AU745602B2 - Interactive electrophysiological measurement of event related neural responses - Google Patents

Description

P/00/001 Regulation 3.2

AUSTRALIA

Patents Act 1990 STANDARD SPECIFICATION Invention Title: "INTERACTIVE ELECTROPHYSIOLOGICAL MEASUREMENT OF EVENT-RELATED NEURAL RESPONSES" The invention is described in the following statement: 6***e 666* 6* 6* "INTERACTIVE ELECTROPHYSIOLOGICAL MEASUREMENT OF EVENT-RELATED NEURAL RESPONSES" FIELD OF THE INVENTION This invention relates to the area of electrophysiological measurement and analysis in the field of neurosciences. More particularly it relates to a method for acquiring event-related data representative of physiological activity in the brain of a subject, for example event-related potentials (ERPS) obtained from electroencephalographic signals or event-related fields (ERFS) obtained from magnetoencephalographic signals.

BACKGROUND TO THE INVENTION Electroencephalography is a non-invasive investigative technique which finds application, inter alia, in the diagnosis, prognosis and treatment of mental illness. A conventional electroencephalogram is (EEG), which is a recording of a time varying potential corresponding to brain electrical activity, can be detected and recorded using electrodes placed in proximity to the scalp of a human subject. The complementary technique of magnetoencephalography (MEG) similarly measures the magnetic signals emitted by the brain.

Existing EEG or MEG techniques record the brain's electromagnetic activity without taking account of the mental state of the subject or any external stimuli present. Other applications for these techniques include identification of individuals and cybernetic control systems.

Event related potentials (ERPS) reflect the variations in brain electrical activity associated with the occurrence of some definable event or psychological process, such as a movement or an external stimulus. Similarly event related fields (ERFS) reflect the variations in brain 5 magnetic activity associated with such an event. ERPs are conventionally extracted from EEG data, in which they are embedded, by a procedure known as "averaging". This procedure may involve the arithmetic averaging or summing of a series of EEG samples which consistent with the event are cumulative, whilst the temporally random potentials tend to cancel one another thus approaching a minimum value.

ERPs may be broadly categorised according to whether they originate in the brain or are of non-cerebral origin. ERPs originating in the brain include: sensory evoked potentials (EPs), obtained when the subject is presented with a brief randomly occurring external stimulus; long latency potentials (eg. P300), associated with ascertaining the meaning of the stimulus, ie. cognitive function; motor potentials, which are associated with the subject's movement; and slow potentials, such as the contingent negative variation (CNV), which are elicited under certain psychological conditions.

In contrast, non-cerebral ERPs originate in sources such as muscles or the eye and must be removed from the EEG data when studying brain electrical activity. It is known, for example, to simultaneously record an electro-occulogram from the eye(s) of a subject in order to facilitate rejection of the eye movement artefacts from EEG data.

ERP studies are used for a wide variety of purposes in relation to diagnosing mental dysfunction. Common examples include use in detecting abnormalities in multiple sclerosis SRi:; I i (MS) or schizophrenia.

In ERP studies, the important assumptions are identical stimuli, reproducible responses to a particular randomly applied stimulus and random stationary background EEG activity which is not correlated with the ERP. Several authors have questioned the assumption of independence of the ERP and background EEG, including Basar et al. 1984 new approach to endogenous Event Related Potentials in man: Relation between EEG and P300 wave", Intern.

J Neuroscience, 24 (Suppl. 1) 1-21) and Squires Donchin 1976 ("Beyond averaging: The use of discriminant functions to recognise event related potentials elicited by auditory stimuli", Electroenceph. Clin. Neurophysiol., 41 449-459).

Various models have described the endogenous potentials, for example the P300 wave, as a dynamic change in the EEG activity already present, including Stampfer 1988 ("An analysis of preparation and response activity in P300 experiments in humans", in Basar and Melnechuk Dynamics of sensory and cognitive processing by the brain (1st Edition), Springer- Verlag, Berlin, 276-286) and Wright et al. 1990 ("Inverse filter computation of the neural impulse giving rise to the auditory evoked potential", Brain Topography, 2 293-302). These previous studies selected EEG data retrospectively for further analysis. However, the P300 and long latency potentials in general can be affected by a large number of factors other than background EEG, such as the ordering of preceding stimuli, the inter stimulus interval (ISI), refractory periods and habituation effects. Accordingly, such retrospective selection leads to difficulties in obtaining a data set balanced for all factors other than the variable under test.

Australian Patent Application No. 60894194 in the name of Hudspeth discloses a system for acquiring EEG waveforms from a subject, in particular EPs are evoked for the purpose of training the subject to effect cybernetic control of a device or for the purpose of diagnosing brain dysfunction in the subject. In either application, the EPs are correlated with sets of pre-existing EPs, known as templates, stored within the system.

oS .In the case of a cybernetic control system, when a high correlation with a particular template is determined, the system issues a predetermined command to the device, which i command corresponds to the template. When used for diagnosis purposes, the subject is presented with sets of stimuli and the EPs generated in response are extracted. The EPs may then be compared with templates recorded from a population of normal subjects using multidimensional vector analysis methods.

The Hudspeth specification recognises that the actual EPs are superimposed on a spontaneous potential (SP) normally obtained from an alert and resting subject. Thus the model used in the invention subtracts a background component approximating the SP from the measured signal in order to isolate the unique EPs corresponding to the transient brain electrical activity occurring in response to the stimuli. However, this method assumes that the spontaneous potential will be similar for every subject and takes no account of the actual brain wave state of the subject immediately prior to the application of the stimuli.

United States Patent No. 5215086 (P086) in the name of Terry, Reese, Wemrnicke, Baker Ross (Cyberonics) discloses a method for detecting an electrophysiological entity, and applying a stimulus for the purpose of controlling migraines. United States Patent No. 4702254 (P254) in the name of Zabara discloses a similar device for the treatment of epilepsy. The entity being detected in these patents, and hence the process itself, differs substantially from that ~ii j~ri.u-~ ri i detected in the disclosed invention. The difference is between a "disease" entity which may have a qualitative, poorly specified corellate in brain activity (eg inter ictal epileptiform activity, or the "desynchronised" EEG state), and the physical brainwave state itself (quantitatively defined)..

In patents P086 and P254, the process begins with the recognition of a complex physiological event (epilepsy or migraine). Descriptions of "aura, paroxysmal activity, desynchronised activity, and circadian programming, exemplify the character of an entity that is not precisely localised in time and space. The brain wave state as used herein, and as described by Lehman et al. 1987 ("EEG alpha map series: brain micro-states by space orientated adaptive segmentation", Electroenceph. Clin. Neurophysiol., 67 271-288.), however, represents the precise combination of voltage and location of cortical activity immediately preceding the stimulus. While, in a limiting sense, a physiological event may be associated with a brain wave state (eg a single spike that initiates a rage attack), and a brain wave state could theoretically describe a physiological event (eg precisely specifying three seconds of spike and waves for an inter ictal episode), this is not the usual situation. As one skilled in the electroencephalographic art, it is not obvious that a complex physiological event, and a brain wave state are the same entities.

This difference is supported by the pattern recognition techniques employed In patents P086 and P254. Spectral analysis (FFT) is not capable of, nor intended for, recognising the short time periods associated with brain wave states.

The patents P086 and P254, are also different to the disclosed process, in the handling of the stimulus. The timing with respect to brain wave states is not a consideration in patents P086 and P254 (eg a train of stimuli cannot be considered to have an exact instant of application), whereas in the disclosed process, a stimulus is delivered immediately following the instant of pattern recognition. This is consistent with responding to (and presumably responding within) a S.i physical brain wave state, rather than responding to and within a physiological state. As one skilled in the electroencephalographic art, it is not obvious that these two different response S processes are linked.

oo United States Patent No. 5325862 in the name of Lewis Ryan-Jones discloses a personal identification and impairment is assessment system which uses brain activity patterns.

Samples of individual brain activity patterns, acquired using EEG or MEG techniques, are stored in a database to distinguish between normal and impaired brain states in any particular individual. The system is characterised by the use of evoked responses and neural network analysis of the resulting brain activity patterns for the purpose of identifying individuals or assessing brain impairment.

The US specification notes that conventional EEG and MEG signals may show variations beyond individual subject variation and that, accordingly, special techniques may be necessary to stabilise activity patterns. It is observed that brain activity can be stabilised by strict control of the conditions under which brain responses are generated, leading to the requirement for ERP techniques. It is concluded that whilst RPs vary considerably amongst individuals, there is stability within individuals over time given the same stimulus.

Furthermore, neural network analysis of the ERP data is seen as being something of a i r 1 :lr.l. -r panacea in respect of the sensitivity of traditional statistical techniques to nonlinear characteristics found in biological data. Whilst the sensitivity of ERP data may be adequate for personal identification, other researchers have found conflicting results (as summarised in Goodin, below) in the use of ERPs for clinical purposes. The use of neural network techniques also suffers from the problem of not being as widely and well understood as conventional statistical techniques in this field.

A number of authors have previously described effects of pre-stimulus brain wave state, specifically alpha waves, on the ERP. Jasiukaitis Hakerem 1988 ("The effect of pre-stimulus alpha activity on the P300", Electroenceph. Clin. NeurophysioL, 25 (Suppl. 2) 157-65) found that larger amplitude P300s were obtained in the high alpha ERP. Stampfer 1988 (supra) reports on an increase in alpha phase alignment with the onset of stimulus, but does not correlate this increase with resulting averaged amplitude. Jansen Brandt, 1991 ("The effect of the phase of pre-stimulus alpha activity on the averaged visual evoked potential", Electroenceph. Clin.

Neurophysiol., 80 241-250. carry out this correlation of alpha phase with amplitude for a visual evoked response.

Goodin 1990 ("Clinical utility of long latency 'cognitive' event related potentials (P3): the pros", Electroenceph. Clin. NeurophysioL, 76) analysed certain patient studies conducted during the 1980s, and found that there were discrepancies amongst the studies regarding the sensitivity of the P3 (or P300). This author observed that this may relate to differences in the i: method of eliciting the P3 response, variability in the severity of dementia among the patients studied and the fact that some patients do not generate a recognisable or reproducible response.

Furthermore inattention can, even in normal subjects, result in either a small or absent P3 response. Goodin concludes that a high rate of absent or non-reproducible P3 responses will detract from its clinical utility and that this rate differs widely between investigators.

Other relevant prior art known to use techniques similar to ERP exists in the fields of biofeedback, photic driving, "conditioning stimuli" and "self-generated ERPS". Biofeedback in the ERP context is a field in which a feedback stimulus is applied after a standard ERP has been S generated and analysed.

Photic driving or "entrainment" is an EEG related field in which a regular repetitive stimulus is applied to produce a desired periodic brain wave state, irrespective of the preexisting brain wave state. This entrained (usually alpha frequency) signal provides general information about the response of a periodic brain waveform to a regular periodic stimulus. The information provided during entrainment is conceptually different to the single brain wave response to a single stimulus form of information provided by an ERP. The information is distributed over many cycles of the response, is limited to characteristics of a given periodic signal, and most importantly the information is limited to brain wave states which have been shown to respond to photic driving, such as alpha waves. A refinement to this method is to use information from an existing periodic waveform to synchronise the stimulating waveform with the brain waves of the subject, such as described in United States Patent No. 5241967 in the names of Yasushi and Saito.

Self-generated ERPs are event-related potentials generated is in response to the subjects own stimulus marker, typically a button press. No external stimulus is provided, and the subject is required to initiate the stimulus when in a selected mental state. Mental state, however, is not necessarily the same as brain wave state and is dependent on a subjective judgement by the C1I=l =li'--iBL 1 ;;.ii-az-r r r patient. In conditioning stimulus paradigms, a conditioning stimulus (CS) is applied in order to force the subject into a required mental state before the application of the operative imperative stimulus (IT).

OBJECT OF THE INVENTION It is an object of the present invention to provide a method for acquiring event-related data, representative of physiological activity in the brain, which ameliorates or overcomes at least some of the problems associated with the prior art.

It is another object of the present invention to provide a method for acquiring eventrelated potential (ERP) data and/or event related potential (ERF) data which is able to take account of the existing brain wave state of a subject during acquisition.

Further objects will be evident from the following description.

DISCLOSURE OF THE INVENTION In one form, although it need not be the only or indeed the broadest form, the invention resides in a method for acquiring event related data, representative of physiological activity in the brain of a subject, said method including the steps of: initially monitoring the subject's brain wave state; comparing the monitored brain wave state with preselected criteria; applying a selected stimulus to the subject when the brain wave state substantially meets the preselected criteria, otherwise returning to the monitoring step; recording the brain wave activity of the subject subsequent to the application of the stimulus; and processing the recording of brain wave activity to extract the event related data for further analysis.

Preferably the step of monitoring the subject's brain wave state includes the step of Speriodically sampling spontaneous potential(s) sensed at one or more scalp locations of the subject.

Alternatively, the step of monitoring the subject's brain wave state includes the step of periodically sampling spontaneous field(s) sensed proximate one or more scalp locations of the o subject.

In preference, the step of monitoring brain wave state of the subject includes the step of determining the nature of background brain wave activity, from the sensed spontaneous potentials or fields, in real time.

Preferably the preselected criteria in the comparing step are representative of a desired brain wave state.

Suitably the preselected criteria include a selected threshold amplitude, frequency distribution and/or wave shape of background brain wave activity.

The preselected criteria may also include a pattern of brain wave activity selected in order to investigate a mental dysfunction.

Ic; i:; The step of determining the nature of background brain wave activity may involve fast Fourier transform, autocorrelation or template matching techniques.

The step of applying the stimulus may be selected from applying one or more of an auditory, visual, olfactory, gustatory, tactual, somatosensory or electromagnetic stimuli.

Preferably the step of recording the brain wave activity of the subject includes the step of recording one or more brain wave states of the subject.

Suitably, the step of recording the brain wave activity of the subject includes the step of recording the occurrence of the stimulus applied to the subject.

The step of recording the occurrence of the stimulus may include the steps of recording the time, duration and/or type of stimulus applied to the subject.

Preferably the step of recording said one or more brain wave states of the subject includes the step of periodically sampling and is recording potential(s) sensed at said scalp locations of the subject.

ooincludes the step of periodically sampling and recording field(s) sensed proximate said scalp locations of the subject.

Preferably steps and of the method are repeated for a predetermined number of cycles in order to assemble a plurality of brain wave activity recordings in response to said stimulus.

If required, the step of applying a preselected stimulus may include the step of selecting the stimulus, from a range of available stimuli, in accordance with the brain wave activity of the subject recorded in a previous cycle.

o The same stimulus may be repeatedly applied to the subject in succeeding cycles or may e be varied after a predetermined number of cycles.

If required, the step of applying the stimulus may be carried out when the brain wave state substantially meets preselected criteria or after a predetermined time delay.

In a second more particular form, the invention resides in a method for acquiring event related data, representative of physiological activity in the brain of a subject, said method including the steps of: initially monitoring the subject's brain wave state; comparing the monitored brain wave state with criteria representative of a brain wave state which is selected for investigating a mental dysfunction; applying a stimulus to the subject when the monitored brain wave state meets the representative criteria, otherwise returning to the monitoring step; recording the brain wave state presented by the subject in response to application of the stimulus; repeating steps to above in order to record a plurality of brain wave state responses; and tv-7777, processing the recording of plural brain wave state responses to extract the event related data for assisting in the investigation of the mental dysfunction.

Preferably, the comparing step employs syntactic analysis techniques in order to compare the monitored brain wave state to the representative criteria.

The term "brain wave state" as used in this document refers to spatio-temporal patterns of brain physiological activity, such as the patterns of brain electrical activity revealed by an EEG recording, and/or patterns of brain magnetic activity as revealed by an MEG recording.

These patterns are similar in concept to the microstates as described by Leeman et al.

1987 ("EEG alpha map series: brain micro-states by space orientated adaptive segmentation", Electroenceph. Clin. Neurophysiol., 67 271-288.), although the specific implementation may vary. The brain wave states as used herein are preferably less then one second in is duration although longer states may be useful for particular applications, for example in detecting drowsiness.

BRIEF DETAILS OF THE DRAWINGS To assist in understanding the invention preferred embodiments will now be described with reference to the following figures in which: of FIG 1 is a schematic diagram of a system operating in accordance with an embodiment •of the invention; FIG 2 is a flow chart illustrating key steps in the method of a first embodiment; 9 FIG 3 is a plot of the P300 amplitude comparing ERP recordings made using a prior art technique with ERP recordings made in accordance with the method of the first embodiment; FIG 4 is a plot of medication response ratio of prior art ERP recordings in an ADHD investigation; and FIG 5 is a plot of recording technique response ratio, comparing prior art ERP responses to ERP responses recorded in accordance with a second embodiment.

O 9• DETAILED DESCRIPTION OF THE DRAWINGS Whilst the "interactive" ERP (IERP) method of the invention can be readily applied to many types of electrophysiological investigation of the brain, the particular system illustrated in FIG 1 is arranged for the conduct of an investigation into the P300 peak generated in an auditory oddball paradigm. The P300 peak is defined as the largest positive peak occurring at the Pz scalp location in the period from 280 to 500 milliseconds (ms) from the application of the stimulus, see Polich et al. 1985 ("Effects of age on the P300 component of the event related potential from auditory stimuli: Peak definition, variation and measurement", Journal of Gerontology, 40(6) 721-726) and see also Polich 1991 ("P300 in clinical applications", Am. J.

EEG Technol, 31 201-231) in relation to the standard auditory oddball paradigm.

The investigation, which is merely an exemplary application for the method of the embodiment, compares the effect on the amplitude of the P300 peak of the standard ERP method, which does not take account of pre-stimulus brain electrical activity, with that of the IERP method of the embodiment which initiates the stimulus when brain electrical activity meets pre-selected criteria. The selection of the threshold level of activity is an important part f:- rbl of the investigation. These threshold values were selected to be comparable with previous studies in which alpha dependent ERPs were generated retrospectively, including those by Jansen Brandt 1991 (supra) and Jasiukaitis Hakerem 1988 (supra).

The system 1 0 includes an array of nineteen (1 9) electrodes (not shown) disposed in substantially the standard 10-20 arrangement at respective sites on the scalp of a subject 1 1.

The Oz channel is used for is recording an electro-occuiogram (from the left eye) rather than as an occipital channel due to the limited number of channels available with the recording equipment 14, a Biologic Brain Atlas 1I. All electrode sites are referenced to earlobes 12 of the subject 1 1, thus twenty-one (21) analogue signal lines 13 feeding into the recording apparatus 14.

The electrical potential generated by the electrophysiological activity of the brain and sensed by the electrodes is very small. Accordingly the output from each electrode, referred to as a signal, must be amplified and filtered to enable subsequent display and processing. The signals from the electrodes feed to the amplification and filtering stage 15. All signal channels have a nominal gain of 20,000, a lower band pass dB attenuation) of 1 Hertz (Hz) and an upper band pass dB attenuation) of 30 Hz. The amplified and filtered signals are then each subject to an analogue-to-digital conversion stage in a recording computer, which 9 samples selected EEG signal channels at 128 Hz. The conversion produces an EEG data stream see of 8 bit bytes representative of the amplitude of the signals on respective channels. The recording apparatus 14 also includes a display means 16 for viewing the EEG waveforms and first data storage means 17 for storing EEG data.

A monitoring computer 19 controls the process by repeatedly sampling and analysing a selection of the amplified and filtered EEG signals 18 and then initiating synchronising signals 20, 22 based on this analysis. The EEG signals 18 are subject to separate A/D conversion in the monitoring computer 19. The IERP method of the embodiment is based on a real time spectral 1. analysis of the EEG stream. The monitoring computer 19 continuously samples a data window on the Oz channel and performs a fast Fourier transform (FFT). Whenever both an relative and absolute thresholds of the transformed signal (indicative of alpha brain wave activity) are exceeded, the monitoring computer generates a collection synchronising pulse 20. The data window is of 128 points (1 second) duration, and is shifted 32 points (0.25 seconds) with each FFT analysis. A Harmnning window is applied in the FFT process.

o: The threshold FFT values were selected to be comparable with previous studies in which alpha dependent ERPs were generated retrospectively. Jansen Brandt 1991 (supra) used an absolute power threshold to determine whether alpha activity was present or not. For the present study, a relative power threshold was used, calculated for each section of background EEG as it was monitored. A relative power value of less than 40% was considered an absence of alpha activity. Jasiukaitis Hakerem 1988 (supra) use a different technique, comparing the highest 25 percentile of alpha amplitudes with the lowest 25 percentile. This highest 25% method was considered an appropriate definition of "high alpha" for interactive recording.

While sampling rates (128 Hz) are the same for the recording apparatus 14 and monitoring computer 19 and both machines share the same amplifiers, the sample points for monitoring and recording computers are not synchronised and can differ by up to 3.6 ms. Using spectral analysis over a full second of data sampled at 128 Hz however, this difference is not I I I r-ir I considered significant. The synchronising pulse 20 is used to generate a marker in the EEG recording apparatus 14 which can be used in off line averaging and artefact control procedures.

A similar stimulus synchronising pulse 22, generated simultaneously by the monitoring computer 19, causes a separate stimulus computer 23 to generate the required stimulus 25. The type of the stimulus may be specified by the monitoring computer 19 via data bus 21. The stimulus computer may also log responses, such as reaction time (as reflected in an optional button-press provided for the subject), in a second data store 24. In the embodiment the selected stimulus comprises an auditory stimulus, although it will be apparent to those with skill in the art that one or more of a visual, olfactory, gustatory, tactual, somatosensory or electromagnetic stimuli may be employed.

A simple auditory oddball paradigm is utilised in the investigation, with frequent stimuli which the subject is asked to ignore and infrequently occurring "oddball" stimuli for which the subject is asked to keep a mental count. The frequent tone was of 250 Hz frequency at 60 dB nHL. The infrequent stimuli was a tone of 2000 Hz at 60 dB nHL. The stimuli 25, generated by the stimulus computer 23 in response to the stimulus synchronising pulse 22, are step function tones with 5 millisecond rise time and 40 millisecond duration. The infrequent tone comprised 25% of the stimuli presented in a pseudo random fashion. The random sequence was adjusted to ensure no more than three consecutive target "oddball" tones would be generated.

With reference to the flow chart shown in FIG 2, some key steps of the method of the embodiment will be discussed. The flow chart focuses primarily on the steps undertaken by the monitoring computer 19 in accordance with the IERP method. Working from the top left of the flow chart, a first external process 1 merely represents the step of reading from memory or S perhaps a peripheral input device, the preselected brainwave state specified for triggering the stimulus. The second external process 2 (top right) represents a background process which continuously samples the amplified and filtered EEG signals 18, as supplied by the recording apparatus 14, performs an A/D conversion and supplies the EEG data to a buffer accessible by the monitoring computer 1 9.

OS b The main thread of the method comprises several nested loops, beginning with the step of processing the EEG buffer contents to determine the subject, s actual brain wave state. As discussed above in relation to the system of the embodiment, this involves obtaining a series ,0 of 128 data points from the buffer and performing an FFT. The subsequent step is a decision S• point involving a comparison of the result of the FFT with the preselected brainwave state. in the embodiment the preselected brain wave state corresponded to an FFT value indicative of alpha activity. If the alpha activity threshold was not met, the process is was repeated.

However, when the threshold indicative of the preselected brain wave state was exceeded, a stimulus synchronising pulse was sent to the stimulating computer 23 and a collection synchronising pulse was sent to the recording apparatus 14 such that the EEG data could be marked to indicate the timing and type of stimulus. This mark facilitates the later processing of the EEG data to determine the average ERP. In some cases the EEG data will not be continuously recorded, but recorded in response to the collection synchronising pulse.

The final decision point is used to control the length of a recording session or block, which in the embodiment is determined in terms of a desired number of ERP sweeps. The external process 3 following the desired ERP sweeps, represents the averaging of the EEG data to produce the resultant ERP value.

While three separate computers were used in the present configuration for convenience in programming, there did not appear to be any significant timing barriers to a single computer carrying out the same task. Artefact elimination was conducted on the basis of all channels, with data being discarded on the basis of any artefact activity (such as that indicated by EOG signal) occurring within 500 ms of application of the stimulus to a subject. The averaged ERPs were formed by post processiig of an EEG recording with each stimulus onset indicated by the digital marker on one channel. The width of the marker was used to is indicate the particular type of stimulus as described in Languis Miller 1989 ("Development and utilisation of single trial evoked potential analysis using forward and back averaging of marked EEG", BASIS Newsletter,Nov: 3-6).

The paradigm used in the investigation consisted of four ERP recording blocks, two practice and two experimental. The aim was to provide a balanced contrasts design within the constraints of providing an identical stimulus sequence. A group of 35 people, including 21 female and 14 male, were selected for the investigation.

The first block consisted of an "interactive" recording, to allow determination of the appropriate threshold parameter. During this block, the threshold values were adjusted manually by the operator to determine an appropriate threshold level which would represent "alpha" activity. The aim of this determination was to allow stimuli only when the background alpha amplitude was in the top 25% (Jasiukaitis Hakerem 1988) of its amplitude distribution.

An initial estimate of the threshold level was obtained from simple EEG data obtained in a previous set of recordings with the same subjects, however substantial adjustment was still required in most cases.

The second ERP recording block followed, in which the stimulus presentation schedule was not linked to EEG alpha, but followed a different pseudo random schedule alone. These two recordings together formed a preparatory session. This session, as well as being is used to S set the threshold value, also provides an essential rehearsal session to minimise the effects of practice on the experimental session. Jodo Inoue 1990 ("Effects of practice on Go/No Go task", Electroenceph. Clin. NeurophysioL 76 249-257) detailed the marked effect of practice on a P300 task, even though the paradigms are different. Both 2 0 practice effects and habituation effects Amochaev et al 1989 ("Topographic mapping and habituation of event related EEG band alpha desynchronisation", Intern. J. Neuroscience 49 151-155) can be minimised in the experimental session by the use of a long (approximately 25 minutes) preparatory recording session.

The third ERP recording block was the "interactive" recording of the experimental session, in which the alpha threshold value was held constant to the final value determined in the first recording block of the preparatory session. This third block, as well being used to generate the "interactive" ERP, also generated an ISI map which consisted of the sequence of inter stimulus intervals which were used. This map was stored in memory for use with the subsequent recording. This block comprises the "alpha" recording based on the "interactive" ERP recording method.

The fourth block was recorded without regard to background alpha activity, but used the same sequence of inter stimulus intervals based on the IS] map from the previous block, and the A 12 i i: i.i--.ii.-ilii rqi~i"~ same sequence of stimuli. This block comprises the "random" recording based on the prior art ERP recording method. It was the effect of background alpha activity in this pair of recording blocks, the experimental session, which was analysed.

The "interactive" methodology relies on comparing the results of applying an identical pattern of stimuli with two different background EEG configurations. For the "alpha" recording, the stimuli should occur only when background alpha activity was present. Subjects for whom such background activity occurred only rarely, were not considered to represent a valid "alpha" recording, and were discarded from analysis. For the "random" recording, stimuli should occur randomly whether background alpha activity is present or not.

Subjects for whom non alpha background activity occurred only rarely, were not considered to represent a valid "random" recording, and were discarded from analysis. In both situations, an acceptable level of background activity occurrence was set at more than Analysis was therefore restricted to subjects for whom the interactive recording of the experimental session contained between 25 and 75 percent of the background activity above the threshold. FIG 3 illustrates the effect of "alpha" IERP recording on the P300 (P3) amplitude at Pz for the subjects in the 25 to 75 percent band. The data points represented by R are the values obtained in response to random pre-stimulus activity whilst A are the respective values obtained in response to alpha pre-stimulus activity.

The interactive recording method provides information about is the effect of background o: EEG on the ERP formed from a complete block of 200 trials over approximately 10 minutes, a realistic P300 paradigm block of trials (Polich 1991). Retrospective selection of sweeps allows conclusions to be made only for the particular sweeps selected. No conclusion can be made on the effect of retrospectively selected pre-stimulus activity on the ERP average over the whole block of trials. As it is this block average which is commonly reported, a method which could Sstudy effects on the average directly was considered to be desirable. In summary, both alpha (Niedermeyer 1993), and ERPs (Picton 1988) have been shown to be strongly influenced by several experimental paradigm variables. In retrospective sweep selection, there is no way of ensuring that these variables are controlled or matched. Interactive recording provides a means of controlling for these variables when studying the effect of pre-stimulus EEG.

In a second embodiment of the method of the invention, an auditory oddball paradigm o.o: was implemented in order to study the effect of the IERP method in a group of subjects with Attention Deficit Hyperactivity Disorder (ADHD). Attention-deficit hyperactivity disorder is a childhood developmental disorder characterised predominantly by poor attention, impulsivity, and overactivity (American Psychiatric Association, 1987).The prevalence of ADHD ranges from 2% in primary care paediatric samples to 6% in school-age children, and from around in community samples to 50% or higher among clinical referrals.

Stimulants have proved among the most effective treatments for childhood ADHD.

ADHD children treated with stimulants display enhanced reaction time, accuracy, and amplitude of the P3b component of brain event-related potentials (ERP) during vigilance, memory and learning tasks, see Douglas 1983 ("Attentional and cognitive problems." In M.

Rutter Developmental Neuropsychiatry. New York: Guilford Press, 280-283.); Douglas et al. 1988 ("Dose effects and individual responsivity to methylphenidate in attention deficit disorder." J. Child Psychol, Psychiatry, 29 453-475) and Klorman et al. 1991 ("How eventrelated potentials help to understand the effects of stimulants on attention deficit hyperactivity 13 i* disorder." In J. R. Jennings, P. K. Ackles and M. G. H. Coles Advances in Psychophysiology, Vol 4. London: Jessica Kingsley Publishers, 107-153). Reviews of Methylphenidate (MPH) outcome studies, in all evaluating thousands of ADHD children typically find that between 73% and 94% were judged to have improved after a brief course of stimulant drugs. Stimulant effects are generally of larger magnitude in ADHD than in normal children or adults. Furthermore, stimulant effects are comparable for a range of ADHD subgroups as well as for acute and chronic administrations.

Considerable research has been carried out comparing ADHD groups with controls in terms of the P3b component. In particular these studies show a distinction in a clinical group performance to tests of sustained attention as opposed to tests of selective attention as discussed is in Klorman 1991 (supra). However, the investigation of the embodiment was designed to test for differences in sustained attention.

Tests of sustained attention have consistently found decreased P3b amplitude in the ADHD sample. Klorman et al 1979 ("Effects of methylphenidate on hyperactive children's evoked responses during passive and active attention", Psychophysiology 16 23-29), Klorman et al 1992 ("Methylphenidate reduces abnormalities of stimulus classifications in adolescents with attention deficit disorder", J. Abnorm. PsychoL 101(1) 138-139) and Holcomb Ackerman 1985 ("Cognitive event related potentials in children with attention and reading deficits", Psychophysiology, 22 656-667) used a visual test of sustained attention, the Continuous Performance Task (CPT). Loiselle et al 1980 ("Evoked potential and behavioural signs of attentive dysfunction in hyperactive 5 boys", Psychophysiology, 17 193-201) and Lubar et al 1990 Discourse on development of EEG diagnostics and biofeedback for attention deficit hyperactivity disorders", Biofeedback Self Rag. 3 201-225) used an auditory vigilance task to again test sustained attention. This auditory task was the basic Auditory Oddball Paradigm described by Ritter et al. 1968 ("Orienting and habituation to auditory stimuli: A study of short Sterm changes to auditory evoked response", Electroenceph. Clin. Neurophysiol. 25 550-556).

Other studies indicate that the reduction in P3b amplitude is not obtained with tasks of selective S attention, or with tests in which the two groups performed comparably, see Callaway et al 1983 (Hyperactive children's event related potentials fail to support under arousal and maturational lag theories", Arch. Gen. Psychiatry 45 1107-1117).

Tests involving ADHD subjects and the effect of psychostimulants on the ERP (Klorman, 1991 supra) indicate that the P3b component amplitude is indeed increased or "normalised" under MPH treatment. Young et al 1995 ("Acute challenge ERP as a prognostic of stimulant therapy outcome in attention-deficit hyperactivity disorder", Biol. Psychiatry 37 33) reports on usefulness of the acute ERP response to a single dose administration of MPH as a prognostic for stimulant therapy outcome after six months in clinically referred ADHD children.

This MPH challenge test focuses on the magnitude of changes in the evoked response potential (ERP) amplitudes between baseline and two-hour post-drug measures.

The investigation of the embodiment tests initially whether a significant increase in average ERP amplitude can be achieved by stimulating interactively. It then tests whether there is any difference in the effect of interactive ERP testing based on the subject's result on the MPH challenge test. The subjects were 66 clinically referred children and adolescents (Aged 6- 17 years: mean =12.3, sd 2.6) presenting with symptoms of attention-deficit hyperactivity disorder referred for ERP testing.

l rt i As with the first embodiment, a simple auditory oddball paradigm was utilised.

Frequent stimuli (500Hz) were ignored and infrequently occurring "oddball" stimuli (2000Hz) were mentally counted. Subjects were instructed to "count the target tones as a check will be made at the end of the test". An important aspect of the investigation was the comparison of standard ERP and the interactive ERP recording techniques used to implement this paradigm. The effects of each technique on the resulting ERP data and the means of determining the inter stimulus interval are of particular interest.

The first trial was a normal auditory oddball task with stimuli applied at random without regard to background EEG. Stimuli consisted of tone bursts of 40 ms duration with 1 0 ms rise and fall times applied to both ears simultaneously. Target tone was at 2000 Hz and non target tone of 500 Hz. 50 target and 200 non-target tones were applied with a fixed inter-stimulus interval of 1.1 seconds. Delivery sequence was random, with manual adjustment of a true randomly generated sequence to ensure no more than three sequential target stimuli were applied. This recording was used to generate the baseline value for later comparison. This value was also termed "pre" in computing the medication response, or "standard" in computing the recording technique response.

After a 10 minute period, the interactive ERP recording technique was carried with stimuli applied in response to a selected pattern of background activity. The data window is of o: 128 points (1 second) duration, and is shifted 16 points (125 milliseconds) with each test. In this embodiment, the application of the auditory stimulus is conditional on a preselected pattern o occurring in the pre-stimulus background EEG. The preselected pattern used has been found to be associated with an increase in ERP signal to noise ratio.

The determination of the occurrence of a matching pattern in the background EEG comprising the pre-stimulus brain wave state can be achieved using either of two broad approaches. The first approach, to what amounts to the analysis and classification of a data set, is the "decision theoretic" or pattern recognition approach. The most popular method of pattern S recognition used in EEG research is the statistical pattern recognition technique, also called decision theoretic or the discriminant approach.

In contrast, the second approach is the structural or knowledge based approach which includes the rule based and computational linguistic methods. The second embodiment uses S" syntactic analysis, which is the mathematical -formulation of the computational linguistic method. The capacity of the computational linguistic method to investigate the structural information contained in an EEG recording is particular advantageous for classification purposes. Syntactic analysis is readily adapted to hierarchical decomposition of a single entity into a number of samples.

In the method of the embodiment, the entity is one second of EEG data corresponding to the brain wave state of the subject. The entity is decomposed into a set of small samples, termed "primitives" which correspond to known classes of EEG component waveforms. The method chosen to identify the EEG waveforms (and thus brain wave states) of interest is that of wavelet detection. One problem to be overcome is the detection of the presence of a wavelet in a noisy signal, as are EEG signals in general. Whilst the approximate shape of the wavelet is known, the time of occurrence must be estimated.

The wavelets are detected using a technique referred to as template matching which is L :Iw based on matched filtering using cross correlation in view of the relatively small number of sample points involved. Assuming a noisy signal then: s(t) where s(t) is the unknown signal, and n(t) is noise.

Then s(t) i Gi i(t ti) where i(t) is a time dependent template of the waveform being sought, and Gi is the multiplicative gain applied to the template to generate the signal.

An estimate of the gain Gi can be obtained by cross correlation of the template i with the noisy signal si. For a constant template normalised for unit power, the minimal cross correlation gain value G, which is used in classification. If the minimal value is greater than a particular threshold, a signal is deemed to be present. The threshold value varies from subject to subject and may be manually varied during the ERP recording.

Individual features of a small portion of a target background EEG waveform, ie.

primitives, are used as the basis of the comparison. The primitives include single cycles (built up from half waves) of various frequencies selected to represent those commonly occurring in the brain wave activity; for example "faster alpha" activity (12 points, 10.7 Hz), "average theta" activity (22 points, 6 Hz), "general beta" activity (8 points, 18 Hz) and "fast delta" (34 points, Hz) where "points" refers to the number of data points required for a sampling rate of 128 Hz.

0*0 The comparison step involved the use of finite state string grammar, wherein a one dimensional string specifying the desired occurrence and/or sequence of primitives is created.

The patterns to be recognised correspond to combinations of alpha, theta, delta, beta and flat wavelets. In the embodiment, for example the grammar is able to express a target brain wave state for ADHD as being: a single cycle of theta which is desynchronised after the theta cycle and which does not show alpha primitives before and after the theta cycle in such a string. It is anticipated that other target brain wave states could be devised for applications in diagnosing a wide variety of mental dysfunctions, including MS and schizophrenia.

The embodiment used only a single trace (Pz) for the purpose of analysis. However other embodiments incorporating topographic information from a plurality of EEG traces could be converted to a single dimension, or 2 dimensional array grammar could be utilised. The "interactive" recording system and method was otherwise as set out in relation to the first embodiment above. This recording was used to generate the "interactive" value for later o comparison.

In this "interactive" recording, the total test time was restricted to no more than the standard recording time. In some subjects, this lead to a lesser number of stimuli during interactive recording. In others, the recording time was less than the random recording time.

Using this method, while the sequence of targets and non targets is the same for random and interactive recordings, the sequences of inter stimulus intervals are not equal. Nor are the number of stimuli or the total recording times.

On completion of the interactive test, the subjects were given a challenge dose of Methylphenidate and allowed to leave. Upon returning 90 minutes later, the recorders were reapplied to the previously marked locations, without additional preparation. Another standard ERP recording, identical to the baseline recording, was then carried out. This recording was used to generate the "post" value for later comparison.

The ERP was generated by off line averaging sections of EEG based on the stimulus marker, generated on the EOG channel. Single sweeps were first visually inspected and any sweeps with eye movement or muscle artefact identified within 500 ms of the stimulus were discarded. The remaining sweeps were then averaged for target and 2 0 non-target stimuli. The topographic maps so generated were used to identify the P3b component for each recording.

The latency and amplitude measures of this component were used for further analysis.

The medication response ratio was computed as the ratio of P3b(Amp) post-MPH to P3b(Amp) pre-MPH. The recording technique response ratio was computed as the ratio of P3b(Amp) using the interactive recording technique, to P3b(Amp) using the standard recording technique. A minimum of 15 target sweeps after elimination of eyeblinks was mandatory for both random and interactive recordings. A total of 66 matched triplets of recordings were obtained for random and interactive pre medication and random recording post medication.

A comparison of random ERP P3b amplitude pre medication versus random ERP P3b amplitude post medication was conducted. This data was used to assess the subjects as Responders R) or Non responders N on the auditory MPH challenge test. The medication response ratio was calculated, and subjects with the ratio above a pre set threshold value (1.3) were rated as responding positively. A clinical rating was also used at a later stage, but initially the response rating was based solely on the physiological response to medication. 18 subjects A comparison was then made of amplitude of P3b peak generated using the standard ERP recording technique and using the interactive recording technique in response to a selected pattern of activity N or R both before medication. A ratio was again calculated (interactive standard) to quantify the response to recording technique variation. FIG 5 summarises the S. recording technique response ratio. Using the Bartlett test for homogeneity of variance, there was a significant difference (Chi square 7.13, df 1, p 0.008) in the recording technique response ratio between the groups of responders and non responders.

The correlation between the recording technique response ratio and the medication response ratio was then analysed explicitly without the use of a grouping threshold. A o significant correlation was found between the two ratios with a Pearson correlation coefficient of 0.543 (Bartlett Chi squared statistic 22.163, N 66, df 1, p 0.001). The probability value was not affected by treating the analysis as an unplanned comparison, and applying a Bonferroni correction factor. FIG. 5 shows how the recording technique response ratio, and the medication response ratio are related.

This graph also highlights a possible bias toward data which are based on a very low baseline value (the denominator in both ratios). To test for such a bias, the correlation value was recalculated after excluding any data point for which the baseline value was less than 5 uvolts. 8 data points were excluded, but there was still a significant correlation (0.453, Chi squared 12.725). To further guard against such possible bias, the correlation between the recording technique response difference (interactive baseline), and the medication response (post baseline) was also tested. A significant correlation was likewise found between the two differences with a Pearson correlation coefficient of 0.467 (Bartlett Chi squared statistic 15.633).

The amplitude data, rather than ratio data, was then analysed, to study the effect of 17 interactive recording. In comparing the raw amplitude data for matched pairs of random and interactive recordings, an assumption of independence clearly cannot be made. A paired samples t-test was therefore used. This comparison, with all data combined, showed no significant difference between standard recording and interactive recording (t 1 86, df 65, p =.853).

The t-test comparison was then repeated with the subjects grouped by the subsequent auditory MPH challenge test result. In the non responder group, a trend was evident (t 1.97, df 47, p .054) toward the interactive recording being lower than the standard recording. In the responder group, a significant difference was found (t 2.56, df 17, p .020), with the interactive recording being of higher amplitude than the standard recording.

To investigate whether the recording technique effect was more likely to be associated with the short term physiological effects of the medication, or the long term clinical effects, the analysis was carried out again based on the clinician's feedback. This feedback was essentially a responder/non responder rating, based on a six month assessment period. There was again a significant difference (Chi squared 17.875, df 1, p .001) in the recording technique response ratio between the groups of responders and non responders. The raw amplitude data was also analysed using groups based on the clinical assessment. In the non responder group, a significant difference was found (t -1.667, df 25, p .018) with the interactive recording !i being lower than the standard recording. In the responder group, a trend only was found (t 1.644, df 39, p .108), with the interactive recording being of higher amplitude than the standard recording. These findings were obviously weaker than the findings based on the physiological ratings.

In summary, the results indicate that the ERP obtained by stimulating during selected background EEG activity in accordance with the method of the second embodiment was significantly greater than the ERP obtained by stimulating at random, but only for subjects who subsequently responded well to MPH. To a lesser extent, the results also indicate that applying stimuli during selected background EEG activity has the opposite effect on the ERP for subjects who did not respond well to medication. The effect in this instance, however, was not as significant as the first finding. Both effects were more marked when response to medication o was measured purely on a physiological basis, but were also found when the clinical response was incorporated.

These results indicate that there is a significant association between the effect on the ERP of medication, and the effect on the ERP of the interactive recording technique. There is no logical a priori reason why this should be so.

The first result found was that "there is a significant difference in the recording technique response ratio between the groups of responders and non responders". The grouping into responders and non responders was based on the value of the medication response ratio, so there appears to be a significant difference in the effect of interactive recording, depending upon the effect of medication. This conclusion is bome out by the high correlation found between the medication response ratio, and the recording technique response ratio.

The most obvious conclusion to be drawn from this investigation is that the effect of interactive ERP recording on the P3b amplitude is the same as the effect of MPH medication on the P3b amplitude. This is exemplified in FIG 4 where the Post MPH values and the 18 91 interactive values cluster together. This conclusion has beneficial implications from a clinical perspective. If the clinical is information derived from the MPH challenge test is available using an interactive test, then the test time can be cut from four hours to 1 hour. More importantly, there would be no need for a test dose of Methylphenidate with clear ethical benefits. In addition, the comparison with clinical assessment, showed that the interactive recording technique had considerable predictive power in its own right. If the positive (81%) and negative predictive power found in this sample were confirmed, then a test based on two ERP recordings compares very favourably with present methods of assessment of ADHD.

Klorman 1991 (supra) suggests that the effect of MPH is to "normalise" the ERP of ADHD children. If this is true, then it could be argued that the effect of interactive ERP recording is likewise to normalise the ERP response. Since this normalisation is based on analysis of the background EEG, then the "abnormality" must also be evident in the background EEG. This finding has implications for theories of attention and ADHD pathology not apparent prior to the implementation of the interactive ERP recording method of the invention.

Conceptually the IERP process is a methodology distinct from both EEGs and ERPS. The process provides new information which is simply not obtainable by the EEG or standard ERP processes. This information, how the ERP interacts with the brain wave state of the EEG, has the potential utility of any new measure in research and clinical practice.

In practice, however, it may at present be considered as an adaptation of the standard ERP technique. The major advantage of the IERP process over the standard ERP process is that it provides an increase in scientific validity through superior theoretical support. The process S has been shown to produce an IERP which is significantly different to the standard ERP for a particular brain wave state (alpha activity). As the background EEG in the typical standard ERP consists of varying length periods of several brain wave states, the resulting standard ERP is the result of an uncontrolled mixture of ERPs to each brain wave state. The IERP process allows the investigation of a pure base on a single brain wave state) ERP, instead of a mixture. The o advantage is obtained by the process allowing the control of a variable (brain wave state) which has been shown to have an effect on the ERP.

S"The, brain wave, state of alpha activity has also been linked with the mental state of attention. The IERP process would allow an ERP to be generated with far greater control of S attention (as reflected in brain wave state) than can be obtained using behaviour techniques.

The process would also allow a similar investigation of brain wave state to mental state linkage in other ERP paradigms.

The use of pure IERP recordings would also have an advantage in the investigation of patient groups where the brain wave state might be expected to be variable. This is the case in mental illnesses such as Attention Deficit Hyperactivity Disorder and Schizophrenia, where both EEG and ERP are recognised as different, but are not specific to the diagnosis. The IERP provides new information which may prove to have higher sensitivity and specificity than the present EEG and ERP information.

It will be appreciated, by those with' skill in the art, that similar considerations apply to the electrical and magnetic techniques introduced above. Accordingly, although the description focuses on ERPs and EEG, the invention disclosed herein finds application in ERFs and MEG and may also find application in related techniques for acquiring event-related data 19 representative of physiological activity in the brain, such as positron emission tomography

(PET).

Throughout the specification the aim has been to describe the preferred embodiments of the invention without limiting the invention to any one embodiment or specific collection of features.

0

*U

*U*

a *0 o

U