Detection of brain state responses and monitoring system

Abstract

A method for aiding the detection of drug effect using anaesthesia-induced patterns in an electroencephalogram (EEG), comprising: obtaining EEG data corresponding to one or more EEG signals of a patient under general anaesthesia measured using an EEG, wherein the one or more EEG signals are time-varying and comprise a time series of segments comprising at least a repeating first segment type and a repeating second segment type, wherein instances of segments corresponding to the first segment type alternate with instances of segments corresponding to the second segment type; and determining a signal metric comprising a series of time ordered metric datapoints, wherein each metric data point is determined based on: a comparison between at least two segments temporally associated with the metric point; and / or a comparison between at least two different measures of at least one segment temporally associated with the metric point, wherein the signal metric is a measurement of an effect of an anaesthetic drug regimen administered to the patient, and related methods and systems.

Description

DETECTION OF BRAIN STATE RESPONSES AND MONITORING SYSTEMField of the Invention

[0001] The invention generally relates to a method for aiding the detection of brain state responses, for example due to drug effects such as anaesthesia-induced patterns in an electroencephalogram (EEG).Background to the Invention

[0002] Brain monitoring with EEG in anaesthetised adults has been used to understand titration of dosing, perioperative outcomes, and the neurophysiologic basis of anaesthesia. In adults, typical EEG changes with inhalational anaesthetics and propofol include global increases in amplitude with gradual slowing of oscillations during anaesthesia induction, followed by frontal alpha (8 to 12 Hz) predominance during anaesthesia maintenance. With further increasing dose, burst suppression develops. Burst suppression is a phenomenon of profound discontinuity ascribed to a causative factor, where discontinuity is a phenomenon of alternating periods of high amplitude mixed-frequency activity (bursts) and of little or no activity (interburst interval, IB I). Burst suppression under general anaesthesia is more likely to be observed in neurologically vulnerable patients such as adults requiring surgery for epilepsy treatment, the aging population and adults with neurodevelopmental disorders.

[0003] EEG changes during general anaesthesia have been reported throughout childhood. Conclusions about specific age-related changes, particularly in the very young, are limited by the broad age ranges of reports in the literature. With general anaesthesia, alpha oscillations emerge around 3 -months of age and become increasingly concentrated in the frontal cortex by seven months of age. Total frontal EEG power increases with age and anaesthetic depth between four months and six-to-eight years of age, thereafter, decreasing with increasing age. Other reproducible changes seen in EEG with general anaesthesia (e.g. alpha oscillation coherence, phase amplitude coupling) are not seen in infants under one year of age. Development of burst suppression with general anaesthesia is more likely at younger ages.

[0004] Reports have described that total EEG power in infants during general anaesthesia tends to be lower in older children. In term neonates, exposure to volatile anaesthesia has been shown to decrease signal power in the 0.5-4 Hz frequency band but have no effect on signal power in frequency ranges from 5-20 Hz and 30-100 Hz. Burst-suppression ratio, 90% spectral edge frequency, relative beta ratio and approximate entropy all show little change in neonates with sevoflurane concentrations between 0.5% and 2% in neonates. Just one study has been published reporting an increase in discontinuity from baseline in term neonates during generalanaesthesia, obtained using amplitude-integrated EEG (aEEG) [1], aEEG is a processed EEG output that automatically grades discontinuity according to the amplitude of IBIs compared with thresholds of 5 pV, 10 pV and 25 pV. In term neonates undergoing cardiac surgery, burst suppression and progression to an isoelectric EEG has been reported during cardiopulmonary bypass and deep hypothermic cardiac arrest.Neonatal EEG

[0005] Typical neonatal EEG differs from that of older children and adults, it exhibits patterns that predictably mirror each stage of brain development. Age differences of just a few weeks lead to different EEG characteristics. Each trace must therefore be interpreted in the context of the subjects’ post-menstrual age (time from conception).

[0006] A striking background feature of preterm EEG is the presence of discontinuity. This phenomenon is a natural part of brain developmental and not a disease process or secondary to a causative factor. In developmental discontinuity, IBIs are defined by periods of relative attenuation below age-specific amplitude thresholds (ranging from 25 pV to 50pV), lasting two seconds or more, against amplitudes up to 300pV. However, the underlying mechanisms for developmental discontinuity are incompletely understood. Burst amplitude and IBI duration steadily decrease with maturation. Symmetry, synchrony, reactivity, and typical graphoelements also follow characteristic patterns with age.

[0007] At term age, the background EEG activity is usually continuous, with an amplitude of 25 pV to 50 pV. The only discontinuity that persists is trace alternans, a subtle form of discontinuity with especially short IBIs that only occurs during quiet sleep and disappears by 40 weeks post-menstrual age.

[0008] The terms burst suppression and discontinuity are often used interchangeably in anaesthesia literature. However, in the developing brain, burst suppression and discontinuity are not synonymous. Remarkably, while developmental discontinuity is well defined, definitions of burst suppression vary widely across all ages and there is no global consensus. Definitions generally refer to prolonged periods of attenuated activity (often defined as under 5 pV) alternating with periods of paroxysmal high amplitude activity. In the developing brain, burst suppression is also distinguished from excess discontinuity by the absence of normal patterns within the bursts. Neonatal burst suppression is described clinically by the “sharpness” of bursts and the durations of interburst intervals. It is an ominous sign in the EEG of any neonate and indicative of a poor prognosis, especially if it is non-reactive and present during wakefulness. Non-reactive burst suppression is associated with severe neurodevelopmentaldisability or death in over 90% of cases. The occurrence of burst suppression in neonates during general anaesthesia, or any harmful association thereof, has not been robustly shown in the literature.Neonatal general anaesthesia

[0009] Any inference that burst suppression might confer harm to neonates during general anaesthesia must also consider the protective effect of general anaesthesia in this population. General anaesthesia suppresses an exaggerated stress response to surgery that would otherwise result in circulatory and metabolic complications in the postoperative period. This effect has been robustly demonstrated with either inhalational or opioid based general anaesthesia, although the dose required for this effect is unknown. Conversely, excessive exposure to general anaesthesia may pose a risk of harm due to direct neurological toxicity. Prolonged exposure to high doses of general anaesthesia has been shown to be harmful in animal models. However, this toxicity has not been found in humans at usual clinical doses and the doses that might be required to cause harm is unknown.

[0010] Too much and too little anaesthesia pose a risk of harm in the neonatal brain. However, there is no way known to measure the effect of anaesthesia directly from the neonatal brain, despite it being the intended site of action. Considering the exponential growth of EEG- derived depth of anaesthesia devices for adults, which are rapidly becoming standard of care in contemporary anaesthesia practice, it is appealing to apply these devices directly to neonates. However, it is well documented that these devices perform poorly in the young, most profoundly in those aged under one year. This is unsurprising due to the distinct characteristics describe above of the usual neonatal EEG and the paucity of data about any impact thereon of general anaesthesia.

[0011] Reference made herein to background art does not constitute an admission that the background art or any specific publication forms a part of the common general knowledge in the art, in Australia or any other country.Summary of the Invention

[0012] According to an aspect of the present disclosure, there is provided a method for aiding the detection of drug effect using anaesthesia-induced patterns in an electroencephalogram (EEG), comprising: obtaining EEG data corresponding to one or more EEG signals of a patient under general anaesthesia measured using an EEG, wherein the one or more EEG signals are time-varying and comprise a time series of segments comprising at least a repeating first segment type and a repeating second segment type, wherein instances of segmentscorresponding to the first segment type alternate with instances of segments corresponding to the second segment type; and determining a signal metric comprising a series of time ordered metric datapoints, wherein each metric data point is determined based on: a comparison between at least two segments temporally associated with the metric point; and / or a comparison between at least two different measures of at least one segment temporally associated with the metric point, wherein the signal metric is a measurement of an effect of an anaesthetic drug regimen administered to the patient.

[0013] The patient may be a neonate.

[0014] Optionally, the first segment type corresponds to burst periods. Optionally, the second segment type corresponds to interburst intervals (IB I). In a typical implementation, the first segment type corresponds to burst periods and the second segment type corresponds to interburst intervals (IB I).

[0015] The metric datapoints are optionally determined, at least in part, in accordance with an inter-segment comparison. For a particular metric datapoint the inter-segment comparison may comprise: identifying a first selection comprising one or more instances of the first segment type being temporally correlated with the particular metric datapoint; identifying a second selection comprising one or more instances of the second segment type being temporally correlated with the particular metric datapoint; and comparing the first selection to the second selection. The first selection may comprise a single instance of the first segment type. The second selection may comprise a single instance of the second segment type. The one or more instances of the first segment type of the first selection and the one or more instances of the second segment type of the second segment may be associated with a temporally contiguous portion of the EEG data. Determining the particular metric datapoint may comprise: determining at least one inter-segment ratio by taking a ratio of a measure of the one or more EEG signals of the first selection and a measure of the one or more EEG signals of the second selection. Determining the particular metric datapoint may comprise: determining at least a first inter-segment ratio and a second inter-segment ratio, wherein at least one of the measure of the one or more EEG signals of the first selection and the measure of the one or more EEG signals of the second selection of the first inter-segment ratio is different to at least one of the measure of the one or more EEG signals of the first selection and the measure of the one or more EEG signals of the second selection of the second inter-segment ratio, wherein determining the particular metric datapoint comprises normalising and combining at least the first inter-segment ratio and the second inter-segment ratio.

[0016] At least one inter-segment ratio may comprise a measure of the one or more EEG signals of the first selection and / or a measure of the one or more EEG signals of the second selection selected from at least one of: a measurement of the EEG signal central tendency of the associated one or more EEG signals, for example a signal mean, median or count; a measurement of the EEG signal strength of the associated one or more EEG signals, for example deviation from a central tendency such as signal variance or standard deviation; a measurement of the rate of change of the one or more EEG signals; a measurement of a deterministic relationship, for example a linear trend, of the one or more EEG signals; a measurement of a change in phase of the one or more EEG signals; a measurement of the duration of the selection of the one or more EEG signals; a measurement of change point estimation of the associated one or more EEG signals, for example change in gaussian distribution or change in correlation between channels of the associated one or more EEG signals; and a measured rate of change of one or more of the preceding measurements. Optionally, either or both of: the measure of the one or more EEG signals of the first selection; and the measure of the one or more EEG signals of the second selection, comprises a measure of EEG signal strength. In an embodiment, an intra-segment ratio comprises a ratio comparing a first measure indicative of EEG signal strength associated with an instance of the second segment type and a second measure indicative of a duration of the instance of the second segment type, wherein the second segment type corresponds to interburst intervals (IB I).

[0017] The metric datapoints are optionally determined, at least in part, in accordance with an intra-segment comparison comprising one or more intra-segment ratios, wherein the, or each, intra-segment ratio is associated with one of either the first segment type or the second segment type. E.g., a particular intra-segment ratio is based on either: one or more segments of the first segment type and not the second segment type; or one or more segments of the second segment type and not the first segment type. For a particular metric datapoint the intra-segment comparison may comprise: for the, or each, intra-segment ratio: identifying an associated intra- segment selection comprising one of either: one or more instances of the first segment type temporally correlated with the particular metric datapoint; or one or more instances of the second segment type temporally correlated with the particular metric datapoint, and calculating a ratio of a first measure and a second measure of the one or more EEG signals of its associated intra-segment selection. For at least one intra-segment ratio, the intra-segment selection may comprise a single instance of either the first segment type or the second segment type. The method may comprise: determining at least a first intra-segment ratio and a second intra- segment ratio, wherein either one or both of: the intra-segment selection of the first intra- segment ratio may be associated with the first segment type and the intra-segment selection ofthe second intra-segment ratio may be associated with the second segment type; and the first measure of the intra-segment selection of the first intra-segment ratio may be different to the first measure of the intra-segment selection of the second intra-segment ratio.

[0018] At least one intra-segment comparison may comprise a first measure and / or a second measure selected from at least one of a measurement of the EEG signal central tendency of the associated one or more EEG signals, for example a signal mean, median or count; a measurement of the EEG signal strength of the associated one or more EEG signals, for example deviation from a central tendency such as signal variance or standard deviation; a measurement of the rate of change of the one or more EEG signals; a measurement of a deterministic relationship, for example a linear trend, of the one or more EEG signals; a measurement of a change in phase of the one or more EEG signals; a measurement of the duration of the selection of the one or more EEG signals; a measurement of change point estimation of the associated one or more EEG signals, for example change in gaussian distribution or change in correlation between channels of the associated one or more EEG signals; and a measured rate of change of one or more of the preceding measurements.

[0019] Optionally, the method comprises: segmenting the EEG data to identify instances of the first segment type and instances of the second segment type. The EEG data may be segmented using a preconfigured signal strength threshold. The EEG data may be segmented using a preconfigured change point detection algorithm. The change point algorithm may be configured to identify step changes in one or more of the following: a measurement of the EEG signal central tendency of the associated one or more EEG signals, for example a signal mean, median or count; a measurement of the EEG signal strength of the associated one or more EEG signals, for example deviation from a central tendency such as signal variance or standard deviation; a measurement of the rate of change of the one or more EEG signals; a measurement of a deterministic relationship, for example a linear trend, of the one or more EEG signals; a measurement of a change in phase of the one or more EEG signals; a measurement of the duration of the selection of the one or more EEG signals; a measurement of change point estimation of the associated one or more EEG signals, for example change in gaussian distribution or change in correlation between channels of the associated one or more EEG signals; and a measured rate of change of one or more of the preceding measurements. The segmenting may ensure an instance of the first segment type immediately precedes an instance of the second segment type and vice versa. The segmenting may require an identified change in the EEG data to persist for a predefined duration before identifying a boundary between twosegments. The segmenting may include identifying a transition region between adjacent segments that is not classified as either a first segment type or a second segment type.

[0020] According to another aspect of the present disclosure, there is provided a method for aiding the detection of drug effect using anaesthesia-induced patterns in an electroencephalogram (EEG), comprising: obtaining EEG data corresponding to one or more EEG signals of a patient under general anaesthesia measured using an EEG, wherein the one or more EEG signals are time-varying; identifying a development of a repeating first segment type and a repeating second segment type, wherein instances of the first segment type alternate with instances of the second segment type; and in response, outputting a discontinuity indication indicative of the presence of the repeating first segment type and a repeating second segment type in the EEG data.

[0021] According to another aspect of the present disclosure, there is provided a method for monitoring a patient using an electroencephalogram (EEG), comprising: obtaining EEG data corresponding to one or more EEG signals of a patient measured using an EEG, wherein the one or more EEG signals are time-varying; identifying a development of a repeating first segment type and a repeating second segment type, wherein instances of the first segment type alternate with instances of the second segment type; and in response, outputting a discontinuity indication indicative of the presence of the repeating first segment type and a repeating second segment type in the EEG data.

[0022] According to another aspect of the present disclosure, there is provided a method for aiding the detection of drug effect using anaesthesia-induced patterns in an electroencephalogram (EEG), comprising: obtaining EEG data corresponding to one or more EEG signals of a patient under general anaesthesia measured using an EEG, wherein the one or more EEG signals are time-varying and comprise at least a repeating first segment type and a repeating second segment type, wherein instances of the first segment type alternate with instances of the second segment type; determining a signal metric based on a comparison between instances of the first segment type and instances of the second segment type wherein the signal metric is time-varying and comprises a series of time ordered metric datapoints; and wherein the signal metric is a measurement of an effect of an anaesthetic drug regimen administered to the patient.

[0023] Optionally, the method further comprises: determining a signal metric based on a comparison between instances of the first segment type and instances of the second segment type wherein the signal metric is time-varying and comprises a series of time ordered metric datapoints; and wherein the signal metric is a measurement of a state of the patient’s brain.

[0024] According to another aspect of the present disclosure, there is provided a system for aiding the detection of drug effect using anaesthesia-induced patterns in an electroencephalogram (EEG), comprising: a processing module configured to interface with an electroencephalography (EEG), wherein the processing module is configured to implement the method of one or more of the previous aspects to thereby generate a signal metric associated with a patient based on one or more EEG signals measured using the EEG of the patient.

[0025] The system optionally comprises the EEG. The EEG may comprise processing means, and the processing module may be a functional computational module of the EEG. The EEG may comprise processing means and, in use, the processing module may be in data communication with the EEG.

[0026] The system optionally comprises a display configured to present a visual and / or textual representation of the determined signal metric. The display may also be arranged to present a visual and / or textual representation of one or more EEG signals, such that the representation of the signal metric may be presented simultaneously with the representation of one or more EEG signals. The signal metric may be generated and presented on the display in substantially real-time such that the signal metric represents a current effect of the general anaesthesia on the patient.

[0027] According to another aspect of the present disclosure, there is provided a computer program comprising executable program code configured to cause a processor to perform the method of the above aspect.

[0028] According to another aspect of the present disclosure, there is provided a nontransient computer readable storage media comprising the computer program of the previous aspect.

[0029] As used herein, the word “comprise” or variations such as “comprises” or “comprising” is used in an inclusive sense, i.e. to specify the presence of the stated features but not to preclude the presence or addition of further features in various embodiments of the invention.Brief Description of the Drawings

[0030] In order that the invention may be more clearly understood, embodiments will now be described, by way of example, with reference to the accompanying drawing, in which:Figure 1 shows a schematic representation of elements of an EEG monitoring system;Figure 2 shows a method for obtaining metric datapoints from EEG data;Figure 3 shows example EEG signals measured from a neonate under general anaesthesia;Figure 4 shows example transitions and segmentation of EEG data;Figure 5 shows an example metric visualised with anaesthetic dose;Figures 6A-6D show example experimental results; andFigures 7A-7C show example time varying experimental results.Description of Embodiments

[0031] Figure 1 shows an EEG monitoring system 10 according to an exemplary embodiment. The system 10 comprises an electroencephalography system (EEG) 11 interfaced with a processing module 12. The processing module 12 is configured to obtain EEG data corresponding to one or more EEG signals 24 (e.g., the one or more EEG signals 24 can correspond to one or more EEG channels) from the EEG 11. Optionally, the processing module 12 is configured to control operation of the EEG 11. For the purposes of the present disclosure, it is assumed that the one or more EEG signals 24 have been digitised when generating EEG data for use in the processing module 12. The processing module 12 is configured to process the obtained EEG data in order to generate a metric comprising a time series of metric datapoints. The processing module 12 comprises a memory for both data storage (e.g., via a non-volatile memory) and for providing a working memory for transient computing purposes (e.g., via a volatile memory) during operation of the processing module. The functionality of the processing module 12 described herein can be associated with computer program code (i.e., a computer program). The computer program is stored in the memory and the processing module 12 reads the computer program during operation. The computer program can be stored in a non-transient computer readable storage media.

[0032] In an embodiment, the processing module 12 can be implemented as a functional computational module of the EEG 11. For example, the EEG 11 itself comprises processing means for receiving and processing the EEG data, in which case the processing module 12 represents an additional function of the EEG 11. In another embodiment, the processing module 12 is implemented by a separate computer processor to the EEG 11 and receives EEG data via a data communication channel with the EEG 11. For example, the data communication channel can comprise a serial bus such as USB or a removal data storage such as a solid-state hard drive or a FLASH memory. In another example, the data communication channel can comprise a data network such as a local intranet and / or the Internet.

[0033] Also shown in Figure 1 is a display 13 interfaced with the processing module 12. The display 13 can also be interfaced with the EEG 11 such that both the EEG data and the metric generated by the processing module 12 can be presented simultaneously. Alternatively, the display 13 can be independent of another display of the EEG 11. The processing module 12 is also interfaced with a data store 14 for storing metric datapoints of the metric.

[0034] Figure 2 shows a patient monitoring method according to an embodiment. The patient can be receiving medical treatment, such as receiving a treatment acting at the brain. Examples include the patient being under general anaesthesia, being treated for seizures, in an induced coma, or suffering from hypothermia. The patient can have an illness of the brain, such as encephalopathy, including (but not restricted to) neonatal encephalopthies such as perinatal hypoxic ischaemic encephalopathy.

[0035] In a particular example, as assumed herein for the purposes of exemplifying embodiments (unless stated otherwise), the patient is a neonatal baby who is undergoing a medical procedure, as was the case in respect of the Experimental Results discussed with reference to Figures 6A-6D and Figures 7A-7C. The medical procedure as described herein involves the patient being put under general anaesthetic. The EGG 11 comprises a number of electrodes (not shown) which, in use, are arranged on the patient.

[0036] At step SI 00, the electrodes of the EEG 11 are coupled to the patient according to known procedures. For example, applicable clinical guidelines for EEG setup can be followed (which may optionally depend on an applicable category of the patient, such as a neonate).

[0037] At step S101, the EEG 11 measures one or more EEG signals 24 from the patient, and the processing module 12 obtains EEG data corresponding to the one or more EEG signals 24. Typically, the processing module stores the EEG data in an interfaced memory (e.g., a non-volatile memory). The one or more EEG signals 24 are time-varying. It may be that measurement of the one or more EEG signals 24 begins before the patient is administered the general anaesthetic. In an alternative, the EEG electrodes are only applied to the patient after administration of the general anaesthetic, which may be applicable to patients, such as neonates, where is difficult to apply to EEG electrodes when the patient is conscious. The EEG signals 24 are also measured after the general anaesthetic is first administered and typically for at least the duration of the medical procedure. In a preferred embodiment, the processing module 12 is configured to continuously obtain the EEG data during the medical procedure performed on the patient. In this case, step S101 and the following steps S102-S109 represent steps which are continuously performed during the medical procedure. In an embodiment, the processing module 12 is configured to perform post-procedure analysis on stored EEG data, in which casesteps S101-S109 can be understood as being performed once on the entire dataset. This latter embodiment can be utilised for research and forensic analysis purposes.

[0038] Typically, the one or more EEG signals 24 are subjected to a pre-processing procedure at step SI 02 in order to generate the EEG data, for example, for the removal of noise and artefacts from the one or more EEG signals 24. This step can include one or more of rereferencing (e.g., changing the reference channel(s) of the EEG data), artefact rejection, nonlinear noise rejection, and other standard EEG processing techniques.

[0039] At step SI 03, the (pre-processed) EEG data is segmented and classified. In an embodiment, the processing module 12 implements a signal classifier 30 as shown in Figure 1. Figure 3 shows an example of EEG signals 24 taken from a term-aged baby, during the maintenance phase of general anaesthesia. The EEG signals 24 show repeating burst periods 20 of relatively high amplitude (e.g., in microvolts (pV)) and relatively short duration separated in time by interburst intervals (IBIs) 21 of relatively low amplitude and relatively long duration). Therefore, a characteristic difference between the burst periods 20 and the IBIs 21 can be an amplitude difference (e.g., in microvolts (pV)). The example shown in Figure 3 are EEG signals 24 measured from a term neonate during inhalational general anaesthesia at Royal Children’s Hospital in Melbourne, Australia. This neonate did not have any neurological disorder or suspicion of such a disorder. As can be seen, in this example, the EEG signals 24 exhibit an anaesthesia-induced pattern of alternating characteristic segments (in the present case, boundaries between segments are visible as discontinuities in the EEG signals 24).

[0040] The signal classifier 30 is configured to segment the EEG data by identifying transitions 22 between burst periods 20 and IBIs 21 (and vice versa). The transitions 22 effectively demarcate individual segments of the EEG data, and the signal classifier 30 is configured to classify the identified segments as either a burst period 20 or an IBI 21. For example, in Figure 4 (which is the same EEG data as shown in Figure 3), the signal classifier 30 identifies eight transitions 22a-22h. Between the first transition 22a and the second transition 22b, the signal classifier 30 identifies a segment and classifies it as a burst period 20. Between the second transition 22b and the third transition 22c, the signal classifier 30 identifies a second segment and classifies it has an IBI 21. In the example of Figure 4, the transitions 22 are associated with marked changes in the amplitude of the one or more EEG signals 24. In this case, a preconfigured absolute amplitude threshold can be utilised to identify the transitions 22.

[0041] However, the inventor has found that other characteristics can differentiate burst periods 20 and the IBIs 21. The inventor has found that the EEG data can often be accurately segmented in situations where there is an insufficient voltage amplitude difference in the oneor more EEG signals 24 between burst periods 20 and IBIs 21. Therefore, an advantage of considering alternative indicators of transitions 22 may be that absolute voltage thresholds cannot account for relatively small voltage amplitude changes between burst periods 20 and IBIs 21. Therefore, relative changes in one or more measures of the EEG signals 24 can be utilised in place of absolute amplitude thresholds, thereby allowing detection of transitions 22 including in cases where absolute amplitude thresholds are not capable of detecting transitions 22.

[0042] For example, and without intending to be limiting, one or more of the following statistical measurements can be analysed by the segmentation module 30 in order to identify transitions 22 between the burst periods 20 and the IBIs 21 (and vice versa): a) a measurement of the signal strength of the associated one or more EEG signals 24, for example deviation from a central tendency such as signal variance or standard deviation; b) a measurement of the rate of change of the one or more EEG signals 24; c) a measurement of a deterministic relationship, for example a linear trend, of the one or more EEG signals 24; d) a measurement of a change in phase of the one or more EEG signals 24; e) a measurement of ensemble variance and / or global field power (GFP) between two or more EEG signals 24; f) a measurement of change point estimation of the associated one or more EEG signals 24, for example change in gaussian distribution or change in correlation between channels of the associated one or more EEG signals 24; and g) a measured rate of change of one or more of the preceding measurements.

[0043] In general, there may be multiple techniques available when measuring a property of the one or more EEG signals 24. Additionally, a sampling window can be defined for a particular measurement, for example defined by a sampling time period, such that an averaged or summed value is measured. For example, in terms of signal strength, an average peak-to- peak amplitude or average standard deviation over the entire sampling window is used as the measurement of the signal strength. The sampling window is therefore implemented as a moving window. The sampling window size should be sufficiently small (e.g., of short duration) to ensure identification of transitions between burst periods 20 and IBIs 21, for example, at most 1 / 10thof the expected duration of a burst period 20. The size of the sampling window can be a settable parameter of the system 10.

[0044] In each case, the statistical measurement typically shows a step change at a transition 22. The signal classifier 30 can therefore be configured to analyse the EEG data according to one or more of the statistical measurements and to identify statistically significant step changes in the one or more EEG signals 24. In an embodiment, one or more change point detection and estimation techniques can be utilised by the signal classifier 30 in order to identify transitions 22 between burst periods 20 and IBIs 21. Change point detection is the problem of finding abrupt changes in data (such as the EEG data) when a property of a time series changes; such changes may represent changes in the underlying data generating process or data states. Change point estimation describes the nature and degree of such changes. In the present case, the change point estimation can be based on one or more statistical measurements of the EEG data.

[0045] The signal classifier 30 can be configurable (e.g., via a user input) in order to adjust one or more signal analysis parameters. For example, a signal analysis parameter can correspond to a required size (absolute or relative) of a step change for a particular measurement to cause the signal classifier 30 to identify a transition 22. In a further example, for a particular measurement, a transition 22 from a burst period 20 to an IBI 21 can be associated with a different signal parameter than a transition 22 from an IBI 21 to a burst period 20.

[0046] In an embodiment, a step change for a particular statistical measurement can be determined to be an increase step (e.g., corresponding to a statistically significant increase in the value(s) of the statistical measurement) or a decrease step (e.g., corresponding to a statistically significant decrease in the value(s) of the statistical measurement). The classification of a particular segment as a burst period 20 or an IBI 21 then depends on both the particular measurement and the determination of the transition 22 as being associated with an increase step or a decrease step.

[0047] Considering the measurement of a signal strength and / or count of the one or more EEG signals 24, an increase step (i.e., an increase in the statistically measured mean or count) is expected to indicate a transition 22 from an IBI 21 to a burst period 20. Similarly, a decrease step (i.e., a decrease in the statistically measured mean or count) is expected to indicate a transition 22 from a burst period 20 to an IBI 21.

[0048] In an embodiment, the signal classifier 30 is configured to ensure that burst periods 20 alternate with IBIs 21 (and vice versa). For example, the signal classifier 30 can be configured to ensure that adjacent segments are not classified the same (e.g., two adjacent segments are not both classified as a burst period 20 or as an IBI 21).

[0049] In an example, the signal classifier 30 utilises ensemble variance as a statistical measurement. The ensemble variance corresponds to a measurement of the variance of instantaneous voltages measured across two or more (e.g., typically all) of the EEG signals 24. Typically, for neonates, there are at least four EEG signals 24, for example, four or six EEG signals 24 associated with a corresponding four or six EEG electrodes attached to the patient. However, other numbers of EEG signals 24 could be applicable (e.g., twelve), for example, depending on the particular EEG 11 utilised and the particular protocol followed. It has been found that burst periods 20 are characterised by an increase in the mean ensemble variance, measured with respect to a suitable measurement window, which can be different to the sampling window. The length of the measurement window can be settable parameter of the system 10. The required increase in mean ensemble variance required to identify a burst period 20 can be a settable parameter of the system. The required increase in mean ensemble variance can be expressed, for example, as a number (which can be an integer but is typically allowed to be a real number) of standard deviations above the mean ensemble variance. In one example, which was used when preparing the graphs of Figures 7A to 7C, the measurement window length was set at 3.5 seconds and the required increase in mean ensemble variance was set at 2 standard deviations. By identifying burst periods 20 in this manner, IBIs 21 are effectively identified as those portions of the EEG data not identified as burst periods 20.

[0050] In an embodiment, the signal classifier 30 implements a machine learning algorithm for segmentation and / or classification of EEG data. A supervised machine learning algorithm can be trained on previously acquired EEG data in the form of EEG training datasets in which segments corresponding to burst periods 20 and IBIs 21 are identified and labelled. For effective training, the EEG training datasets should include EEG data obtained in a variety of circumstances, such as a variety of ages of the patient category (in the case of neonates, different neonatal ages), a variety of general anaesthetic regimens, a variety of medical procedures, and at a variety of points during the medical procedures.

[0051] At step SI 04, a burst period selection is identified comprising one or more burst periods 20. In an embodiment, a single burst period 20 is selected per burst period selection. In another embodiment, two or more burst periods 20 can be selected for a particular burst period selection. In this latter case, it can be a requirement that the selected burst periods 20 represent a sequence of directly adjacent burst periods 20 (here, adjacency is on the basis of ignoring any intervening IBIs 21).

[0052] At step SI 05, an IBI selection, associated with the burst period selection, is identified comprising one or more IBIs 21. In an embodiment, a single IBI 21 is selected per IBI selection.In another embodiment, two or more IBIs 21 can be selected for a particular IBI selection. In this latter case, it can be a requirement that the selected IBIs 21 represent a sequence of directly adjacent IBIs 21 (here, adjacency is on the basis of ignoring any intervening burst periods 20).

[0053] Note that steps SI 04 and SI 05 are independent of one another and can therefore be undertaken in any order or simultaneously. This is emphasised in Figure 2 by showing the steps as occurring in parallel.

[0054] The burst period selection and the associated IBI selection are together considered a paired selection and are typically required to be temporally correlated. For example, in an embodiment in which a single burst period 20 is selected for each burst period selection and a single IBI 21 is selected for each IBI selection, the selected IBI 21 of a particular paired selection can be required to be directly adjacent to the selected burst period 20 of said paired selection (e.g., such that the selected burst period 20 and the selected IBI 21 of a particular paired selection represent a contiguous portion of the EEG data). In another example, in an embodiment in which two or more burst periods 20 are selected for each burst period selection and / or two or more IBIs 21 are selected for each IBI selection, the combination of burst periods 20 and IBIs 21 can be required to represent a contiguous portion of the EEG data.

[0055] At step SI 06, each selection pair is analysed in order to generate a metric datapoint associated with the selection pair. The metric datapoint is typically temporally correlated with the one or more burst periods 20 and the one or more IBIs 21 of the selection pair. That is, the metric datapoint is understood as related to the same period of time as covered by said one or more burst periods 20 and the one or more IBIs 21. The metric datapoint can be assigned a timestamp that is temporally correlated with the selection pair. The timestamp can be, for example, a single value such as equal to the beginning of the selection pair, a midpoint value, or the end of the selection pair.

[0056] In an embodiment, the analysis comprises comparing the one or more selected burst periods 20 to the one or more selected IBIs 21 of the selection pair. The comparison can involve comparing one or more statistical measures of the one or more burst periods 20 to one or more statistical measures of the one or more IBIs 21.

[0057] In an embodiment, one or more inter-segment comparisons are calculated for each metric datapoint. A particular inter-segment comparison can comprises a ratio of at least one measure of the one or more EEG signals 24 of the one or more selected burst periods 20 to at least one measure of the one or more EEGs signals 24 of the one or more selected IBIs 21 (or vice versa). One or more measures can, for example, correspond to a measure utilised by thesegmentation module 30 to identify transitions 22. However, at least one measure can correspond to a different measure to those used in to identify transitions 22.

[0058] Example measures that can be applied to the one or more selected burst periods 20 and / or the one or more selected IBIs 21 include: a) a measurement of the signal strength of the associated one or more EEG signals 24, for example deviation from a central tendency such as signal variance or standard deviation; b) a measurement of the rate of change of the one or more EEG signals 24; c) a measurement of a deterministic relationship, for example a linear trend, of the one or more EEG signals 24; d) a measurement of a change in phase of the one or more EEG signals 24; e) a measurement of the duration of the selection of the one or more EEG signals 24; f) a measurement of ensemble variance between two or more EEG signals 24; g) a measurement of change point estimation of the associated one or more EEG signals 24, for example change in gaussian distribution or change in correlation between channels of the associated one or more EEG signals 24; and h) a measured rate of change of one or more of the preceding measurements.

[0059] Depending on the embodiment, different measures of the one or more EEG signals 24 can be determined. The measures are applied to the EEG data associated with the associated one or more selected burst periods 20 or one or more selected IBIs 21, as appropriate. A sampling window can be associated with one or more of the measures, this can be the same sampling window as used for by the segmentation module 30. Alternatively, a different sampling window can be used. The particular sampling window can be a settable parameter of the system 10.

[0060] In one example, a measure corresponds to a measurement of the signal strength of the associated one or more EEG signals 24, for example the standard deviation, or equivalently a root mean square of the EEG signal. In another example, a measure corresponds to a measurement of central tendency of the associated one or more EEG signals 24, for example the mean, median or count. In another example, a measure corresponds to a measurement of the rate of change of the variance or standard deviation of the associated one or more EEG signals 24. In yet another example, a measure corresponds to a measurement of the duration of the associated one or more EEG signals 24.

[0061] In an embodiment, a first inter-segment comparison is calculated as a ratio of burst amplitude to IBI amplitude. Therefore, the first inter-segment comparison can be calculated by determining a measure of the amplitude of the one or more burst periods 20 of the burst period selection (“burst amplitude”) and a measure of the amplitude of the one or more IBIs 21 of the IBI selection (“IBI amplitude”). These measures can correspond to a determination of the variance or standard deviation of the one or more EEG signals 24 (or related, e.g., an RMS value) of the one or more EEG signals 24. The inter-segment comparison can then be determined by dividing the burst amplitude by the IBI amplitude.

[0062] The inventor has found that the mean variance of the one or more EEG signals 24 during an IBI 21 tends to remain relatively constant during administration of an anaesthetic. However, the mean variance of the one or more EEG signals 24 during a burst period 20 has been found to decrease with a reduction in anaesthesia dose (and to increase with an increase in anaesthesia dose). Therefore, the first inter-segment comparison can be expected to also decrease with a reduction in anaesthesia dose.

[0063] In an embodiment, a second inter-segment comparison is calculated as a ratio of IBI duration to burst duration. The IBI duration corresponds to the duration of the one or more selected IBIs 21. In the case of two or more selected IBIs 21, this can be a total duration of the combined IBIs 21 or an average duration of each IBI 21. Similarly, the burst duration corresponds to the duration of the one or more selected burst periods 20. In the case of two or more selected burst periods 20, this can be a total duration of the combined burst periods 20 or an average duration of each burst period 20. The second inter-segment comparison can be calculated by dividing the IBI duration by the burst duration.

[0064] The inventor has found that the burst duration tends to remain relatively constant during administration of an anaesthetic. However, the IBI duration has been found to decrease with a reduction in anaesthesia dose (and to increase with an increase in anaesthesia dose). Therefore, the second derived parameter can be expected to also decrease with a reduction in anaesthesia dose.

[0065] In an embodiment, one or more intra-segment comparisons are calculated for each metric datapoint. A particular intra-segment comparison can comprise a ratio of at least one first measure to at least one second measure of the one or more EEG signals 24 of either the one or more selected burst periods 20 or the one or more selected IBIs 21.

[0066] In an embodiment, a first intra-segment comparison is calculated as a ratio of burst amplitude to burst duration for the one or more selected burst periods 20. These can be determined as described in relation to the first and second inter-segment comparisons. As noted,the mean burst amplitude has been observed to decrease with reduction in anaesthesia dose (and vice versa) while the burst duration has been observed to be relatively constant with anaesthesia dose. Therefore, the first intra-segment comparison, when expressed as the burst amplitude divided by the burst duration, can be expected to also decrease with a reduction in anaesthesia dose.

[0067] In an embodiment, a second intra-segment comparison is calculated as a ratio of IB I duration to IBI amplitude for the one or more selected IBIs 21. Again, these can be determined as described in relation to the first and second inter-segment comparisons. As noted, the IBI duration has been observed to decrease with reduction in anaesthesia dose (and vice versa) while the IBI amplitude has been observed to be relatively constant with anaesthesia dose. Therefore, the second intra-segment comparison, when expressed as the IBI duration divided by the IBI amplitude, can be expected to also decrease with a reduction in anaesthesia dose.

[0068] Although not shown in Figure 2, an embodiment may utilise only intra-segment comparisons and therefore not require step SI 06 as such. Additionally, in a case where an embodiment only utilises one or more intra-segment comparisons associated with burst periods 20, then step SI 05 may not be required. Similarly, in a case where an embodiment only utilises one or more intra-segment comparisons associated with IBIs 21, then step S104 may not be required.

[0069] At step SI 07, a metric datapoint is generated in accordance with the one or more inter-segment comparisons and / or one or more intra-segment comparisons. For example, the metric datapoint can be a single value or a vector (or an array of vectors).

[0070] In an embodiment, a single segment ratio is generated at step SI 06 and its value is assigned to the metric datapoint as a single value.

[0071] In another embodiment, at least two segment ratios are generated and combined. In order to effectively combine two different segment ratios, a normalising factor can be applied to each (or equivalently, all but one of the segment ratios). The normalising factor is typically unique for each segment ratio. The normalising factor(s) can be predefined. In an embodiment, a user is enabled to set and / or adjust the normalising factor(s), as the optimal normalising factor(s) may vary depending on the patient.

[0072] In another embodiment, in which two or more segment ratios are generated at step SI 06, each segment ratio value is separately assigned to the metric datapoint thereby forming a vector. In a variation, at least one vector entry can be generated by combining two or more segment ratios as described above.

[0073] The metric datapoints generated at step SI 07 are typically stored in the memory of the processing module 12, preferably at least a non-volatile memory, at step SI 08.

[0074] In the case in which the metric datapoints are continuously generated in real-time during a medical procedure being performed on the patient, typically a graphical or textual representation is generated and presented on a display viewable by the anaesthetist, at step SI 09. This can be the same display arranged to present the EEG data itself, or the display may be different.

[0075] Figure 5 shows an example visual representation of metric datapoints comprising single values, derived from the combination of a first inter-segment comparison and second inter-segment comparison (see EEG-derived index line 48). The combined normalised anaesthesia dose (see anaesthesia dose line 47) represents effective hypnotic dose from the medications administered (this is not a validated approach and is provided for illustrative purposes only). The metric changes with anaesthesia dose. During onset, with a delay that is consistent with pharmacological onset. During offset, it does not exhibit the same delay.

[0076] In order to aid in exemplifying the invention, the disclosure herein has focused on the identification of burst periods 20 and IBIs 21, and comparisons of measurements between burst periods 20 and / or IBIs 21. It is anticipated that other repeating features may be present in EEG data captured from a patient under general anaesthetic. The burst periods 20 can therefore be understood as a particular example of a first segment type and the IBIs can be understood as a particular example of a second segment type. The signal classifier 30 can therefore be understood as being configured to identify segments based on identifiable transitions 22 within the EEG data, of which transitions 22 between burst periods 20 and IBIs 21 and vice versa) may be an example. The signal classifier 30 can also be understood as being configured to assign a segment type to each segment based on one or more predefined segment types (e.g., two segment types may be labelled a first segment type and a second segment type). Similarly, the described burst period selections can be understood as being a particular example of first segment type selections each comprising one or more (depending on the embodiment) segments of the first segment type. Also similarly, the described IBI selections can be understood as being a particular example of second segment type selections each comprising one or more (depending on the embodiment) segments of the second segment type.

[0077] A patient’s EEG may only begin to include distinguishable repeating segments (e.g., burst periods 20 and IBIs 21 distinguishable by the segmentation module 20) after the patient has received a sufficient dose of general anaesthetic. In an embodiment, the processing module 12 is configured to output metric values comprising a value indicative of the lack ofdistinguishing repeating segments while this is the case. For example, for a metric which tends to zero with decreasing dose of general anaesthetic, the processing module 12 can be configured to output a value “0” until distinguishable repeating segments are identified (at which point, the processing module 12 outputs calculated values for the metric datapoints as described herein).

[0078] In an embodiment, the processing module 12 is configured to output (e.g., to the display 13 and / or data storage 14) a discontinuity indication upon initially identifying distinguishable repeating segments, separately or alternatively to calculating the metric datapoints. The discontinuity indication can be useful to an anaesthetist during a medical procedure as the presence of a discontinuity (i.e., the presence of distinguishable repeating segments) may indicate the patient is unconscious.

[0079] Advantageously, an anaesthetist may use the described metric in several ways. In a particular example, the metric can be referred to as “Relative Anaesthesia Discontinuity” or “RAD”. It is anticipated to be useful for at least one of the following: a) Monitoring in real time the degree of effect at the brain of general anaesthesia; b) Assisting with determining if a dose of anaesthesia that a neonate has received has been sufficient to cause RAD; c) Predicting the time to wake up after general anaesthesia; and d) Investigating the relative potency of different medications for producing RAD.Example Experimental Results

[0080] Figures 6A to 6D show average measures over a number of selection pairs at different stages of anaesthesia of a patient. The data source is the same as that of Figure 3, although over a significantly longer overall duration. Figures 6A and 6C relates to selection pairs associated with portions of the EEG data associated with the patient under general anaesthesia at a relatively constant dose. Figures 6B and 6D relates to selection pairs associated with portions of the EEG data associated with washout of the general anaesthesia when the dose of anaesthesia was low. The centreline is median, the box is interquartile range, bars are 95% confidence interval, and the crosses are outliers.

[0081] Comparing Figures 6A and 6B, it is apparent that the amplitude of the EEG signals 24 during the burst periods 20 is noticeably higher when the patient is under general anaesthesia (Figure 6A) compared to the washout phase (Figure 6B). However, it is also apparent that there is no significant change in amplitude of the IBIs 22. Similarly, comparingFigures 6C and 6D, it is apparent that the duration of the IBIs 21 is longer when the patient is under general anaesthesia (Figure 6C) compared to the washout phase (Figure 6D).

[0082] Figures 7A-7C show example experimental data associated respectively with three different patients (each being a neonate). In preparing each figure, the EEG data has already been segmented, such that it is associated with alternating burst periods 20 and IBIs 21. The burst periods 20 and IBIs 21 can be considered as a single time series of alternating burst periods 20 and IBIs 21, or as two separate time series comprising a time series of burst periods 20 and another time series of IBIs 21. In the latter case, the timestamps for the burst periods 20 alternate with the timestamps for the IBIs 21.

[0083] In each figure, the topmost panel A shows the median amplitude of ninety consecutive periods in a sliding window (i.e., burst amplitude line 40a-c shows the median amplitude of ninety consecutive burst periods 20 using a sliding window and IBI amplitude line 41a-c shows the median amplitude of ninety consecutive IBIs 21 using the same sliding window). Also in each figure, the second-top panel B shows the median length of ninety consecutive periods in a sliding window (i.e., burst duration line 42a-c shows the median duration of ninety consecutive burst periods 20 using a sliding window and IBI duration line 43a-c shows the median duration of ninety consecutive IBIs 21 using the sliding window). Generally, of course, other smoothing techniques and window lengths can be utilised as desired.

[0084] In each figure, the third-top panel C shows the administered anaesthesia dose to the patient over time. The patients associated with Figures 7A and 7B were administered both propofol 45a-b and sevoflurane 46a-b, whereas the patient of Figure 7C was administered sevoflurane 46c only.

[0085] Finally, in each figure, the bottommost panel D shows changes in an intra-segment comparison (being a ratio in this case) over time (see intra-segment comparison lines 44a-c), where the intra-segment comparison comprises a ratio of IBI 21 amplitude to IBI 21 duration (normalised to the range [0,1]). In effect, this corresponds to the ratios of the IBI amplitude line 41a to IBI duration line 43a (Figure 7A), IBI amplitude line 41b to IBI duration line 43b (Figure 7B), and IBI amplitude line 41c to IBI duration line 43c (Figure 7C). The intra- segment comparison shows a negative correlation with anaesthesia dose to some degree, suggesting a potential role in the measurement of anaesthesia drug effect for such patients.

[0086] Considering the particular intra-segment comparison of Figures 7A-7C, values closer to zero reflect longer and smaller amplitude IBIs 21, which are typically seen with deeper anaesthesia. These results were analysed using Spearman’s correlation (also known as“Spearman's rank correlation coefficient”) in order to assess the strength of monotonic association. In each case, the correlation was statistically significant with p<0.01.

[0087] The correlation between sevoflurane concentration and the intra-segment comparison was negative in all three participants: a) Participant 1 : p = -0.83 b) Participant 2: p = -0.62 c) Participant 3 : p = -0.53

[0088] This indicates that as the dose of sevoflurane increased, the intra-segment comparison tended to fall, consistent with deeper levels of anaesthesia. The strength of this relationship varied between individuals but was consistently moderate to strong.

[0089] Only the first two participants (i.e., see Figures 7A and 7B) received propofol. The first received a single bolus at induction; the second received an infusion over time while sevoflurane gradually washed out. a) Participant 1 : p = -0.02 b) Participant 2: p = -0.42

[0090] In the first case (a), the correlation was negligible, likely reflecting the limited duration of exposure. In the second (b), a clear, moderate negative correlation was observed, though weaker than that seen with sevoflurane.

[0091] To explore the relationship between total anaesthetic exposure and this particular intra-segment comparison, a simple combined dose estimate was used. Correlations between this combined dose and the intra-segment comparison were as follows: a) Participant 1 : p = -0.75 b) Participant 2: p = -0.48 c) Participant 3 : p = -0.53

[0092] For the second participant, where both drugs were given during overlapping periods, the combined dose showed a somewhat stronger association than propofol alone (-0.48 vs - 0.42), though not stronger than sevoflurane alone (-0.62). This suggests the additive approach may be useful, at least in some cases.Applicability to other patients

[0093] Although the present disclosure is primarily focused on neonate patients, embodiments described herein may be applicable to other classes of patient, for example, where discontinuities as described are expected to form when a patient is under general anaesthetic. For example, discontinuities may be expected in neuroatypical patients, patients with encephalopathy, patients with acquired head injury, patients with concurrent pharmacotherapy (such as anticholinergics or other neuroactive medications), the elderly, and also healthy older children and adults.

[0094] Further modifications can be made without departing from the spirit and scope of the specification.References[1] Stolwijk LJ, Weeke LC, Vries LS de, Herwaarden MYA van, Zee DC van der, Werff DBM van der, et al. Effect of general anesthesia on neonatal aEEG — A cohort study of patients with non-cardiac congenital anomalies. Pios One. 2017; 12(8):e0183581.

Claims

Claims:

1. A method for aiding the detection of drug effect using anaesthesia-induced patterns in an electroencephalogram (EEG), comprising: obtaining EEG data corresponding to one or more EEG signals of a patient under general anaesthesia measured using an EEG, wherein the one or more EEG signals are timevarying and comprise a time series of segments comprising at least a repeating first segment type and a repeating second segment type, wherein instances of segments corresponding to the first segment type alternate with instances of segments corresponding to the second segment type; and determining a signal metric comprising a series of time ordered metric datapoints, wherein each metric data point is determined based on: a comparison between at least two segments temporally associated with the metric point; and / or a comparison between at least two different measures of at least one segment temporally associated with the metric point, wherein the signal metric is a measurement of an effect of an anaesthetic drug regimen administered to the patient.

2. The method of claim 1, wherein the patient is a neonate.

3. The method of claim 1 or claim 2, wherein the first segment type corresponds to burst periods, and the second segment type corresponds to interburst intervals (IB I).

4. The method of any one of claims 1 to 3, wherein the metric datapoints are determined, at least in part, in accordance with an inter-segment comparison, wherein for a particular metric datapoint the inter-segment comparison comprises: identifying a first selection comprising one or more instances of the first segment type being temporally correlated with the particular metric datapoint; identifying a second selection comprising one or more instances of the second segment type being temporally correlated with the particular metric datapoint; and comparing the first selection to the second selection.

5. The method of claim 4, wherein: the first selection comprises a single instance of the first segment type; and / or the second selection comprises a single instance of the second segment type.

6. The method of claim 4 or claim 5, wherein the one or more instances of the first segment type of the first selection and the one or more instances of the second segment type of the second segment are associated with a temporally contiguous portion of the EEG data.

7. The method of any one of claims 4 to 6, wherein determining the particular metric datapoint comprises: determining at least one inter-segment ratio by taking a ratio of a measure of the one or more EEG signals of the first selection and a measure of the one or more EEG signals of the second selection.

8. The method of claim 7, wherein determining the particular metric datapoint comprises: determining at least a first inter-segment ratio and a second inter-segment ratio, wherein at least one of the measure of the one or more EEG signals of the first selection and the measure of the one or more EEG signals of the second selection of the first inter-segment ratio is different to at least one of the measure of the one or more EEG signals of the first selection and the measure of the one or more EEG signals of the second selection of the second inter-segment ratio, wherein determining the particular metric datapoint comprises normalising and combining at least the first inter-segment ratio and the second inter-segment ratio.

9. The method of claim 7 or claim 8, wherein at least one of: the measure of the one or more EEG signals of the first selection; and the measure of the one or more EEG signals of the second selection, comprises a measure of EEG signal strength.

10. The method of any one of claims 1 to 9, wherein the metric datapoints are determined, at least in part, in accordance with an intra-segment comparison comprising one or more intrasegment ratios, wherein the, or each, intra-segment ratio is associated with one of either the first segment type or the second segment type.

11. The method of claim 10, wherein for a particular metric datapoint the intra-segment comparison comprises: for the, or each, intra-segment ratio: identifying an associated intra-segment selection comprising one of either: one or more instances of the first segment type temporally correlated with the particular metric datapoint; or one or more instances of the second segment type temporally correlated with the particular metric datapoint, and calculating a ratio of a first measure and a second measure of the one or more EEG signals of its associated intra-segment selection.

12. The method of claim 11, wherein, for at least one intra-segment ratio, the intra- segment selection comprises a single instance of either the first segment type or the second segment type.

13. The method of claim 11 or claim 12, comprising: determining at least a first intra-segment ratio and a second intra-segment ratio, wherein either one or both of: the intra-segment selection of the first intra-segment ratio is associated with the first segment type and the intra-segment selection of the second intra-segment ratio is associated with the second segment type; and the first measure of the intra-segment selection of the first intra-segment ratio is different to the first measure of the intra-segment selection of the second intra- segment ratio.

14. The method of any one of claims 11 to 13 when dependent on claim 3, comprising an intra-segment ratio comprising a ratio comparing a first measure indicative of EEG signal strength associated with an instance of the second segment type and a second measure indicative of a duration of the instance of the second segment type.

15. The method of any one of claims 1 to 14, comprising: segmenting the EEG data to identify instances of the first segment type and instances of the second segment type.

16. The method of claim 15, wherein the EEG data is segmented using a preconfigured change point detection algorithm.

17. The method of claim 15 or claim 16, wherein the segmenting ensures an instance of the first segment type immediately precedes an instance of the second segment type and vice versa.

18. The method of any one of claims 15 to 17, wherein the segmenting requires an identified change in the EEG data to persist for a predefined duration before identifying a boundary between two segments.

19. The method of any one of claim 15 to 18, wherein the segmenting includes identifying a transition region between adjacent segments that is not classified as either a first segment type or a second segment type.

20. A method for aiding the detection of drug effect using anaesthesia-induced patterns in an electroencephalogram (EEG), comprising: obtaining EEG data corresponding to one or more EEG signals of a patient under general anaesthesia measured using an EEG, wherein the one or more EEG signals are timevarying; identifying a development of a repeating first segment type and a repeating second segment type, wherein instances of the first segment type alternate with instances of the second segment type; and in response, outputting a discontinuity indication indicative of the presence of the repeating first segment type and a repeating second segment type in the EEG data.

21. A method for monitoring a patient using an electroencephalogram (EEG), comprising: obtaining EEG data corresponding to one or more EEG signals of a patient measured using an EEG, wherein the one or more EEG signals are time-varying; identifying a development of a repeating first segment type and a repeating second segment type, wherein instances of the first segment type alternate with instances of the second segment type; and in response, outputting a discontinuity indication indicative of the presence of the repeating first segment type and a repeating second segment type in the EEG data.

22. The method of claim 21, further comprising: determining a signal metric based on a comparison between instances of the first segment type and instances of the second segment type wherein the signal metric is timevarying and comprises a series of time ordered metric datapoints; and wherein the signal metric is a measurement of a state of the patient’s brain.

23. A system for aiding the detection of drug effect using anaesthesia-induced patterns in an electroencephalogram (EEG), comprising: a processing module configured to interface with an electroencephalography (EEG), wherein the processing module is configured to implement the method of any one of claims 1 to 22 to thereby generate a signal metric associated with a patient based on one or more EEG signals measured using the EEG of the patient.

24. A system as claimed in claim 23, comprising the EEG.

Images