GB2567533A

GB2567533A - Medical sensor

Info

Publication number: GB2567533A
Application number: GB1813118.5A
Authority: GB
Inventors: Alexander Bernstein David; Kahagala-Gamage Sankha
Original assignee: Sankha Kahagala Gamage
Current assignee: Sankha Kahagala Gamage
Priority date: 2017-08-11
Filing date: 2018-08-10
Publication date: 2019-04-17
Also published as: GB201712892D0; GB201813118D0

Abstract

The probability of a seizure occurring in the imminent future is determined based on measurements of core body temperature and heart rate variability (HRV). A first one of the parameters is monitored and a change or deviation from normal values leads to a determination of a first probability of seizure. A change or deviation in the second parameter leads to a modification of the first probability to determine a second probability of seizure. Alternatively a single probability based on both parameters may be generated. An alarm may be raised when probabilities exceed a threshold. The combination of measurements allows the ruling out of false positives, and enhanced certainty that the conclusion is correct, giving a user advanced warning of imminent seizure. The system may be wearable, for example incorporated into a garment such as a vest (figs. 2 & 6). The system may communicate with a mobile phone 8, which may or may not be used for data processing.

Description

Medical sensor

Field of the Invention

The invention relates to medical sensors, and more particularly to medical sensors for predicting a seizure.

Background

At any given time, around 1% of the population suffers from epilepsy, with around 4% or the population being affected at some point in their life. Perhaps the most commonly recognised feature of epilepsy is seizures, which can range from momentary losses of awareness (known as absence seizures) to uncontrolled jerking movements. For a sufferer, a seizure of any severity can be a troubling and traumatic event. Worse still, as the severity increases, so too does the risk of physical injury, caused by a user biting themselves, or but uncontrolled jerking causing a sufferer to hit themselves or objects in their vicinity.

While seizures can also have non-epileptic causes, epileptic seizures characterised by abnormally excessive or synchronous neuronal activity in the brain receive much of the focus of the medical community. All types of seizures require medical attention, but since they occur sporadically and without obvious advance warning, even regular sufferers are often caught by surprise by the onset of a seizure. Unless it occurs in a hospital or with a medical professional close at hand, it is unusual for a suffer to receive medical attention until after the seizure has run its course and any damage due to the unexpectedness of the seizure having already occurred.

The present invention aims to address some or all of these problems.

Summary

Disclosed herein is a system for predicting a seizure; comprising: a processor; a temperature sensor for monitoring core body temperature; a heart rate sensor for monitoring heart rate; wherein the processor is configured to: monitor heart rate data output from the heart rate sensor to determine a heart rate variability in order to obtain a first probability of a seizure; monitor core body temperature; and based at least in part on the core body temperature modify the first probability to obtain a second probability of a seizure. While changes in certain vital signs may be indicative of the imminent onset of a seizure, there can be other causes for particular vital signs changing. The inventors have found that where the heart rate variability changes, indicating a certain probability of an imminent seizure, the probability of an imminent seizure is markedly increased when a change in the core body temperature is also detected. The system and method described herein is designed to predict seizures of all types, but particularly epileptic attacks, and even more specifically occipital lobe epilepsy attacks. Note that the order of the steps is not critical, and, for example, the temperature may be monitored and lead to a first probability, which is then modified to form a second probability in the light of heart rate variability data. Equally, there may be a steady output of probability of an imminent seizure, based on both temperature and heart rate data, such that one parameter is not strictly correlated with the first probability and the other is not strictly correlated with the modification of the first to form the second. In any event, the overall effect is to either use a second measurement of a parameter to confirm the initial finding that an anomalous reading in one parameter is indeed indicative of an imminent seizure, which is notified to a user by the probability of a seizure being increased. In the alternative, the second parameter measurement is used to contradict the first parameter measurement, e.g. to determine that it is a false alarm. This is communicated to the user as a reduced probability of an imminent seizure.

In this latter case, it may be more appropriate to think of the system as monitoring first and second vital signs and in the case where a change in the first vital sign is correlated with a change in the second vital sign, outputting an increased probability of a seizure.

Typically, the strongest indication of an imminent seizure is where the heart rate increases and/or the core body temperature drops, although other combinations of vital sign changes can also provide useful indications. Indeed, by constant monitoring of heart rate; heart rate variability; core body temperature; and/or rate of change of core body temperature, and correlating these data sets with one another, an accurate picture of the likelihood of an imminent seizure can be provided. It is possible to detect relatively small changes in core body temperature and heart rate for use in determining a probability of an imminent seizure. In some cases, the magnitude of the change in core body temperature or heart rate is directly correlated with the change in probability of a seizure.

Providing advanced warning of a seizure in the form of an elevated probability of an imminent seizure occurring can give a user of the system time to prepare by alerting a medical professional or other helper, moving away from hard or sharp objects and towards softer surroundings, or putting themselves into a recommended safety position to minimise harm, for example. A user may have up to ten minutes advanced warning under ideal conditions. Typically a user has around five minutes advanced warning, but clearly even one minute advanced warning represents a significant benefit. Consequently, where this document refers to “imminent” seizures, this shall mean a seizure which will occur within the next ten minutes or so.

As used herein, heart rate variability relates to situations where the frequency of heart beats changes. This may, for example be provided as:

BPMi - BPM_bBPM_b

In which BPM, is the instantaneous heart rate at any given time in beats per minute and BPM_b is a suitable background heart rate in beats per minute. This form has the effect that a constant heart rate has a variability of 0, a sudden 50% increase in heart rate has a variability of 1/2, and a halving of heart rate has a variability of -1/2. Other definitions may also be used, which use a differential form to take into account a rate of change of heart rate with time.

Similarly, core body temperature is used to distinguish from the temperature at the skin surface of a user because the external skin temperature is vastly affected by extraneous parameters such as air temperature, the body part being measured, degree of contact between the sensor and the skin, etc.

Any appropriate sensors may be used for the parameters of heart rate (such as LEDbased sensors, piezoelectric sensors, ECG devices, etc.) and temperature (such as thermistors, thermocouples, thermopiles, fluid thermometers, infrared-based thermometers, etc.)

Optionally at least a portion of the system is portable, and preferably wearable. This allows a user to live their life relatively unaffected, but secure in the knowledge that a seizure is less likely to catch them unaware as they know that their vital signs are constantly being monitored to give them advanced warning. In some cases the wearable portion of the system is attachable to a garment, or is, or includes a garment. This allows a user to integrate the system into their life without undue complexity.

The sensors can be removed from the garment if necessary for e.g. cleaning the garment or the sensors.

Particularly suitable garments include shirts, vests, t-shirts, jackets, blouses, jumpers, etc. as they are worn on the torso, and so are suitable for mounting the sensors in a position to measure heart rate and/or core body temperature. Other garments which are considered are underwear, trousers, shorts, belts, skirts, scarves, hats, gloves, socks, etc.

Advantageously the garment is configured to be worn on a human torso or abdomen. In some cases the wearable portion of the system includes the heart rate sensor and/or the temperature sensor. The torso or abdomen is an area which is particularly suitable for unobtrusive measurements of heart rate and core temperature due to the proximity to, respectively, the heart, and the armpit (which is a suitable place to measure core body temperature).

Where the sensors or other parts of the measurement system (processor, communications module, etc.) are attached to the garment, they may be waterproof (indeed, the entire measurement system may be waterproof) to avoid damage due to sweat, rain, etc. In some cases the garment may even be a swimming costume, which will also advantageously have waterproof measurement elements, if any are provided on the garment.

The sensors may be configured to measure their respective properties without requiring direct skin contact. For example, the core body temperature sensor may be a noncontact infrared thermometer. Having sensors which do not require contact with the skin can result in a much more comfortable experience for a user.

The garment may be adjustable, for example to improve comfort, or to expand the device. Epilepsy can affect children, and it may be preferred for children to have one garment which they keep for an extended period, which can be expanded as they grow.

The wearable portion of the system may comprise a power source in some cases, where the power source is selected from: a battery; wireless energy; energy harvester; a connection to mains electricity; and/or a power generator. For example, solar power, Peltier effect based wearable power generation, energy harvesters for harvesting energy from background electromagnetic radiation, fuel cells, inductive or capacitive coupling to a power source etc. could all be considered in some designs.

In some examples a complete system may be provided in which the wearable portion of the system further includes the processor. In other cases, the processor is not part of the wearable portion of the system, which keeps the complex and expensive parts of the system safe.

The processor may be part of a: computer; mobile device; virtual computer; tablet; cloud computing; and/or embedded system. This allows the calculations to be performed in any location which is convenient for a user. In particular, these devices are conveniently accessible by a user of the system, and readily provide the necessary processing power to make a determination. They also typically have good connectivity, allowing a signal to be sent to parties other than the user.

The modification of the first probability to obtain the second probability may include determining a correlation between the heart rate variability and change in core body temperature. For example, in the event that the heart rate variability is correlated with the change in core body temperature, the second probability is higher than the first probability, indicating that a seizure is more likely than either the heart rate variability or the change in core body temperature alone would have indicated. This correlation helps to rule out false positives, thereby improving the reliability of the determination of probability of an imminent seizure. The correlation may be determined by a statistical model, for example Multivariate Statistical Process Control (MSPC).

The system may further comprise an alert means. This provides a manner by which a user or medical professional may be alerted to an imminent seizure. The alert means may comprise one or more of: sound generation means (a buzzer, speaker or the like); light generation means (LEDs, bulbs, fibre-optic communications and the like); movement generation means (e.g. vibrating parts for haptic alerting); and/or inter-computer communication means (wired or wireless signalling capability). Each of these allows a user or a nearby medical professional to be alerted to the situation and respond accordingly. In some cases, the medical professional may be some distance away, but are nevertheless able to take action, e.g. moving to where the user is. In some cases, for example where the alert means is inter-computer alert means, the system may also include location identifying means (GPS-based, RFID-based, etc.) and include a location in the alert message, to assist a medical professional or other recipient of the alert to locate the user (who may be unable to respond in the usual way if the seizure has already started, for example).

When an alert is issued, it may start gently, and gradually increase in intensity. A sudden loud sound, strong vibration, bright light, etc. can be shocking to a user and increase their heart rate (among other effects). As noted above, changes in heart rate can be indicative of an imminent seizure. In this scenario an alert may cause a feedback loop in the monitoring algorithm where an alert triggers a raise in heart rate, which triggers an alert, further raising the heart rate, and so on. Worse, the raised heart rate may itself increase stress levels, and contribute to a seizure, depending on the nature of the seizure and the individual concerned. These effects can be mitigated or eliminated by the form of the alert being that described above, with a gentle start and a gradually increasing intensity, optionally further where the intensity is capped at a relatively low level.

The system may be configured to alert a user if the first and optionally additionally the second probability exceeds a respective threshold. In some cases, the first probability can exceed its threshold and trigger an alert on its own. This may be the case if the first probability exceeds a high threshold. That is, where one or other of the parameters exceeds a “normal” range by a large amount, simple versions of the algorithm may interpret this as a sharp increase in the probability of a seizure. However, as noted above, there is a possibility that this corresponds instead to a change in vital signs having a different cause. In this case, the uncertainty in cause is captured by requiring a higher probability threshold for a single parameter than for two parameters in a correlated way.

The probability threshold of the second probability can be substantially lower than the first threshold, due to the increased degree of confidence provided by both measurements in conjunction. Waiting until the modified (or updated) second probability exceeds a threshold can help to reduce the incidence of false alarms. As noted above, heart rate can change for many reasons (exercise, stress, watching a horror movie, etc.), not all of which will signal an imminent seizure. Examples of a probability threshold for a single parameter to cause an alarm to be issued are if it exceeds a probability of 80% in some cases, whereas for two parameters, this number might be as low as 50%.

Additionally, the system may be configured to alert a user if both of the heart rate variability and core body temperature readings reach a respective threshold value. These thresholds can be useful in determining that the user is approaching an unsafe level. In the cases where a single parameter exceeds its respective threshold, this may trigger a recalibration of the system. This is because where one parameter changes, it is usually correlated in some way with other parameters. A single parameter exceeding its respective threshold value can therefore be indicative of a sensor error, and the recalibration is a way of providing a robust measurement and monitoring system which self-corrects.

In the context of this disclosure, a parameter exceeding its threshold can mean that the value is greater than a particular amount, or less than a particular amount. In general, heart rate variability and core body temperature each have a “normal” value which can vary between individuals. Small deviations from this “normal” value occur naturally throughout the day. The threshold values referred to above are best thought of in some cases as a range around the “normal” value which will not be treated as unusual. In these cases, exceeding a threshold means where the relevant sensors measure values of: the heart rate; heart rate variability; core body temperature; or rate of change of core body temperature which do not fall within this broadened region around the normal value. The “normal” value and/or range can be determined by the system by collecting data over a long period, and training itself on the typical values a user experiences. In some cases, a user may inform the system that they are e.g. exercising, so the heart rate and body temperature are expected to increase. In these cases, the system may be configured to suppress warnings, or to require a higher probability of a seizure to be determined before alerting a user.

An alert may be issued in the form of an audio message (recorded words, klaxons, alarms, etc.); a visual message (flashing lights, words on a screen, etc.); haptic events (e.g. vibration); wired communication; and/or wireless communication (e.g. to a computer or mobile telephone device).

The temperature sensor may comprise a plurality of sensing units. This can help provide redundancy in case one sensor malfunctions, or it can provide verification of the results of one sensor.

Where sensors are provided on the wearable portion of the system, the placement of at least one of the temperature sensors may be adjustable relative to the wearable portion of the system. This can help to adapt the device to different body shapes, or to position the temperature sensor in the best place to measure the core body temperature.

The system may further comprise software downloadable to a portable device. This can form part of the alerting system, e.g. on a user’s mobile telephone, tablet, computer, etc., or it can allow statistical monitoring of seizure timings, frequencies, etc. to a medical professional to assist in diagnosis and provision of treatment regimes. As part of this, the system may include sensors for other vital signs, for example blood pressure and respiration rate which can be factored into a more holistic view of the underlying causes of the seizures.

The downloadable software may allow a user, medical professional or other carer/helper to annotate the data with other information of interest, including but not limited to: height, weight, diet, blood sugar levels, ambient air temperature and prevailing weather, sleeping patterns, medication usage, etc. to assist in gaining a broad picture of the situation. This additional information can be synchronised with the measured vital signs data, to correlate the effect of the various recorded events with the measurements of the vital signs.

The wearable portion of the system may further include a communication module. This allows the sensor information (and optionally the calculated probability of an imminent seizure) to be transmitted to another device, e.g. to alert a user or update a medical record. The communication may use any suitable communications method, such as: Bluetooth (RTM); Wi-Fi (RTM); Near-field communication; Light-based communication; RF communications; wireless communications; and/or wired data links. These communication means can allow the communication module to communicate between the wearable portion of the system and an external device selected from: a computer; a mobile device; a virtual computer; a tablet; a cloud computing; and/or an embedded system. The external device may have an application installed on it to assist in communicating with the system, for example to provide the necessary levels of security.

Also disclosed herein is a method of predicting an epileptic seizure comprising: monitoring heart rate data output from the heart rate sensor to determine a heart rate variability in order to obtain a first probability of a seizure; monitoring core body temperature; and based at least in part on the core body temperature modifying the first probability to obtain a second probability of a seizure. While certain changes in vital signs may be indicative of the imminent onset of a seizure, there can be other causes for particular vital signs changing. The inventors have found that where the heart rate variability changes, indicating a certain probability of an imminent seizure, the probability of an imminent seizure is markedly increased when a change in the core body temperature is also detected. Note that the order of the steps is not critical, and, for example, the temperature may be monitored and lead to a first probability, which is then modified to form a second probability in the light of heart rate variability data. Equally, there may be a steady output of probability of an imminent seizure, based on both temperature and heart rate data, such that one parameter is not strictly correlated with the first probability and the other is not strictly correlated with the modification of the first to form the second.

Typically, the strongest indication of an imminent seizure is where the heart rate increases and/or the core body temperature drops, although other combinations can also provide useful indications. Typical changes in core body temperature of 1 degree or more and in heart rate of 5% of resting rate are easily detected and represent a sufficiently significant change to trigger a change in probability. In some cases, the magnitude of the change in core body temperature or heart rate is directly correlated with the change in probability of a seizure.

Providing advanced warning of a seizure in the form of an elevated probability of an imminent seizure occurring can give a user of the system time to prepare by alerting a medical professional or other helper, moving away from hard or sharp objects and towards softer surroundings, or putting themselves into a recommended safety position to minimise harm, for example. A user may have up to ten minutes advanced warning under ideal conditions. Typically a user has around five minutes advanced warning, but clearly even one minute advanced warning represents a significant benefit.

Where the method modifies the first probability to obtain the second probability, this may include determining a correlation between the heart rate variability and change in core body temperature. For example, in the event that the heart rate variability is correlated with the change in core body temperature, the second probability is higher than the first probability, indicating that a seizure is more likely than either the heart rate variability or the change in core body temperature alone would have indicated. This correlation helps to rule out false positives, thereby improving the reliability of the determination of probability of an imminent seizure. The correlation may be determined by a statistical model, for example Multivariate Statistical Process Control (MSPC).

In some examples, a part of the method is performed using a portable, or preferably a wearable system including a garment. This may be, for example, the garment or system set out above, or any other appropriate garment. In any case, this allows a user to live their life relatively unaffected, but secure in the knowledge that a seizure is less likely to catch them unaware as they know that their vital signs are constantly being monitored to give them advanced warning, by allowing a user to integrate the system into their life without undue complexity.

Advantageously the garment is configured to be worn on a human torso or abdomen. In some cases the garment includes the heart rate sensor and the temperature sensor, e.g. attached to the garment. The torso or abdomen is an area which is particularly suitable for unobtrusive measurements of heart rate and core temperature due to the proximity to, respectively, the heart, and the armpit (which is a suitable place to measure core body temperature).

The garment may be adjustable, for example to improve comfort, or to expand the device. Epilepsy can affect children, and it may be preferred for children to have one garment which they keep for an extended period, which can be expanded as they grow. The method may therefore include adjusting the garment to improve the measurement and/or increase comfort.

In some examples a complete system may be provided in which the wearable portion of the system further includes a processor for receiving data on the heart rate and the core body temperature, providing first and second probabilities, outputting alerts, performing statistical analyses, and so forth. In other cases, the processor is not part of the wearable portion of the system, which keeps the complex and expensive parts of the system safe.

In some examples, readings are taken from the sensors at predetermined intervals, for example every 100ms or so. Readings of this frequency are useful for calculating rates of change of the measured parameters. Storing many such readings can allow the detection of longer term trends, which can be used to assist the system in identifying abnormal vital signs. For example, heart rate variability relates to changes in heart rate. This means that it is useful to identify a range of “normal” behaviour as a baseline. Measurements of the heart rate over a long period of time can be used to train a processor to identify abnormal behaviour. Similarly, general trends in core body temperature can be used to train the processor to identify unusual changes with greater confidence.

Moreover, in some cases the readings may be averaged or smoothed over a time window to discount false alarms. The longer term trends mentioned above may be based on either a series of instantaneous measurements of one or both parameters (e.g. heart rate and core body temperature), or on the smoothed or averaged measurements of one or both parameters, or indeed on both data types of one or both parameters.

The method may further include alerting a user if the first and optionally additionally the second probability exceeds a respective threshold. In some cases, the first probability can exceed its threshold and trigger an alert on its own. This may be the case if the first probability exceeds a high threshold. That is, where one or other of the parameters exceeds a “normal” range by a large amount, simple versions of the algorithm may interpret this as a sharp increase in the probability of a seizure. However, as noted above, there is a possibility that this corresponds instead to a change in vital signs having a different cause. In this case, the uncertainty in cause is captured by requiring a higher probability threshold for a single parameter than for two parameters in a correlated way.

Additionally, the method may include alerting a user if both of the heart rate variability and core body temperature readings reach a respective threshold value. These thresholds can be useful in determining that the user is approaching an unsafe level. In the cases where a single parameter exceeds its respective threshold, this may trigger a recalibration of the system. This is because where one parameter changes, it is usually correlated in some way with other parameters. A single parameter exceeding its respective threshold value can therefore be indicative of a sensor error, and the recalibration is a way of providing a robust measurement and monitoring system which self-corrects. As noted above, “exceeding a threshold” can be interpreted in some cases as where a parameter is measured which falls outside of a “normal” range of values.

An alert may be issued in the form of an audio message (recorded words, klaxons, etc.); a visual message (flashing lights, words on a screen, etc.); haptic events (e.g. vibration); wired communication; and/or wireless communication (e.g. to a computer or mobile telephone device).

The method may include outputting an alert if both of the heart rate variability and core body temperature readings reach a respective threshold value. This provides an alternative method of providing the user with a warning that a seizure is imminent.

The alert may be output via an alert means. This provides a manner by which a user or medical professional may be alerted to an imminent seizure. The alert means may comprise one or more of: sound generation means (a buzzer, speaker or the like); light generation means (LEDs, bulbs, fibre-optic communications and the like); movement generation means (e.g. vibrating parts for haptic alerting); and/or inter-computer communication means (wired or wireless signalling capability). Each of these allows a user or a nearby medical professional to be alerted to the situation and respond accordingly. In some cases, the medical professional may be some distance away, but are nevertheless able to take action, e.g. moving to where the user is. In some cases, for example where the alert means is inter-computer alert means, the method may make use of location identifying means (GPS-based, RFID-based, etc.) and include a location in the alert message, to assist a medical professional or other recipient of the alert to locate the user (who may be unable to respond in the usual way if the seizure has already started, for example).

The core body temperature may be performed using a temperature sensor comprising a plurality of sensing units. This can help provide redundancy in case one sensor malfunctions, or it can provide verification of the results of one sensor.

Where sensors are provided on the wearable portion of the system, the placement of at least one of the temperature sensors may be adjustable relative to the wearable portion of the system. This can help to adapt the device to different body shapes, or to position the temperature sensor in the best place to measure the core body temperature. Accordingly, the method may include adjusting the position of the sensor(s) for comfort or to improve the data quality.

The method may further comprise downloading software to a portable device. This can form part of the alerting processes, e.g. on a user’s mobile telephone, tablet, computer, etc., or it can allow statistical monitoring of seizure timings, frequencies, etc. to a medical professional to assist in diagnosis and provision of treatment regimes. As part of this, the method may make use of sensors for other vital signs, for example blood pressure and respiration rate which can be factored into a more holistic view of the underlying causes of the seizures. The downloadable software may allow a user, medical professional or other carer/helper to annotate the data with other information of interest, including but not limited to: height, weight, diet, blood sugar levels, ambient air temperature and prevailing weather, sleeping patterns, medication usage, etc. to assist in gaining a broad picture of the situation. This additional information can be synchronised with the measured vital signs data, to correlate the effect of the various recorded events with the measurements of the vital signs.

The garment may further include a communication module. This allows the sensor information (and optionally the calculated probability of an imminent seizure) to be transmitted to another device, e.g. to alert a user or update a medical record. The communication may use any suitable communications method, such as: Bluetooth (RTM); Wi-Fi (RTM); Near-field communication; Light-based communication; RF communications; wireless communications; and/or wired data links. These communication means can allow the communication module to communicate between the wearable portion of the system and an external device selected from: a computer; a mobile device; a virtual computer; a tablet; a cloud computing; and/or an embedded system. The external device may have an application installed on it to assist in communicating with the system, for example to provide the necessary levels of security.

For the avoidance of doubt, the present disclosure also relates to a system configured to carry out any of the methods described herein, even where such systems are not explicitly described. Similarly, where the present disclosure relates to a method performed on a specific system, this should be interpreted as a disclosure of the method per se, even where such method are not explicitly described separately from the specific system.

Brief Description of the Drawings

Embodiments of the invention will now be described with reference to the drawings, in which:

Figure 1 shows a medical sensor according to the invention;

Figure 2 shows a front and rear view of the medical sensor of Figure 1 attached to a garment;

Figure 3 shows a medical sensor of the present invention communicating with a mobile device;

Figure 4 shows a medical sensor of the present invention communicating with a mobile device for processing data;

Figure 5 shows a medical sensor according to the invention including an integrated power unit; and

Figure 6 shows a front and rear view of the medical sensor of Figure 6 attached to a garment.

Detailed Description

The present invention describes a wearable device to predict imminent seizures by the wearer. Heart rate changes occur during and leading up to a seizure, and the inventors have developed an algorithm to measure the heart rate variability, and determine that a seizure will occur imminently. Additionally, a drop in core body temperature also correlates with an imminent seizure. An algorithm takes inputs of measurements from sensors for heart rate and core temperature and outputs a probability of an imminent seizure based on both parameters. This greatly increases the confidence in the result, and allows seizures to be predicted with good accuracy. The sensors need not be directly attached to a patient’s skin, thereby avoiding discomfort to the patient. Overall, the device is cost-effective, and likely to be worn by the patient outside of a specialist setting, while providing a high degree of confidence that the user will be alerted in advance of a seizure.

Throughout the Figures, similar elements are labelled with the same numbers, and operate in broadly the same manner, so will not be described in detail in each Figure.

Consider now Figure 1, which shows a medical sensor for predicting a seizure. The sensor comprises four temperature sensors 1 a-1 d, a processor 5, a heart rate monitor 2 and an alert generating means 3. The heart rate monitor 2 and the temperature sensors 1a-1d are connected to the processor 5 by wires 4 to provide a path along which signals from the sensors 1, 2 can be provided to the processor. In use, a user arranges the sensors 1, 2 in such a way as to measure their core body temperature and their heart rate. The readings from the sensors 1, 2 are provided to the processor 5 along wires 4. In this way, the processor 5 is able to monitor the heart rate and detect any variability in the heart rate. This heart rate variability is then used to determine a first probability of an imminent seizure. The core body temperature is also received by the processor 5. In the event that a change in the core body temperature is detected, the algorithm incorporates this information into the probability determination and modifies it. This is possible because an imminent seizure is more likely in cases where both heart rate variability and core body temperature changes correlate with one another. Similarly, a change in a first parameter may look like a seizure is imminent, but the second parameter remaining static can be used to determine that the change in the first parameter had another cause (or even was due to sensor 1, 2 or processor 5 error).

Where the processor 5 determines that the probability of a seizure is high enough to warrant it, a signal may be sent to the alert generating means 3, which outputs a human or machine interpretable alert. On registering this alert, a user can take action such as alerting medical staff, getting to a safe location or putting themselves in a safe position.

Turning now to Figure 2, which shows the sensor of Figure 1 attached to a garment, in this case a vest 6. Note here that because the sensors 1, 2 are mounted on garment 6, the user can easily take off the sensors 1, 2 if they become uncomfortable or if the user wishes to have a shower. Likewise, because the garment 6 conforms to the user’s body shape, they can put the garment 6 on again, and be confident that the sensors will be located in the correct position, with only small adjustments being needed.

The wiring 4, sensors 1, 2, alerting means 3, and processor 4 are mounted on the inside of the garment 6, so as to prevent damage to them, and hold them close to a user’s skin to improve the measurement. The wires 4 are held in place relative to the garment to reduce the likelihood that they will become tangled. The temperature sensors 1a-1d are positioned in the left pectoral, right pectoral, left stomach and right stomach positions to provide an accurate core body temperature measurement. The heart rate sensor 2 is placed over any artery, or close to the heart. The alerting means is placed in an appropriate place depending on the type of alert is it intended to output. For visual alerts, this would be on the front of the vest 6, preferably in the natural line of sight of a user. For audio alerts, this is close to the ears. For haptic alerts, this is somewhere with a reasonably high nerve density, for example placed under the neckline of the garment to be in contact with skin below the collarbone.

A vest 6 (along with t-shirts, shirts, jackets, etc.) is a particularly suitable garment to use with the present invention, since these are all worn on the torso and/or abdomen, so can be placed so as to allow sensors to easily measure a user’s core body temperature and heart rate.

Turning now to Figures 3 and 4, the device of Figure 1 is shown communicating with a mobile device 8 via communications pathway 7. In each Figure, the device is provided with a communication module (not shown), in order to enable the communication between the sensors 1, 2 and the processor 5 on the one hand and the mobile device 8 on the other hand. The communication module may be part of the processor or be formed as a separate unit. The communication between the device and the mobile device 8 may take any suitable form, including by not limited to wired, wireless or fibre optic communications. This communication is useful to provide the user with an alternative alerting means, for example, or indeed for any of the other situations set out above, as shown in Figure 3.

In Figure 4, the mobile device 8 takes on a degree of data processing. This means that the processor 5 need not be powerful. Its only task may be to collect the data form the sensors 1,2 and forward it to the mobile device 8 for processing. This allows the processor 5 to be small, lightweight, and cheap, while the mobile device carries out the bulk of the calculations. In cases where the algorithm is complex, this also has advantages of speed, and of not causing overheating of the processor 5, which may be uncomfortable for a user.

In both Figures 3 and 4, the use of an external device (which can in some cases be a computer, laptop, cloud storage, etc.) allows the data to be passed on to other parties, e.g. medical professionals, carers, helpers, parents, etc. As noted above, the data can include that a seizure is anticipated or is happening, where the user is, what time the relevant events occurred, etc.

Finally, consider Figures 5 and 6 which show a device (and garment 6) similar to those shown in Figures 1 and 2. The main difference here is the provision of a power source 9 as part of the medical sensor. This allows the device to power the sensors and the processor, without requiring plugging in to mains power. The power source is connected to the processor 5 for powering it, and also if needed for powering the sensors. In some examples, the sensors may be passive in that they do not require external power sources to generate a signal, for example thermocouples generate a voltage from temperature differences and piezoelectric sensors can be used to generate a voltage directly from a movement (e.g. from a pulse) which is indicative of a heart beat. In cases where there is a communications module for communicating with an external device, the integrated power source 9 can be used to power the communications module as well.

In the above discussion, many embodiments show four temperature sensors, but as noted above, there may be fewer than this, or indeed more. The device can work with only a single temperature sensor, for example.

Similarly, the discussion focussed on providing a first probability of an imminent seizure from the heart rate, and this probability is modified to form a second probability based on a measurement of a core body temperature. However, the order in which the measurements are taken and used can be broader than this. For example, the temperature may be monitored and lead to a first probability, which is then modified to form a second probability in the light of heart rate variability data. Equally, there may be a steady output of probability of an imminent seizure, based on both temperature and heart rate data, such that one parameter is not strictly correlated with the first probability and the other is not strictly correlated with the modification of the first to form the second, but rather the two parameters continually contribute to ongoing updates of the probability of a seizure.

The garment in all cases is shown as a vest, but as noted above a wide range of garments may be appropriate. Most preferred are torso or abdominal garments (shirts, tshirts, vests, waistcoats, cummerbunds, blouses, jackets, jumpers, hoodies, etc.), but scarves, trousers, shorts, socks, shoes, hats, gloves, swimwear, underwear etc. may all be appropriate garments for attaching all or part of the system to. Each of these garments may be adjustable in part or in full, e.g. to adjust to a growing child or to improve user comfort.

The location of the sensors, processor, alerting means, and power source where appropriate, are all for illustration purposes, and can be located at any appropriate point on the body. In this case “appropriate” means suitable for carrying out their role, so for sensors this means able to measure the relevant parameter to a good degree of accuracy. For other components, this means being unobtrusive and comfortable. For the communications module, this means having a clear communications channel to a separate device, which may mean ensuring a clear line of sight, ensuring there is no blocking effect of wireless signals, or ensuring that a plug for receiving a cable to form a wired connection is accessible while the garment (and medical sensor) are being worn on a user’s body.

Claims

1. A system for predicting a seizure; comprising:

a processor;

a temperature sensor for monitoring core body temperature;

a heart rate sensor for monitoring heart rate;

wherein the processor is configured to:

monitor heart rate data output from the heart rate sensor to determine a heart rate variability in order to obtain a first probability of a seizure;

monitor core body temperature; and based at least in part on the core body temperature modify the first probability to obtain a second probability of a seizure.

2. The system according to claim 1, wherein at least a portion of the system is wearable.

3. The system according to claim 2, wherein the wearable portion of the system is attachable to a garment.

4. The system according to claim 3 further including the garment.

5. The system according to claim 3 or 4, wherein the garment is configured to be worn on a human torso or abdomen.

6. The system according to any one of claims 2 to 5, wherein the wearable portion of the system includes the heart rate sensor and the temperature sensor.

7. The system according to any one of claims 2 to 6, wherein the wearable portion of the system further includes the processor.

8. The system according to any one of claims 2 to 6, wherein the processor is not part of the wearable portion of the system.

9. The system according to claim 8, wherein the processor is part of a:

computer;

mobile device;

virtual computer;

tablet; cloud computing; and/or embedded system.

10. The system according to any preceding claim, wherein the modification of the first probability to obtain the second probability includes determining a correlation between the heart rate variability and change in core body temperature.

11. The system according to claim 10, wherein the correlation is determined by a statistical model.

12. The system according to claim 11, wherein the statistical model is Multivariate Statistical Process Control (MSPC).

13. The system according to any one of claims 10 to 12, wherein, in the event that the heart rate variability is correlated with the change in core body temperature, the second probability is higher than the first probability.

14. The system of any preceding claim, further comprising an alert means.

15. The system of claim 14, wherein the alert means comprises one or more of: sound generation means; light generation means; movement generation means; and/or inter-computer communication means.

16. The system of any preceding claim, further comprising alerting a user if the second probability exceeds a threshold.

17. The system according to claim 16, wherein the alert comprises at least one of: audio; visual; haptic; wired communication; and/or wireless communication.

18. The system according to any preceding claim, wherein the temperature sensor comprises a plurality of sensing units.

19. The system according to claim 18, wherein the placement of at least one of the temperature sensors is adjustable relative to the wearable portion of the system.

20. The system according to any preceding claim, further comprising software downloadable to a portable device.

21. The system of any preceding claim, wherein the wearable portion of the system further includes a communication module.

22. The system according to claim 21, wherein the communication module is configured to communicate using:

Bluetooth (RTM);

Wi-Fi (RTM);

Near-field communication;

Light-based communication;

RF communications;

wireless communications; and/or wired data links.

23. The system according to claim 21 or 22, wherein the communication module is configured to communicate between the wearable portion of the system and an external device selected from:

a computer;

a mobile device;

a virtual computer;

a tablet;

a cloud computing; and/or an embedded system.

24. The system of claim 23, further including an application installed on the external device.

25. The system of any preceding claim, wherein the wearable portion of the system comprises a power source selected from:

a battery;

wireless energy;

energy harvester;

a connection to mains electricity; and/or a power generator.

26. A method of predicting an epileptic seizure comprising:

monitoring heart rate data output from the heart rate sensor to determine a heart rate variability in order to obtain a first probability of a seizure;

monitoring core body temperature; and based at least in part on the core body temperature modifying the first probability to obtain a second probability of a seizure.

27. The method according to claim 26, wherein the modification of the first probability to obtain the second probability includes determining a correlation between the heart rate variability and change in core body temperature.

28. The method according to claim 27, wherein the correlation is determined by a statistical model.

29. The method according to claim 28, wherein the statistical model is Multivariate Statistical Process Control (MSPC).

30. The method according to any one of claims 26 to 29, wherein, in the event that the heart rate variability is correlated with the change in core body temperature, the second probability is higher than the first probability.

31. The method of any one of claims 26 to 30, wherein a part of the method is performed using a wearable system including a garment.

32. The method of claim 31, wherein the garment is configured to be worn on a human torso or abdomen.

33. The method of any one of claims 31 or 32, wherein the monitoring the core body temperature and/or the heart rate is performed by sensors attached to the garment.

34. The method according to any one of claims 31 to 33, wherein the processor is attached to the garment.

34. The method according to any one of claims 31 to 33, wherein the processor is not attached to the garment.

35. The method of any one of claims 26 to 34, wherein readings are taken from the sensors at predetermined intervals.

36. The method of any one of claims 26 to 35, wherein readings from the sensors are averaged over a time period.

37. The method according to claim 26 to 36, wherein method includes monitoring changes in trends of heart rate variability and/or core body temperature.

38. The method according to claim 37, wherein the processor is configured to output an alert if both of the heart rate variability and core body temperature readings reach a respective threshold value.