure cycle after the forecast is generated.

[0137] The second seizure likelihood forecast 206B comprises an hourly probability. For example, the hourly probability may be a probability or likelihood that the subject will have a seizure for each hour of the second time period. The system may determine the second seizure likelihood forecast 206B using one or more temporal probabilistic models 204 generated according to methods described in PCT / AU2018 / 050575 and / or in PCT / AU2021 / 051118. The second seizure likelihood forecast 206B is based at least in part on subject- specific historical seizure data that relates to seizure events that occurred during and / or before the first time period, such as user data 102, for example. The one or more temporal probabilistic models 204 may be generated according to the method shown in Figure 1 of PCT / AU2018 / 050575, as described between paragraphs

[0092] and

[0132] . The one or more temporal probabilistic models 204 could be generated according to the method shown in Figures 1 and 4 of PCT / AU2021 / 051118, as described between paragraphs

[0036] and

[0040] , and

[0065] and

[0078] .

[0138] At S608, the system determines a risk-adjusted actual seizure rate 210 of the subject experiencing epileptic events across the first time period. The risk-adjusted actualseizure rate 210 is based on the actual seizure rate 208 and the first seizure likelihood 206A. In determining the risk-adjusted actual seizure rate 210, system may initially normalise the actual seizure rate 208 of the first time period to generate a respective normalised actual seizure rate. The normalised actual seizure rate may be normalised between 0 and 1, for example.

[0139] The system adjusts the actual seizure rate 208 or normalised actual seizure rate by the first seizure likelihood 206A, determined at S602, to determine the risk-adjusted actual seizure rate 210 over the first time period. The system may divide the actual seizure rate 208 or normalised actual seizure rate by the first seizure likelihood 206A to determine the risk-adjusted actual seizure rate 210 over the first time period. The system may divide the actual seizure rate 208 or normalised actual seizure rate by a function of the first seizure likelihood 206A to determine the risk-adjusted seizure rate over the first time period. The function of the first seizure likelihood 206A may be (1 - first seizure likelihood 206 A).

[0140] At S610, the system determines an expected seizure rate 212 of the subject experiencing epileptic events across the second time period. The expected seizure rate 212 is based on the risk-adjusted actual seizure rate 210 and the second seizure likelihood 206B. The expected seizure rate 212 may be a value indicating an expected seizure rate per hour for the second time period.

[0141] The system then adjusts the risk-adjusted actual seizure rate 210 over the first time period by the second seizure likelihood 206B to determine the expected seizure rate 212 over the second time period. The system may multiply the first risk-adjusted actual seizure rate 210 by a function of the second seizure likelihood 206B to determine the expected seizure rate 212. The function of the second seizure likelihood 206B may be (1 - second seizure likelihood 206B).

[0142] At S612, the system projects the expected seizure rate 212 into the second time period to forecast future seizure activity. The second time period may be a future time period relative to the first time period. In some embodiments, S606, S608, and S610 maybe omitted from method 600. That is, at S612, the system 300, 502 may project the first actual seizure rate into the second time period to forecast future seizure activity.

[0143] The first time period may comprise any one of: a period of relatively high seizure likelihood, a period of relatively moderate seizure likelihood, a period of relatively low seizure likelihood, a period of transition from a relatively high seizure likelihood to a relatively low seizure likelihood, or a period of transition from a relatively low seizure likelihood to a relatively high seizure likelihood, or a combination thereof. The second time period may comprise any one of: a period of relatively high seizure likelihood, a period of relatively moderate seizure likelihood, a period of relatively low seizure likelihood, a period of transition from a relatively high seizure likelihood to a relatively low seizure likelihood, or a period of transition from a relatively low seizure likelihood to a relatively high seizure likelihood, or a combination thereof.

[0144] Referring to Figure 7, there is shown a graphical example of projecting variability over time of likelihood of occurrence of a seizure in a subject having epilepsy across a future time period. As shown in Figure 7, the first time period, ti, is a duration of 10 weeks, and the second time period, t2, is a duration of 5 weeks. Further, the seizure likelihood during the first time period, Pi, is determined to be 0.4, and number of seizures during the first time period, ai, is determined to be 12 seizures. That is, the subject has been monitored for 10 weeks, the first time period, and their seizures have been recorded and the first seizure likelihood measure calculated, for example.

[0145] Upon projecting variability over time of likelihood of occurrence of a seizure in a subject having epilepsy across a future time period, the first actual seizure rate 208, A1, is determined, the second seizure likelihood 206B is forecasted, P2, and the expected seizure rate 212, E2, is calculated. For example:

[0146] The number of seizures in the second time period can then be forecasted,2, being 1.8 seizures per week for 5 weeks, totalling 9 expected seizures in the second period.

[0147] Referring back to Figure 2A, and with reference to explanatory Figure 2C, the set- specific temporal probability model for a particular set may be generated based only on historical epileptic seizure events from the record belonging to the pertinent set. Such a configuration is illustrated in Figure 2C by the solid lines joining set-specific seizure activity data 106A-106C to the respective forecast model 108-112. In that case the process of Figures 6 and 7 may be performed as described. That is, a temporal probability model for a particular set pertaining to a distinct seizure type, for example forecast model 108 for seizure type A, may be generated to forecast the distinct seizure type, and its distinct cyclical pattern, separately from alternative seizure types, wherein the distinct seizure type and alternative seizure types include any one of the following seizure types and their sub-seizure types as previously described: focal onset, generalised onset, or unknown onset, for example. In this particular example, generating the temporal probability model for a particular type uses, from among the comprehensive seizure activity data 102B, only the seizure activity data from the same particular type. For example, forecast model 108 for type A uses only type A seizure activity data 106A, forecast model 110 for type B uses only type B seizure activity data 106B, forecast model 112 for type C uses only type C seizure activity data 106C. These exclusive data relationships are indicated by the solid lines in Figure 2C.

[0148] In an alternative case, the set-specific temporal probability model for a particular set may be generated based on historical epileptic seizure events from the record belonging to the pertinent set and other historical epileptic seizure events from the record (such as those belonging to other seizure sets). In that case, the historical epileptic seizure events belonging to the pertinent set are processed as seizure events for the first time period (for example at S604). Again, this is indicated in Figure 2C by the solid lineslinking the set- specific seizure activity data 106A-C to respective forecast models 108- 112. The historical epileptic seizure events belonging to other seizure sets are not treated as seizure events for the first time period in the same way, rather, they are used more in the same way as physiological data or environmental data. That is, they are modelled over the same first time period and compared to the seizure events over the same first time period to identify correlations or relevant patterns. The correlations or relevant patterns may then be leveraged in generating the set-specific temporal probability model. This is indicated by the dashed lines linking the set-specific seizure activity data 106A- C to respective forecast models 108-112. In generating the type B forecast model 110, the set-specific seizure activity data for type B events 106B is treated as historic seizure events, whereas the set- specific seizure activity data for type A and C events 106 A and 106C, is treated as background or environmental information that is processed to determine whether any pattern or correlation exists between seizure events in those sets and seizure events satisfying the profile of a type B seizure event.

[0149] Referring to Figures 8A and 8B, there is shown an example forecast visualisation 800, being a visual presentation of generated temporal probability models resulting from forecast visualisation 116. The example forecast visualisation 800 is a screen capture of an application running on a personal computing device. Figure 8A shows a first portion of the example forecast visualisation 800 screen capture and 8B shows a second portion of the example forecast visualisation 800 screen capture, such that Figures 8A and 8B when combined form the example forecast visualisation 800. That is, Figure 8B is an extension of Figure 8 A and each of Figures 8 A and 8B share a portion of the example forecast visualisation 800 screen capture.

[0150] Forecast visualisation 800 includes:- regions and / or data highlighted by, or presented on a background of, a first colour 801 representative of a low risk rating,- regions and / or data highlighted by, or presented on a background of, a second colour 802 representative of a baseline risk rating,- regions and / or data highlighted by, or presented on a background of, a third colour 803 representative of a high risk rating,- regions and / or data highlighted by, or presented on a background of, a fourth colour 804 representative of a low risk rating in a non-current month, and- regions and / or data highlighted by, or presented on a background of, a fifth colour 805 representative of a baseline risk rating in a non-current month.Each of the first colour 801, second colour 802, and third colour 803 of the forecast visualisation 800 may be shown as a texture, or any suitable visual representation for conveying the respective relevant information. Forecast visualisation 800 further includes a calendar portion 806 of a month desired to be shown. In this example, the month of July in the year 2023 is shown, however, it will be understood that a user may be able to adjust the month shown, for example, by using the month selection option 810.

[0151] Forecast visualisation 800 further includes a legend 808, indicating each of the first colour 801, the second colour 802, and the third colour 803 and what they represent. Each day shown in the calendar portion 806 is represented by one of the first colour 801, the second colour 802, or the third colour 803 to provide the user with an indication of the average forecast seizure risk rating for the respective day for the membership criteria of user preferences 104. For example, 3 July indicates a low seizure risk rating for the day for the respective membership criteria, and 4 July indicates a baseline seizure risk rating for the day for the same membership criteria.

[0152] Forecast visualisation 800 further shows a daily seizure risk breakdown 812, which indicates the patients seizure risk rating throughout the selected day. As shown in forecast visualisation 800, the 5th of July is selected as indicated by selection indicator 814. The seizure risk breakdown 812 includes an hourly timeline 816 that shows a 24- hour time scale. The current time is indicated by time indicator 818. For example, the current time when the screen capture of the forecast visualisation 800 was recorded is indicated to be roughly 9am. The time indicator 818 may allow a user to quickly identifywhat their seizure risk rating is at the current time when looking at the daily seizure risk breakdown 812.

[0153] The daily seizure risk breakdown 812 includes a baseline 820, indicated by the second colour 802, which is the patients average seizure risk rating determined by generation of the temporal probability models from forecast visualisation 116. From the baseline 820 there extends further risk ratings, such as risk ratings 822 and 824, for example, which show the patients risk for an hour relative to the baseline 820. For example, risk rating 822 indicates that the patient is expected to have a high seizure risk rating from 4am to 5am, further indicated by the third colour 803. For example, risk rating 824 indicates that the patient is expected to have a low seizure risk rating from 11pm to 12am, further indicated by the first colour 801. The seizure risk ratings are determined to be baseline risk, high risk, or low risk for each hour of the selected day, as shown in Figure 8B. The seizure risk ratings shown are dependent on the generated temporal probability models resulting from forecast visualisation 116.

[0154] The forecast visualisation 800 further shows optional menu selections 826 for navigating the application running on the personal computing device. A reporting button 828 may also be included for allowing a user to navigate to a user interface for reporting seizure activity data 102B, such as seizure characterisation data and / or seizure timeline data. An additional selectable help prompt 830 may also be shown for providing explanations and information pertaining to the visualised forecast shown in the daily seizure risk breakdown 812.

[0155] It will be appreciated by persons skilled in the art that numerous variations and / or modifications may be made to the above-described embodiments, without departing from the broad general scope of the present disclosure. The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive.

[0156] The steps, features, integers, compositions and / or compounds disclosed herein or indicated in the specification of this application individually or collectively, and any and all combinations of two or more of said steps or features.

Claims

CLAIMS:

1. A method comprising: receiving a record of historical epileptic seizure events in a patient, including, for each recorded historical epileptic seizure event, seizure timeline data representing date and time of the seizure, and seizure characterisation data characterising duration, symptom profile, and / or severity of the seizure event; dividing the recorded seizure events of the record of historical epileptic seizure events into a plurality of seizure sets defined by respective membership criteria, the membership criteria comprising, for the respective seizure set, one or more value or value ranges of the seizure characterisation data; for at least one seizure set individually, using the record of historical epileptic seizures to generate a set-specific temporal probability model, wherein the set-specific temporal probability model projects variability over time of likelihood of occurrence of a seizure event fitting the membership criteria for the respective individual seizure set into a future time period.

2. The method of claim 1, wherein in generating the set-specific temporal probability model for each seizure set individually, only the recorded seizure events belonging to the respective individual seizure set are used to generate the said set-specific temporal probability model.

3. The method of claim 1, wherein in generating the set-specific temporal probability model for each seizure set individually, the recorded seizure events belonging to the respective individual seizure set and recorded seizures belonging to one or more other seizure sets among the plurality of distinct seizure sets are used to generate the said set- specific temporal probability model.

4. The method of any one of claims 1 to 3, wherein the one or more value ranges of the seizure characterisation data forming the membership criteria for the seizure sets are defined, at least in part, according to user preferences input to an application running on a personal computing device.

5. The method of any one of claims 1 to 4, wherein the one or more value ranges of the seizure characterisation data forming the membership criteria for the seizure sets are defined, at least in part, according to a clustering algorithm executed on the seizure characterisation data of the record of historical epileptic seizures.

6. The method of any one of claims 1 to 5, wherein the seizure characterisation data is self-reported by the patient via a user interface of an application running on a personal computing device or via a dedicated seizure reporting device in data communication with an application running on a personal computing device.

7. The method of any one of claims 1 to 6, wherein the seizure timeline data is self-reported by the patient via a user interface of an application running on a personal computing device or via a dedicated seizure reporting device in data communication with an application running on a personal computing device.

8. The method of any one of claims 1 to 6, wherein the seizure characterisation data is automatically generated by one or more of: a wearable monitoring device, an environmental monitoring device, and an implantable monitoring device.

9. The method of any one of claims 1 to 8, wherein the seizure timeline data is automatically generated by one or more of a wearable monitoring device and an implantable monitoring device.

10. The method of any one of claims 1 to 9, further comprising, at a user interface of an application running on a personal computing device, presenting the temporal probability model for a user-selected one of the plurality of distinct seizure sets to a user on a display screen.

11. The method of claim 10, wherein the presenting includes, in accordance with user preferences, highlighting time windows in which seizure-set-specific likelihood of seizure event occurrence exceeds a threshold, the threshold being either an absolute value or a defined amount or proportion above an average or other characteristic value.

12. The method of any one of claims 1 to 11, further comprising, weighting the temporal probability models for the plurality of seizure sets according to the one or more value ranges of the seizure characterisation data forming the membership criteria of the respective sets, and summing the weighted temporal probability models to obtain a weighted aggregate temporal probability model representing weighted aggregate likelihood of seizure event occurrence for the plurality of seizure sets collectively for the future period of time.

13. The method of claim 12, further comprising, at a user interface of an application running on a personal computing device, presenting the weighted aggregate temporal probability model on a display screen.

14. The method of claim 12 or claim 13, including highlighting future time windows in which weighted aggregate likelihood of seizure occurrence for the plurality of seizure sets collectively exceeds a predefined threshold.

15. An apparatus comprising memory hardware and processor hardware, the memory hardware storing processing instructions which, when executed by the processor hardware, cause the apparatus to perform the method according to any of claims 1 to 14.

16. A system comprising the apparatus according to claim 15, and at least one of: a dedicated seizure reporting device in data communication with an application running on a personal computing device, the dedicated seizure reporting device being configured to record epileptic seizure event data; a wearable monitoring device, the wearable monitoring device being configured to record epileptic seizure event data; an environmental monitoring device, the environmental monitoring device being configured to record environmental data; andan implantable monitoring device, the implantable monitoring device being configured to record epileptic seizure event data.

17. A computer program which, when executed by processor hardware of an apparatus comprising memory hardware and processor hardware, causes the apparatus to perform the method according to any of claims 1 to 14.

18. A non-transitory computer-readable medium storing the computer program according to claim 16.

Images