CN101583308A - Analysis of brain patterns using temporal measures - Google Patents

Abstract

A set of brain data representing a time series of neurophysiologic activity acquired by spatially distributed sensors arranged to detect neural signaling of a brain (such as by the use of magnetoencephalography) is obtained. The set of brain data is processed to obtain a dynamic brain model based on a set of statistically-independent temporal measures, such as partial cross correlations, among groupings of different time series within the set of brain data. The dynamic brain model represents interactions between neural populations of the brain occurring close in time, such as with zero lag, for example. The dynamic brain model can be analyzed to obtain the neurophysiologic assessment of the brain. Data processing techniques may be used to assess structural or neurochemical brain pathologies.

Description

Utilize of short duration Measurement and analysis brain connection pattern

Technical field

The present invention relates to nervous physiology analysis (neurophysiologic analysis) by and large, more specifically, relates to but is not limited to the service time sequence and represent the form analysis brain connection pattern.

Background technology

As any other organ of health, the function of brain also needs to assess, to estimate its health and morbid state.Yet, be different from any other organ of health, also there is not preferable brain function test at present.Typical behavior inspection comprises standard neurologic examination, psychiatric interview or neuropsychological test.Electroencephalogram (electroencephalogram; EEG) almost can not provide any information,, as lie in a comatose condition unless it is serious disorderly to exist epilepsy grand mal or brain function to exist.Be used to assess brain structure (for example nuclear magnetic resonance (magneticresonance imaging; MRI)), chemical property (magnetic resonance spectroscopy (magnetic resonancespectroscopy for example; MRS)), can not replace the assessment of brain function based on the positron emission tomography (fluoro-deoxy-glucose based positron emission tomography (PET)) of fluorodeoxyglucose or pharmacology's (based on the PET of ligand) method.At last, " functional " MRI (fMRI) reaches based on O ¹⁵PET relevant with the brain that in particular task, is activated zone, but not relevant with brain function itself.

Neuropathy comprises for example cognitive competence obstacle (cognitive impairment), is the significant problem that constantly increases the weight of.For example, be called as Alzheimer's disease (Alzheimer ' s Disease; Under the situation of cognitive competence obstacle AD), effectively get involved and depend on EARLY RECOGNITION.Mild cognitive ability obstacle (the mildcognitive impairment of amnesia form; MCI) be syndrome in mistake intelligence early stage among the old people, usually can develop into Alzheimer's disease.Although the Clinical symptoms of Alzheimer's disease and mild cognitive ability obstacle is very definite usually, yet really mistaken diagnosis can take place, thereby make research and treatment make great efforts to become complicated.

People wish to have the objective test that is used for Alzheimer's disease, cognitive competence obstacle or other sacred diseases, but the whole bag of tricks that is proposed so far all has significant disadvantage, thereby have limited its potentiality as sensitivity, reliable diagnostic or appraisal tool.

For example, the method that the 6th, 463, No. 321 United States Patent (USP) is illustrated a type is to bring out reaction potential (evoked response potential; ERP) utilize electroencephalogram (EEG) to measure in the test.Gather data collected, and produce the single vector of representative overall experimenter's reaction of ERP test from EEG.Then, with the experimenter of this vector and known health and diagnosed out and have sacred disease experimenter's the vector of (for example Alzheimer's disease, depression or schizophrenia) compares.Based on a shortcoming of the measurement of ERP is to bring out reaction and can cause very active other brain zone that becomes, some brain zone then to keep inactive relatively what stimulate.Therefore, the EEG that is gathered measures the brain zone that main representative is activated.Utilize this kind method, can't obtain the measurement of the overall cerebration of representative that the activity of active regions is not too also taken into account.When using traditional E EG gauge, this problem is even more serious, because traditional E EG gauge mainly detects near the electrical activity of outer surface of brain, the sensitivity in darker brain zone then significantly reduces.

The 7th, 177, No. 675 U.S. Patent Publications are a kind of to be the method that the patient that diagnosed selects therapy by comparing with the data base that the symptom individuality is arranged who various therapies is had active responding.To for example be compared with the data-base recording of reference individuality by the quantitative nervous physiology information that EEG/QEEG/MEG obtained, for having the active someone of similar EEG/QEEG/MEG, which kind of therapeutic process is the most effective with prediction.Yet disclosed measurement and data analysing method relate generally to Spectrum Analysis, can not identify the delicate peculiar sign of some disease or state from collected all are measured.But the EEG/QEEG/MEG data are to classify according to therapeutic outcome on the whole.

At (Exp.Brain Res. such as Leuthold " Time Series Analysis of Magnetoencephalographic Data " that the people showed, 2005) in, the author has described following experiment: when the experimenter carries out the exercises task and stand various visual stimulus (being included in the image of seeing variation in the eye gaze task), gather the MEG data.Use time domain ARIMA Box-Jenkins modeling pattern to analyze MEG data in-25 to+25ms short-term interacts.Described data are carried out prewhitening, and utilize intercorrelation function (cross correlationfunction; CCF), auto-correlation function (autocorrelation function; ACF) and partial auto correlation function (partial autocorrelation function; PACF) analyze from the paired interaction between the data sequence of MEG acquisition.The closely motion of the motion of monitoring hands and eye, and utilize these to move to make MEG output relevant with the experimenter's activity that is just taking place.Sampling period is a little more than 1kHz.

This works is assessed each to the interaction between the time series in the pick off.During the experimenter carries out these tasks, just observing/having or not in the negative intercorrelation pattern selected each pick off is exported.Although this works has formed some interesting (for example the seeing clearly to the measuring technique that is used to obtain the MEG reading, advantageously utilize the 1kHz sampling, and data are carried out pretreatment with its prewhitening), yet this works only utilizes the interaction of separated sensor signal, and the whole brain model that will not study a large amount of sensor group is taken into account.In fact, hereinafter disclosure is with conspicuous reason owing to basis, and people's such as Leuthold disclosure can not be analyzed the brain state or the diagnosis brain state of cerebration to characterize the experimenter.

In view of this reach other shortcoming of known technology, need a kind of actual solution that is used for automatically analyzing cerebration, this solution should be able to detect reliably and discern various different experimenters the peculiar obvious neuro-pattern of the heart and brain state that closes.

Summary of the invention

One aspect of the present invention relates to be analyzed and classifies experimenter's's (human experimenter who for example inquires into or study or neuropath) nervous physiology activity.Receive experimenter's data acquisition system as input, this experimenter's data acquisition system representative is by the active time series of the nervous physiology that pick off obtained of many spatial distributions, and described pick off is arranged in order to survey experimenter's neural signaling during the idle condition of opening at eyes.Described number of subjects according to can by magneticencephalogram measure or certain other can provide the suitable measurement of time serial message and enough measurement sensitivity to obtain.

The various template stores that to classify according to different brain states for example are stored among the data base in data storage.In the described template each is all represented from each neural colony that known at least one other experimenters that present set brain state measure on statistics the independently selected subclass of of short duration measurement.Described on statistics independently of short duration measurement can comprise in described experimenter's data acquisition system the set of the part intercorrelation between the different time sequence of packets, synchronous substantially interaction between each neural colony in the described experimenter's of described part intercorrelation set representative the brain.

Handle described experimenter's data acquisition system, with the dynamic model of the of short duration measurement in each the neural colony that obtains the described experimenter of representative.Why described dynamic model is dynamic, is because its functional form with the time is represented of short duration measurement.In an exemplary embodiment, described dynamic model comprises the paired seasonal effect in time series part intercorrelation of taking from described number of subjects certificate.Described dynamic model can comprise all each in the time series all to or a certain subclass.

At least a portion of described dynamic model is compared with described a plurality of templates,, produce described experimenter's nervous physiology classification of activities with when described dynamic model is corresponding with in described a plurality of templates at least one.

In another aspect of this invention, by the systematic analysis experimenter's who comprises data input pin and processor nervous physiology activity.Described data input pin can comprise communication interface, for example computer network interface.Described data input pin receives experimenter's data acquisition system, and described experimenter's data acquisition system representative is by the active time series of each collected nervous physiology in the pick off of many spatial distributions, and described pick off is arranged in order to survey patient's neural signaling.Described processor be coupled to described data input pin with communication mode and be programmed in order to: handle described experimenter's data acquisition system, obtaining dynamic brain model, described dynamic brain model is represented independently of short duration measurement on statistics in each neural colony of brain of described experimenter.Then, described system can analyze described dynamic brain model, to estimate described experimenter's nervous physiology state.

Another aspect of the present invention relates to a kind of active system of nervous physiology that is used to analyze first experimenter.Described system comprises: data input pin, be used to receive a plurality of brain activity data set of the idle condition of opening corresponding to eyes (for example eye gaze task), wherein each set representative is by the active time series of the collected nervous physiology of the pick off of many spatial distributions, and described pick off is arranged in order to survey corresponding experimenter's neural signaling; And processor, be coupled to described data input pin with communication mode.

Described processor be programmed in order to: handle the set of each brain activity data, to produce corresponding neural activity dynamic model, described neural activity dynamic model is represented between each neural colony of described first experimenter coupling with time correlation, comprise: handle described brain activity data, to produce the prewhitening time series, described prewhitening time series has meansigma methods, variance, reaches autocorrelative stationarity feature; Calculate the paired part intercorrelation of described prewhitening seasonal effect in time series, each is to the intensity of the signaling between the pick off and the estimated value of symbol in described many pick offs to produce, and described estimated value is represented the paired interaction of neural colony; Described part intercorrelation is carried out classification, and with the measurement of the dependency that produces described brain activity data and verified reference data, described verified reference data is corresponding to multiple different nervous physiology states.

Embodiments of the invention comprise the diagnostic tool that is used for clinical setting or are used for estimating at research environment experimenter's instrument.More generally, each side of the present invention is provided for utilizing data handling system to obtain automatically instrument structural or the nervous physiology assessment that the neuro chemistry encephalopathy (HIE) becomes.The system and method for different aspect is applicable to the nervous physiology state of monitoring experimenter's potential variation, for example progress of disease according to the present invention.In addition, others of the present invention are provided for monitoring the solution of patient's curative effect.

In addition, the automatic nervous physiology classification that provides from different one group of known or unknown state experimenter's brain state is provided various aspects of the present invention.For example, embodiments of the invention can be used for providing from one or more of following state and differentiate classification accurately: normal condition, Alzheimer's disease, mistake intelligence syndrome in early stage, mild cognitive ability obstacle, schizophrenia, dry syndrome, alcoholism, alcohol damaged, fetal alcohol symdrome, multiple sclerosis, Parkinson's disease, bipolarity emotion disease, traumatic brain injury, depression, autoimmune function disease, neurodegenerative disease, pain, the disease that influences the central nervous system or its combination in any.

Preferably, embodiments of the invention can obtain brain sign, diagnosis and other result from may only continuing a few minutes or shorter single test or measurement phase (measurement session).This use that is used in the measurement device of collecting the number of subjects certificate has low relatively cost and high relatively volume of production.For the experimenter, short and do not have a situation that the comfortableness of the collection period (for example using the MEG instrument) of invasive obviously is better than repeatedly going to the clinic or must bears very long test.By the measurement quantitative, that be correlated with on statistics to multiple brain state is provided, embodiments of the invention can overcome the shortcoming that reaches other clinical data according to the observation, uses current indirect diagnostic techniques that present common disease of brain are carried out subjective assessment qualitatively.Another beneficial effect that each side of the present invention provided is can accept without pretreated data as input.Therefore, can utilize SMI output, and need not to carry out special device upgrade, and can carry out thoroughly evaluating to the experimenter, and need not to carry out the prediction of any apriority or prepare for selecting state some conjecture or that infer at various states.

The present invention also provides various other advantages, and by disclosed content hereinafter, these advantages will become apparent.

Description of drawings

Accompanying drawing may not be to draw in proportion, and in these accompanying drawings, identical numbering is described the essence components identical.Different instances with same reference numeral essence components identical of different letter suffix.Accompanying drawing is the various embodiment that discussed with mode cardinal principle graphic extension this paper for example and not limitation.

Fig. 1 graphic extension exemplary MEG instrument;

Fig. 2 A and 2B are with visual representation form graphic extension experimenter's synchronous dynamic network;

The exemplary classification chart that Fig. 3 graphic extension utilizes typical discriminant function (canonical discriminant function) to form;

The spatial model of three separated sensors of Fig. 4 A-4C graphic extension;

The spatial model of three other pick offs of Fig. 5 A-5C graphic extension;

The different instances of the network of the extensive interconnection of Fig. 6-9B graphic extension;

Another exemplary classification chart that Figure 10 graphic extension utilizes typical discriminant function to form;

The example by network struction of Figure 11 graphic extension subject matter of the present invention;

Figure 12 A and 12B graphic extension according to various aspects of the invention, with the corresponding method of example of analyzing the experimenter;

Figure 13 graphic extension according to an aspect of the present invention, a kind of skeleton diagram that is used to analyze experimenter's process; And

Figure 14 is the information flow chart according to one aspect of the invention, the information exchange in the graphic extension system.

Although the present invention can be changed into various modifications and alternative form, yet also describe detail of the present invention hereinafter in detail with the way of example demonstration in the accompanying drawing.However, it should be understood that the present invention does not really want to be restricted to specific embodiment as herein described.On the contrary, the present invention is contained still to belong to by claims and spirit of the present invention that equivalent scope defined thereof and all modifications form, equivalents and the alternative form in the scope of enclosing.

The main element description of symbols

1P: left front frontal lobe 2P: left back frontal lobe

3P: left side-frontal lobe-temporo 4P: left parietal bone of head

5P: left parietal bone of head-occipital bone 6P: right occipital bone

7P: right parietal bone of head-temporo 8P: right temporo

1N: left front frontal lobe cortex 2N: left back frontal lobe

3N: left side-frontal lobe-temporo 4N: left parietal bone of head

5N: occipital bone 6N: right parietal bone of head

7N: right frontal lobe

1000: system 1100: central server

1105: the data base 1110: server

1120: terminating machine 1200: communication network

1310: client computer platform 1312: native processor

1314: pick off 1320: the client computer platform

1330: client computer platform 3000: flow of information

3010: clinic 3012: experimenter's gauge

3014: doctor or Laboratory Technician 3016: network node

3018: instrument output 3020: patient's archives

3022: diagnostic cast 3024: result relatively

3026: report 3028: questionnaire

3030: data storage 3032: feedback/tracking

The specific embodiment

Below describe in detail and comprise the quoting of accompanying drawing, these accompanying drawings constitute the part of this detailed description.Accompanying drawing shows with way of illustration can be in order to put into practice the specific embodiment of subject matter of the present invention.These embodiment are also referred to as " example " in this article, hereinafter enough at length set forth these embodiment, so that the those skilled in the art can put into practice subject matter of the present invention.Each embodiment can be made up mutually, can utilize other embodiment, perhaps can make the change of structure, the electrical aspect of logic OR, this does not deviate from the scope of subject matter of the present invention.Therefore, hereinafter describe in detail and should not be considered as having limited significance, and the scope of subject matter of the present invention is defined by enclose claims and equivalent scope thereof.

In presents, as in the patent document common, use word " (a or an) " to comprise one or more than one.In presents, word " or " be used for representing to have not exclusiveness, except as otherwise noted.In addition, the full content of all publications, patent or the patent document of being quoted in presents is incorporated herein with way of reference, seemingly its be respectively with way of reference incorporate into general.If the application between presents and the file so incorporated into way of reference is inconsistent, then the application in incorporating into the reference paper should be considered as replenishing presents; For implacable discordance, should be with the master that is applied as in the presents.

The application of magneticencephalogram (magnetoencephalography)

Biomagnetism is meant the measurement to the magnetic field of originating in the body.These magnetic fields are to be produced by magnetic material that is associated with biological activity or ion current.An example of physiology record comprises human heart and as the induction coil or the Other Instruments of detector.Other example of biomagnetism comprises by the motion of following the tracks of magnetic particle and carries out the research of digestive system and measure magnetic pollutant in metallists's lung.Some application relates to the activity of measuring human brain.The example of magnetic field detectors comprises and is called as superconducting quantum interference device (Superconducting QuantumInterference Device; SQUID) superconducting device.SQUID is regarded as flight data recorder (black box) with its feedback electronics that is associated, and is used to provide the output voltage that is directly proportional with the magnetic flux that is applied to its search coil.Magnetic signal generally can be because of not being affected by soft tissue and skeleton, and the signal of telecommunication is then isolated by the soft tissue conduction and by skeleton.

In an example, detector comprises that being arranged to the helmet-type that carries out " whole head " analysis designs, and can comprise more than 100 SQUID passages.In an example, use 248 passages, yet, also contain more or less passage according to different embodiment.When being activated together, these coils may detect the magnetic field that electrostatic current produces when a neuron group (for example 10,000 or more).The supposition that utilizes these measured values and distribute about neural activity can calculate active position.In an example, magneticencephalogram is mainly used in preoperative mapping at present clinically.

Magneticencephalogram is a kind of functional imaging instrument, and can use the anatomy coverage diagram from MRI or CT, so that the lip-deep active position visualize of surveying of cortex.For being no more than 3 centimeters dark sources, magneticencephalogram has the of short duration resolution of millisecond scope that is in and several millimeters spatial accuracy.Exist two types MEG search coil to be used for the brain record at present: magnetometer and axial gradient meter.Magnetometer is used to write down the instantaneous strength in known spatial locations magnetic field at any time.Gradiometer then is used for measuring in the known moment magnetic flux gradient or the space derivation of local magnetic field.Gradiometer is often more responsive to darker thalamus source; Magnetometer is then mainly measured the cortex source.

The traditional analysis of MEG signal is concentrated on always the location (with identification in brain and definite single or multiple virtual bipolar positions, the described virtual bipolar preferable percentage ratio that is used for explaining the variance that observes at physical record) of Magnetic Field Source.Other of MEG signal analyzed the modeling then relate to distributed source, to derive the estimated value of the electric current density in the specific brain regions zone of being concerned about.Fig. 1 shows traditional magneticencephalogram (MEG) instrument.

According to one embodiment of present invention, in normal subjects, MCI experimenter and the AD experimenter in old age, during 45 seconds vision is watched task attentively, the dynamic short-term of using MEG research to be write down interact (25 to+25 milliseconds) between the prewhitening signal of people's cortex by 248 axial gradient meters.As noticing, find that obviously increase with the intensity of the negative correlation at age, this is very outstanding in MCI and AD for positive correlation.In addition, MCI and especially AD, with seen in the normal subjects less than new negative correlation be associated.Negative correlation means the feedback by the relay cell that connects each neuronal pool (neuronal pool) usually.In this research, in the very wide zones of different of being separated by of brain, observe negative correlation, and these relevant strengthen and enlarge with the age in MCI and AD with several milliseconds interval.Described of short duration be inconsistent at interval with the activation of relay cell, and also the opposite reaction to same stimulation can not appear simultaneously.Seem that these negative correlation are normal parts of neural background synchronicity, and its increase reflected the enhancing of neural synchronicity, cost is to carry out neurally handling required vigor or degree of freedom and losing synchronously.

Example 1

Magneticencephalogram (MEG) can be used for old experimenter and has MCI and the experimenter of AD.

In an example, dynamical state or the dynamic function that task and MEG assess the brain of three groups watched in use attentively: old experimenter (77.1 ± 1.5 years old, on average ± SEM, N=11), normal subjects (N=4,76.5 ± 2.1 years old) has the experimenter (N=4 of MCI, 75.7 ± 3.7 years old) and experimenter (N=3,79.7 ± 0.3 years old) with AD.When the experimenter watches certain point attentively and reaches 45 seconds, collect data from 248 axial gradient meters (Magnets3600WH, 4-D Neuroimaging), and these data of pretreatment with remove heart illusion electric wave or nictation the illusion electric wave.

By match autoregression integration rolling average (AutoRegressive Integrative Moving Average; ARIMA) model and get residual and, calculate all paired zero lag part intercorrelations with after the time series prewhitening, thus the directly estimated value of synchronous coupled intensity and symbol (positive and negative) provided between the neural colony with 1 millisecond of short duration resolution.

Carry out following covariance analysis: wherein positive part correlation is a dependent variable, and group and pick off are fixed factors, and the distance between age and the pick off is a covariant.Demonstrate group have the statistics on very significantly the influence (p＜10 ^-12The F test).Use Bang Folunni to proofread and correct (Bonferroni correction) and carry out showing behind the paired comparison that there are not significant difference (p=0.23) each other in normal group and MCI group, but AD group has and is starkly lower than normal group (p＜10 ^-7) or MCI group (p＜10 ^-11) average portion relevant.Positive synchronously neural interaction of AD group shows the intensity of reduction.

In an example, carry out following covariance analysis: wherein minus part correlation is a dependent variable, and group and pick off are fixed factors, and the distance between age and the pick off is a covariant.Demonstrate group have the statistics on very significantly the influence (p＜10 ^-6, the F test).Use Bang Folunni to proofread and correct (Bonferroni correction) and carry out showing behind the paired comparison that there are not significant difference (p=0.23) each other in normal group and MCI group, but AD group has and is starkly lower than normal group (p＜10 ^-4) or MCI group (p＜10 ^-6) negative average portion relevant.Seen in described variation and the positive interaction to variation be in same direction (be dependency reduce).Synchronous neural interaction of AD group shows the intensity of reduction.

Interaction between the neural colony is from sleep and awakens to the basis of all brain functioies of higher cognitive process.One aspect of the present invention recognizes, estimate these interactional intensity and spatial model can help in fact to understand brain function and with the relation of behavior.

One aspect of the present invention relates to synchronous neural (the synchronous neural interaction of interaction; SNI), in order to utilize magneticencephalogram (MEG) to assess dynamic brain function with high of short duration resolution.Technology according to an embodiment comprises that the dynamic synchronization between measurement and the corresponding neural colony of brain function interacts.Embodiments of the invention can be used for providing from one or more of following state differentiates classification accurately: normal condition, Alzheimer's disease, lose intelligence syndrome in early stage, mild cognitive ability obstacle, schizophrenia, dry syndrome, alcoholism, alcohol damaged, fetal alcohol symdrome, multiple sclerosis, Parkinson's disease, bipolarity emotion disease, traumatic brain injury, depression, the autoimmune function disease, neurodegenerative disease, pain, influence central nervous system's disease, or its combination in any.

A kind of method helps realizing the classification of the differentiation of health volunteer's scope, ill group, seriousness that the experimenter shows certain state or the measurement of degree, and permission and progression of disease or treatment intervention are side by side monitored the variation of brain function.In an example, this method can be used to assess dynamic brain function routinely, and helps to carry out the effect of Differential Diagnosis and monitoring intervention.In an exemplary embodiment, obtain classification score and posterior probability (posteriorprobability), in order to the seriousness that quantizes cerebral disturbance and monitor its therapeutic process and effect.

This method comprises by the interaction of the dynamic synchronization between the neural colony, analyzes the MEG data.This method can be used for distinguishing various brain injury, includes but not limited to have AD, the experimenter of chronic alcoholism, MCI, multiple sclerosis, schizophrenia and dry syndrome.Subject matter of the present invention can be used as and is used to assess the test of dynamic brain function and is used for the auxiliary Differential Diagnosis of carrying out.

In an example of the present invention embodiment, to MEG signal application Time series analysis method, with estimate between each neural colony because of different dynamic interaction of time, thereby prediction action behavior and musical sound (music), deriving synchronous neutral net related in task, and assess the AD of neural colony and the variation of chronic alcoholism.This Time series analysis method has been proved very practicality and has been expected to be used to estimate the state of brain function.

In an example of the present invention embodiment, service time sequence analysis without deriving the synchronous dynamic network average and rough single test, described single test is to use intercorrelation function (crosscorrelation function from a plurality of; CCF) during the eye gaze task, write down simultaneously from 248 MEG pick offs with 1 millisecond of short duration resolution.This kind analysis obtains the visual form of synchronous dynamic brain network closely similar and powerful between each health population.The dynamic property and the brute force of highdensity space samples, the network that is not capped show that it is suitable as the test that is used for assessing statically dynamic brain function.

Mark health volunteer on these networks with have AD or MCI and detoxifying during be between the experimenter of chronic alcoholism and on statistics, significantly distinguish.In addition, the CCF estimated value table of regional connectivity reveals the resolving ability that is enough to experimenter's individuality is divided into particular demographic (for example healthy group, AD group, MCI group).In addition, genetic search algorithm makes this kind taxonomic hierarchies can extend to other group, comprises schizophrenia and chronic alcoholism.

Genetic search algorithm or genetic algorithm (genetic algorithm; GA) be developed the chromosome array that is used for big and seek gene.Use genetic algorithm (for example to search for big locking phase mutual effect set in certain embodiments of the present invention, when using 248 pick offs is 30,628 interactions), to find the subclass that to predict the classification of disease of brain and state in these locking phase mutual effects.

Example 2

Method

The experimenter lies on the back in bed and a point being required to watch attentively its front reaches 1 minute.Require the experimenter to make its eyes keep watching this point attentively and can not blinking.Then, experimenter's its eyes that close reach other 3 minutes, finish this test thus.Use is carried out all analyses from the MEG data of watching the cycle attentively.In one embodiment, collected data are applicable to identification and remove signal illusion electric wave, for example heart illusion electric wave during the state of sleeping.

The MEG instrument

Use the MEG instrument to collect data.The experimenter lies on a bed in being subjected to the chamber of magnetic shield, and (0.1-400Hz takes a sample with the frequency of 1017Hz from 248 axial gradient meters in the whole persistent period (～4 minutes) of test, Magnes 3600WH, 4-D Neuroimaging, San Diego CA) collects the MEG signal.

Data analysis

Date processing.Can utilize event synchronization subtraction method to remove heart illusion electric wave.Because of the persistent period in eye gaze cycle is lacked (1 minute) very much, the illusion electric wave because of causing nictation can not appear in expectation, but if present, can be detected and remove from data or delete.

Time series modeling to the MEG data.Can benefit from inherent dynamic change in high of short duration resolution and the MEG signal removing the not average data of analyzing heart and/or nictation single test after the illusion electric wave, with prewhitening (soon the MEG time series converts immobilized white noise sequence to) afterwards by calculate these 248 pick offs all each between all intercorrelation functions assess functional interaction between the neural colony big in the set task.This is by utilizing autoregression integration rolling average (AutoRegressive IntegrativeMoving Average; ARIMA) analysis is carried out modeling to original series and is got residual and realize.Then, for all possible in these " prewhitening " sequences to calculating CCF.These CCF can be regarded as connective weight by these 248 pick offs as in the extensive interconnection neutral net of node.The part correlation that utilization is derived from zero lag CCF, structure synchronous dynamic network.Show positive and negative interactional example among Fig. 2.

In Fig. 2 A and 2B, the lines representative is subjected to the part correlation (through Fisher z conversion) of threshold.Fig. 2 A shows that positive part is relevant, and Fig. 2 B shows that then negative part is relevant.Exist 30,628 lines of drawing (that is, and 248 pick offs all possible to), but only show that those exceed the lines with lower threshold value.(Bonferroni inequality) adjusts the statistical significance threshold value according to the Bang Folunni inequality, to take into account 30628 multiple comparisons: nominal significance threshold value is p＜0.001, corresponding to threshold value p＜0.001/30628 (=p＜0.00000003) of reality use.Analysis is based on 45 seconds long time cycles and does not add average and carry out smoothly.

Identification and classification is analyzed

Implement to analyze according to selected measurement, to derive the identification and classification function of some experimenter group, then these identification and classification functions are applied to new case, it is divided in one of them original group.This analysis obtains the posterior probability and of the classification of each group and specifically measures (mahalanobis distance square value), and this concrete measurement is the distance of specific cases apart from each classification group.This measurement can be used for monitoring the potential variation of brain function, to approach the brain function of different groups.

For example, consider AD.Use is from experimenter (AD group) who is suffered from disease by diagnosis and the normal healthy controls person's that conduct is complementary experimenter's (C) data, and derive two linear discrimination classification functions from the SNI data: one of them is used for AD, and another is used for C.Make the new experimenter of the potential initial diagnosis result with mild cognitive ability obstacle (MCI) stand the SNI test.By using AD and C classification function, estimate new experimenter's health degree or AD degree.This assessment is not binary system assessment, but successive assessment, is measured by the mahalanobis distance square value apart from the center of AD group and C group among the typical discriminant function figure.These relative distances can be used for monitoring of diseases progress (experimenter will become more " as AD "), return (more " as C ") or get involved effect (after Drug therapy more " as C ").Nearness is meant relative distance, and does not need to show with graphic form for having implication.The linear discrimination classification analysis can be used for various states or disease, includes but not limited to for example AD, C and MCI data and the experimenter with dry syndrome.Can be by in one period, monitoring so that trend is detected.

The experimenter

In an example, the experimenter comprises the experimenter who has various medical histories, is in 13-90 year the range of age, has both comprised that the male also comprised the women and comprises all race and national groups.Subject matter of the present invention can be used for healthy experimenter and the experimenter who suffers from the disease of the system that can affect the nerves, and these diseases comprise that multiple sclerosis (MS), dry syndrome (SS) or other may relate to central nervous system's autoimmune function disease, MCI, AD, schizophrenia and chronic alcoholism.Embodiments of the invention can be used for having the experimenter of various brain states, for example are subjected to the people of ethanol or drug influence.

Data collection

In an example, utilize axial gradient meter systems (Magnes 3600WH, 4D-Neuroimaging, San Diego, CA) the collection MEG data of 248 passages.The low temperature helmet-shaped Dewar flask (dewar) of MEG is arranged in the chamber that is subjected to electromagnetic shielding to reduce noise.Frequency with 1017.25Hz is collected data (0.1-400Hz).For guaranteeing to prevent that the experimenter from moving, started continuously after five signal coils were digitized before MEG collects and reached before data collection, thereby head is positioned with respect to pick off.The MEG data comprise by～60, the time series that 000 value/each experimenter and pick off are formed.Can get rid of the illusion electric wave relevant by the repeated collection data with eyes.

Data analysis

The data pretreatment comprises and utilizes synchronous event subtraction method to remove heart illusion electric wave.After removing heart illusion electric wave, with time series " prewhitening ", so that can be in the interaction of estimating under the situation that can not cause deflection between each time series because of the self correlation of sequence self.Utilize Bai Kesi-Charles Jenkins (Box-Jenkins) ARIMA modeling to carry out prewhitening, come the of short duration structure of recognition data seasonal effect in time series corresponding to 25 intervals of ± 25 milliseconds to use.～60, implement this analysis on 000 time point.(comprise auto-correlation function (the autocorrelation function that calculates and estimate residual in ARIMA modeling and work up; ACF) and part A CF (PACF)) afterwards, use have 25 AR rank, the 1st jump divides and the ARIMA model of the 1st rank MA obtains with respect to the immobilized residual of meansigma methods, variance and autocorrelation structure.Use SPSS statistical package (SPSS that is used for Windows, Chicago, IL, 2000) to estimate residual.Utilize IMSL statistics storehouse (Compaq Visual FortranProfessional 6.6B version) the DCCF routine computes each to the zero lag intercorrelation between the static residual.In view of the above, for all the sensors to calculating part zero lag intercorrelation and the statistical significance thereof between i and j the pick off.For carrying out linear discriminant analysis, this intercorrelation is carried out the z conversion ,] so that its distribution normalization: z=0.5[ln (1+r)-ln (1-r).

Utilize the DDSCRM subroutine of IMSL Visual Fortran to carry out the linear discrimination classification analysis.Utilization is implemented this analysis through the crosscheck method of staying of exchange and verification, and utilizes the method for reclassifying to obtain the classification function of each group.For the classification of indivedual cases, computational discrimination score, posterior probability reach the mahalanobis distance square value with each group's quality centre of form (centroid).

Aspect is that identification is right with lower sensor: described pick off be to will correctly being classified into its group separately with case with zero uncertainty, be 1.0 and be 0.0 for the posterior probability of all other groups for the posterior probability of correct group promptly.Total available intercorrelation set is N=30,628 (between=248 pick offs all possible to).Suppose the predictive value set sizes for example for k=7, N=30 then, in 628 k=7 the number that might arrange be very large.Therefore, need certain special

Algorithm is searched for this huge space effectively and is discerned the set that can exactly the experimenter be divided to the k=7 of concrete group.In an example, use genetic algorithm.

Figure shown in Figure 3 shows the result that the linear discrimination classification of zero lag part intercorrelation is analyzed.Fig. 3 shows the classification chart that uses genetic search algorithm to obtain for 40 intercorrelations of use of selecting in 50.The centre of form of group is by assembling closely and isolating clearly and distinguish.

Example 3

Use prewhitening (static) magneticencephalogram signal with synchronous dynamic brain network visualization.In an example, collect the collection data from 248 axial gradients.At match autoregression integration moving average model(MA model) and after getting residual, calculate all paired zero lag part intercorrelation PCC between i pick off and j the pick off _IJ ^OThereby, provide between each neural colony the directly estimated value of synchronous coupled intensity and symbol (positive and negative) with 1 millisecond of short duration resolution.In an example, 51.4% PCC _IJ ^OBe positive, 48.6% PCC _IJ ^OBe minus.On average, positive PCC _IJ ^OThan minus PCC _IJ ^OOccur with short transducer spacing more continually, and than minus PCC _IJ ^OStrong by 72%.According to estimated PCC _IJ ^O, structure dynamic neural network (one of each experimenter) to show different features, comprises a plurality of local interactions.These features are powerful between each experimenter, and can be used as the source (blueprint) that is used to estimate dynamic brain function.

A purposes of full head magneticencephalogram (MEG) is that the neural activity source is positioned.Because this problem does not have unique settling mode, therefore measure (the shape of skull according to supposition (single still multiple source), reality, " just drilling modeling (forward modeling) "), specific analytical method and subjective judgment and decide, the result of these analyses can be different.In addition, data are filtered to ≈ 45Hz and following usually, and implement to analyze according to the meansigma methods of test many times.Although it is can be comparatively practical to utilize MEG that activity is positioned, yet other functional neuroimaging method provides more clear and definite information (promptly do not rely on supposition etc. information).These methods comprise fMRI and PET.With respect to the time course of of short duration resolution and cerebration variation, MEG and EEG have the edge.In these researchs, handle data from each single-sensor, average and make it in the particular event of being concerned about, to align to many tests usually, and the shape of review time process.Resulting MEG track (perhaps being the current potential relevant with incident in EEG research) provides the periodic valuable information of brain incident with respect to behavior.Similar approach at fMRI (design relevant) with incident although in practical, lack the of short duration precision of MEG and EEG signal.

According to embodiments of the invention, use full head, high density MEG data to study the interaction between neural colony.Interaction between the neural colony is from sleep and awakens to the basis of all brain functioies of higher cognitive process.Estimate these interactional intensity and spatial model can help in fact to understand brain function and with the relation of behavior.

Method

Ten right-handed experimenters of custom (five women five men) have participated in an experiment (the range of age: 25-45 year; On average ± and SEM, 33 ± 2 years old).

Produce stimulation and utilize the liquid crystal display projector to present to the experimenter by computer.The blue luminous point that the experimenter watches black screen central authorities attentively reaches 45 seconds.Utilize concavo-convex mirror system to present point of fixation, image is placed on the about 62 centimeters screen in experimenter's eyes the place ahead by concavo-convex mirror system.Utilize axial gradient meter systems (the Magnes 3600WH of 248 passages; 4D-Neuroimaging, San Diego) collection MEG data.The low temperature helmet-shaped Dewar flask (dewar) of MEG is arranged in the chamber that is subjected to electromagnetic shielding to reduce noise.Frequency with 1017.25Hz is collected data (0.1-400Hz).For guaranteeing to prevent that the experimenter from moving, started continuously after five signal coils were digitized before MEG collects and reached before data collection, thereby head is positioned with respect to pick off.Paired distance between the pick off is calculated as the geodesic line on the MEG helmet surface.Utilize electrooculography (electrooculography) record eye motion.For this reason, place three electrodes in position around each experimenter's right eye.Frequency with 1017.25Hz is taken a sample to electronystagmogram graphy figure signal.Collected MEG data comprise by～45, the time series that 000 value/each experimenter and pick off form.

In an example, analyze each to the interaction between the time series of pick off.For this reason, it is immobilized need making each independent sequence, i.e. " prewhitening "; Otherwise, the relatedness that the nonstatic of sequence itself can lead to errors.Therefore, analysis comprises the residual that time series is carried out modeling and derived static (or accurate static) to measure in order to calculate paired relatedness, for example intercorrelation.Below described analysis be to carry out to the unsmooth of single test and without average data.Carry out Bai Kesi-Charles Jenkins (Box-Jenkins) autoregression integration rolling average (ARIMA) modeling analysis, come the of short duration structure of recognition data seasonal effect in time series corresponding to 25 intervals of ± 25 milliseconds to use.On 45,676 time points, implement this analysis.At ARIMA modeling widely and work up (comprising the auto-correlation function and the partial auto correlation function that calculate and estimate residual) afterwards, judge and have 25 autoregression rank (equal ± 25 milliseconds interval), the 1st jump divides and the ARIMA model of the 1st rank rolling average is enough to obtain with respect to the in fact immobilized residual of meansigma methods, variance and autocorrelation structure.Use is used for SPSS 10.1.0 version statistical package (SPSS, Chicago) the estimation residual of Windows.Utilize DCFF routine (the Compaq VisualFortran Professional 6.6B version in international mathematics and statistics storehouse (International Mathematics and Statistical Library), Compaq Houston) calculates each to the zero lag intercorrelation between the static residual.According to these data, for all the sensors calculates part zero lag intercorrelation PCC between i and j the pick off _IJ ^OAnd statistical significance.For calculating descriptive statistic and other statistics, utilize Fisher ' s z conversion with PCC _IJ ^OBe transformed into Z _IJ ^O, so that its distribution normalization:

$Z_{\mathrm{IJ}}^O=0.5\left[\ln (1+{\mathrm{PCC}}_{\mathrm{IJ}}^O)\ln (1-{\mathrm{PCC}}_{\mathrm{IJ}}^O)\right].$

The result of exemplary is as follows: under situation with 248 pick offs, each experimenter can have altogether 248! / 2!=30,628 PCC _IJ ^O, thereby have 30,628 * 10 experimenter=306,280 PCC altogether _IJ ^OIn these are relevant, have the record of illusion electric wave nictation in eliminating after, analyzes 285,502 (93.2%) and is correlated with; During these are relevant 81,835/285,502 (28.7%) has statistical significance (P＜0.05).At all effective PCC _IJ ^OIn, 146,741 (51.4%) is positive, 138,761 (48.6%) is minus.Average (± SEM) positive Z _IJ ^OIt is 0.0112 ± 0.00004 (maximum $Z_{\mathrm{IJ}}^O=0.38;$ ${\mathrm{PCC}}_{\mathrm{IJ}}^O=0.36$ ); Average negative Z _IJ ^OBe-0.0065 ± 0.00002 (minimum $Z_{\mathrm{IJ}}^O={\mathrm{PCC}}_{\mathrm{IJ}}^O=-0.19$ )。The absolute value of these meansigma methodss differs very big (P＜10 ^-20Student t check), average |+Z _IJ ^O| than average |-Z _IJ ^O| exceed 72%.The example of the spatial model that the synchronous coupling between pick off and all other pick offs distributes is shown among Fig. 4 A-4C and Fig. 5 A-5C.Only be depicted in statistically evident PCC _Ij ^o(Bonferroni inequality) adjusts the statistical significance threshold value according to the Bang Folunni inequality, to take into account 247 multiple comparisons of each figure: nominal significance threshold value is p＜0.05, corresponding to threshold value P＜0.05/247 (being P＜0.0002) of reality use.Show positive and minus PCC _Ij ^oPoint is represented this 248 positions of pick off projection on the plane.Data are from an experimenter.

PCC _Ij ^oAnd the relation between the transducer spacing.Generally, PCC _IJ ^OWith between pick off i and the j apart from d _IjChange.Generally speaking, more close mutually pick off is tending towards having positive PCC _IJ ^OMinus z _Ij ^oAverage transducer spacing d _IjThan positive z _Ij ^oAverage transducer spacing d _IjLong 24%.Particularly, d _Ij(z _Ij ^o) be 198.92 ± 0.21mm (n=138,700), and d _Ij(z _Ij ^o) be 160.12 ± 0.24mm (n=146,675).Generally, at z _Ij ^oWith d through logarithmic transformation _Ij, i.e. ln (d _Ij) between exist strong and negative association that have very high explicitly.Signed z _Ij ^oWith ln (d _Ij) between Pearson's correlation coefficient (Pearson correlationcoefficient) be-0.519 (P＜10 ^-20).This kind relation shows that synchronous stiffness of coupling is tending towards sharply descending with transducer spacing.

The synchronous dynamic neutral net.PCC _Ij ^oBe the synchronous coupled estimated value between the neural colony, wherein absolute value and PCC _IJ ^ORepresent stiffness of coupling and kind respectively.If will be the node in the extensive interconnection neutral net (for example in Fig. 6-7, carrying out the neural network diagram of visual extensive connection), then PCC by the neural assembly that these 248 pick offs are taken a sample is stereoscopic _IJ ^OCan be used as the interactional estimated value of dynamic synchronization between these nodes.Can be by connecting these 248 nodes with line and showing that every line is that representative positive coupling or minus coupling are visual with this kind large-scale internetwork.The band threshold of this network and the view of ratio after Fig. 6 and 7 is presented at and averages among these 10 experimenters; Between the experimenter, have interactional regional the variation, and these variations are consistent.

The feature that is marked in this network comprises: (i) most of neighbours' interaction is positive; (ii) most of minus interactions occur in the distance situation far away; (iii) relative rare with the interaction of middle pick off; And (iv) the interaction between the cerebral hemispheres is not frequent, may be because related distance is far away.In addition, can change by the systematicness that hereinafter described (with counterclockwise) distinguishes interactional local density qualitatively.There are nine positive interaction zones (Fig. 6), form by the pick off that covers following brain zone: left front frontal lobe (1P), left back frontal lobe (2P), left side-frontal lobe-temporo (3P), left parietal bone of head (4P), left parietal bone of head-occipital bone (5P), right occipital bone (6P), right parietal bone of head-temporo (7P), right temporo (8P), and right frontal lobe (9P).For minus interaction ( Fig. 7), can distinguish several zones, form by the pick off that covers following brain zone: left front frontal lobe cortex (1N), left back frontal lobe (2N), left side-frontal lobe-temporo (3N), left parietal bone of head (4N), occipital bone (5N), right parietal bone of head (6N), and right frontal lobe (7N).In these positive and minus interactions several are spatially overlapping.

The brute force of network between the experimenter

Obviously, the neutral net of constructing as mentioned above closely similar between the experimenter (Fig. 8 A and 8B, and Fig. 9 A and 9B).Can be by all z of computing network _Ij ^oPearson's correlation coefficient between (that is all i to j pick off), all experimenters between quantification and evaluated total network similarity.Higher and highly significant (intermediate value=0.742 of the correlation coefficient that is obtained; Scope: 0.663-0.839; Relevant for all, P＜10-20; Degree of freedom＞20,000).These table of discoveries are understood common network foundation.

This exemplary embodiment is assessed the synchronous dynamic of utilizing the ARIMA modeling and becoming between immobilized each single test MEG time series and is coupled.The result who is obtained be this kind coupled be not subjected in the original MEG data the estimated value of the nonstatic contact scar that exists usually.Use the synchronous coupling between two time serieses of zero lag intercorrelation estimation.Relevant according to these, calculating section is relevant, thereby the direct-coupled symbol between two sensor sequence and the estimated value of intensity are provided, because be removed by may influencing of causing indirectly of other pick off.In addition, PCC _IJ ^OAllow the following dynamic network synchronously of structure: PCC wherein _IJ ^OSymbol and intensity as the estimated value of the coupled symbol of direct neural colony and intensity.Generally speaking, PCC _IJ ^OLess than primary intercorrelation, because taken into account all other 246 possible associations in all sensor combinations.Yet the interaction of small magnitude is the stability rule in the large-scale internetwork.For example, formerly in the research to this kind network, normalized bonding strength is from-0.5 to 0.5 beginning when network trained, and is from approximately-0.2 to 0.2 in the stabilizing network that has trained.By contrast, utilize subject matter of the present invention, PCC _IJ ^OScope then be similar (0.19 to 0.36).According to specific pick off to and the distance decide PCC _IJ ^OFor positive or minus, and be different intensity, thereby make PCC _IJ ^OBe tending towards becoming big more in short-term in distance.This kind trend may be to check identical neural source and cause owing to a plurality of detectors.Although there is not the quantitative measure of this kind factor, its not leading associative mode of being seen of results verification.Particularly, magnetic field intensity will make the associative mode of signal very compact owing to this kind factor with the quick decline of distance and the influence of gradiometer coil, can't see distance and interact on identical ratio.Yet, the analysis showed that the complexity interaction pattern on the whole cortex is visible on single amplitude proportion.For close pick off, shown in relevant grow, but neural activity is generally more relevant partly.The calculating of part correlation is owing to eliminate potential pseudo-effect.

In the research formerly, utilized to whole data acquisition system or in the specific frequency spectrum frequency band and used frequency domain or time-domain analysis, the association between each neural population (being registered as EEG, MEG or local field potentials) has been studied.In these are analyzed, do not test its static property according to the data computation measurement of correlation usually.In the two, static property (or accurate static property) provides between the time series because of constantly different interactional accurate measurement (contrasting with total trend and/or circulation) in time domain (by calculating intercorrelation) and frequency domain (by calculating the concordance square value).Intercorrelation or concordance estimated value based on original nonstatic data can obtain erroneous estimate and spurious correlation.

The symbol of intercorrelation does not provide the information about basic zest or inhibitory synapse mechanism, but the kind of co-variation (simultaneous covariation) when only representing: the unidirectional co-variation (increase/increase of positive relevant expression with respect to serial mean, reduce/reduce), and the rightabout co-variation of minus relevant expression (increase/reduce, reduce/increase).Generally speaking, PCC _IJ ^OBe tending towards in the pick off space changing with orderly fashion, make its be tending towards between the adjacent sensors for positive away from pick off between for minus.Although this kind trend is well-known, yet also there is other obvious and different exception, comprises the PCC between the adjacent sensors _IJ ^OFor minus away from pick off between PCC _IJ ^OFor positive.In addition, decide space PCC on the position of reference sensor _IJ ^OPattern is difference to some extent.Result of study shows to have brute force and fasten orderly relational structure in the pass, but has different local specificity.In fact, these characteristics are base attributes, and it makes resulting large-scale internetwork have the peculiar structure shown in Fig. 6-9B.

A feature of this structure is the regionality variation that whole network is divided into the interactional intensity of plus or minus.Describe these mixed interactions and attempting utilizing electric current density for example or beam forming technique the interaction between Naokong is positioned is another step in this kind method.

Heart illusion electric wave removes algorithm

Can utilize to comprise for example following various programs, remove heart illusion electric wave:

Select representational heart beating as starting template.In an example, use previous template from different experimenters of preserving.Then, segmental related the matching of overlapped data (or slipping over described data) that whenever makes this representativeness template and this template and in each step, calculated next point.Use resulting correlation time process to determine the position of heart beating.When described relevant when surpassing threshold value, the local maximum place record heart beating in the correlation time process.

For reducing because of noise causes the probability that mistake is surveyed, minimum and maximum heart rate are selected to form the time window with respect to last heart beating, in this time window, expect heart beating each time.Ignore the peak value in relevant before this window begins, and get the heart beating at the highest correlation peak place in this window, rather than satisfy the heart beating first time of threshold value.If there is not the peak value that satisfies threshold value in window, then there is the heart beating of losing in supposition and expands this window till it comprises the peak value that satisfies threshold value.At each experimenter, the two is adjusted to time window and dependent thresholds so that real survey maximization and mistake surveyed minimize.

If survey undesirablely, then average the template that forms improvement, and carry out the correlation computations second time with the template of improvement by heart beating to actual detection.In second time, surveying generally can sufficiently complete, makes to carry out more that the detection of multipass will be unfavorable for improveing template again.After second time, check heart rate track, and preserve the detection realize complete (or approaching complete) and do not have the high threshold of the mistake detection of significant amounts with a series of dependent thresholds.Then, use these to survey and form final average cardiac waveform.Get one second data slot, make it comprise whole heart beating around this detection.Heart beating with strong noise is not contained in the meansigma methods, and makes waveform DC move to zero mean after reaching before the heart beating of expansion.Thus, form the average waveform of each channel, wherein only be used to signal from this channel.Then, when detecting heart beating each time, from channel, deduct average waveform.If heart rate is higher, then afterbody excision to the heart beating next time with waveform just begins the place.The result can remove heart illusion electric wave and can not introduce wrong being correlated with at interchannel.

If the individual data file is less than a few minutes, then by fetching the weighted mean from a plurality of average waveform of same record phase, the average waveform that is used in subtraction is more clean.

MEG seasonal effect in time series prewhitening

If time series is static in fact, promptly in the sequence meansigma methods of different time points and variance can not change and self correlation (ACF) and partial auto correlation function (PACF) smooth, but the relation between the evaluation time sequence then.If do not satisfy these conditions, then can be because of previous X value influence, the trend that may exist and the preceding value of given X value introduced on currency during for time t noise, and acquisition error result.

In an example of subject matter of the present invention, come deal with data by prewhitening.Prewhitening is meant these dependences that removes as indicated above.Prewhitening also can (but nonessential) need guarantee that these dependences are effectively removed.

In an example, prewhitening comprises that use autoregression integration rolling average constituent element carries out modeling to time series, and gets residual.

Because each time series all is unique, thereby may be loaded down with trivial details, complicated and multiple process to the seasonal effect in time series modeling, relate to the important constituent element of (a) model of cognition, (b) estimate modeling coefficient and stability thereof, (c) get residual, and the static property of (d) assessing residual.

Carry out the identification of time series models by the shape of checking ACF and PACF.Can carry out the estimation of modeling coefficient differently, especially those relate to the coefficient of rolling average constituent element.The estimation of residual is directly carried out, as utilizes ACF and the PACF assessment residual.If residual is static inadequately, then change model parameter etc. to repeat this process, till realizing static property.

Because of this modeling relates to combination of (a) various factors (autoregression, difference, rolling average) and (b) a plurality of potential grade in each factor (for example autoregression exponent number, difference number, and the exponent number of rolling average), and therefore length (～60 is considered in the place, 000 time point) indivedual time serieses come down to unique, thereby do not have the single rule that is used to realize static property target.

For this reason, can use some kinds of different calculating to search the combination of satisfying the requirement of static property.As removing of heart illusion electric wave, estimate at the same time obtained as a result the time relate to the judgement element.Although the mathematical description to elementary operation (for example difference) is directly, yet determines that suitable combination meeting is more loaded down with trivial details.

In this process, several platforms be can use, FORTRAN, MATLAB, SPSS statistical package and BMDP statistical package comprised.Concrete sequence is depended in the actual employing of platform specific.Regardless of selected particular combinations, target all is to obtain enough static property.

Example 4

The neural associated subset of this examples show part zero lag is differentiated the experimenter and it correctly is divided into the ability of six different nervous physiology state groups.

Method

52 experimenters have participated in research as the volunteer who charges.6 groups are arranged, comprise healthy collator and experimenter with Alzheimer's disease, schizophrenia, multiple sclerosis, dry syndrome and chronic alcoholism.Shown in each group composed as follows: Alzheimer's disease (N=6 male, 76.8 ± 1.8 years old age, meansigma methods ± SEM); Schizophrenia (N=9[7 male, 2 women], 48.2 ± 2.9 years old age); Multiple sclerosis (N=4[2 male, 2 women], 42.5 ± 6.4 years old age); Dry syndrome (N=4 women, 56.3 ± 5.2 years old age); Chronic alcoholism (N=3 male, 57 ± 0.9 years old age); Health volunteer's (contrast group) (N=25[17 male, 8 women], 47.0 ± 3.6 years old age, scope 23-82 year).

The experimenter who belongs to an experimenter group has the functional brain disease, and its diagnosis is to be undertaken by the expert of medical domain separately.The experimenter who suffers from chronic alcoholism did not drink in 24 hours before the research beginning and to use breath analyzer to test soft.Contrast group comprises experimenter and other the healthy experimenters who is complementary with each experimenter group age.All experimenters except belonging to those that contrast group, all take the medicine relevant with its disease of brain; In these medicines some be used for the treatment of psychotic.

Data collection

The experimenter lies on the back on the MEG instrument and a point at its front of eye gaze～62 centimeters places reaches 45-60 second (for different experimenters), meanwhile, from 248 axial gradient meter (0.1-400Hz, frequency with 1017Hz is taken a sample, Magnes 3600WH, 4-D Neuroimaging, San Diego CA) collects the MEG data.Obtain each experimenter's data acquisition system thus, described data acquisition system is made up of 248 time serieses with 45000-60000 time point.As indicated above, utilize the event synchronization subtraction to remove heart illusion electric wave in each sequence.

Data analysis

All of the following stated analyze all be to single test without level and smooth and carry out without average data.After the prewhitening that utilizes Bai Kesi-Charles Jenkins (Box-Jenkins) ARIMA modeling time of implementation sequence, to (N=30628), calculate the part zero lag intercorrelation PCC between i pick off and j the pick off (N=248 pick off) for all the sensors _Ij ^oUtilize Fisher ' s z conversion with PCC _Ij ^oBe transformed into z _Ij ^o, with its distribution normalization:

$z_{\mathrm{ij}}^o=0.5[\ln (1+{\mathrm{PCC}}_{\mathrm{ij}}^o)-\ln (1-{\mathrm{PCC}}_{\mathrm{ij}}^o)].$

Then, judge whether to exist and the experimenter correctly can be divided into its z of group separately _Ij ^oSubclass.For this reason, utilize the powerful crosscheck method of staying, and require each experimenter 100% correctly is divided to its corresponding group to accept z _Ij ^oGiven subclass as the good classification factor, thereby implement linear discriminant analysis.

A kind of brute force method that is used to discern all these subset will be possible and heavy on calculating.For example, for the subclass that forms by 5 predictor, the sum of possible combination will for:

$\frac{(30628!)}{(5!)(30623!)}=224527811140901268000$

For this reason, utilize genetic algorithm to shorten computation time and make the search optimization.By genetic algorithm, choose z randomly by a certain quantity _Ij ^oThe initial subclass that predictor (coming from 30628 obtainable predictor) forms, and make group size=5 and the even little GA that intersects move 24 hours.If at searching period, the separating of fitting function for 2 * 10 ⁵In individual generation, is invariable, then chooses new random subset and repeats this computing.

The result

Utilize genetic search algorithm, easily find each experimenter 100% correctly to be divided to the z of group separately _Ij ^oSubclass.The instance graph (mapping to two-dimensional space) that in Figure 10, shows this kind division.In the CDF space of 2 typical discriminant functions of pro-(CDF), clearly pick out all group's zero laps.

CDF be the predictor variable weighting linearity and, and obtain by statistical analysis.For L group (supposition L is less than the predictor quantity k in the subclass), there be j=L-1 discriminant function CDF _i, i=1 wherein, 2 ..., j.Supposing has six groups in these data, then have five CDF when k＞L all the time.For given predictor subclass, pass through the statistical significance that Wilk ' s Λ test statistics is assessed i=1 to 5,2 to 5,3 to 5 etc. CDF.For i=1 to 5 (P＜10 ^-22), 2 to 5 (P＜10 ^-6) and the CDF of 3 to 5 (P=0.004), Wilk ' s Λ highly significant; For the CDF of 4 to 5 (P=0.16) and CDF 5 (P=0.71), then not remarkable.The experimenter correctly is divided in its group separately by 100%.At last, utilize the multivariate analysis of variance to test original initial data 20 dimension z _Ij ^oThe equality null hypothesis of the centre of form of these six groups in the predictor space (null hypothesis of equality), and in higher (P＜10 of significance level ^-31, Hotteling ' s track check) time refuse described equality null hypothesis.

The chance test

The quantity of Yu Qi qualified subclass depends on the quantity k of the actual prediction factor in the subclass at random, institute might predictor z _Ij ^oWhole size, group's quantity and with the experimenter's that is classified quantity.Suppose that M experimenter belongs to L group and can have the subclass of the identical size that is formed by k predictor in the individual predictor of N (=30628), then will make each experimenter by 100% s of anticipated number at random that correctly is divided to the subclass of its group be:

$s=Q{\left(\frac{1}{L}\right)}^M,$ $Q=\frac{N!}{k!(N-k)!}$

This formula supposition test size be all possible subclass Q of k, under the situation of given bigger N value (=30628), in addition less k value, this also is a kind of task of in fact impossible realization.Yet, this problem is studied in following tractable mode.First kind of analysis and utilization the following fact: for less k, L and M, the limit search is feasible.Under k=2, N=30628, L=3 and M=17 (6 experimenters suffer from Alzheimer's disease, and 3 experimenters suffer from chronic alcoholism, 8 contrast experimenters for being complementary) situation, carried out this kind search.With expection at random～4 subclass (s=3.63) compare, the assessment of this limit obtains 560 makes each experimenter all by 100% subclass of classification correctly.Clearly (P＜10-50), this shows the good set that has excessive (being higher than at random) to difference to these ratios (in Q=469021878 all possible subclass) for binomial theorem, normal deviate z=23.4.At last, following analysis is implemented in the more big collection that can't implement the limit search.For k=10, N=30628, L=6 and all M=52 experimenter, the expection number of subsets quality entity that can obtain 100% correct classification is zero (s=0.0069).Yet, utilize the identification and classification program of genetic algorithm behind the operation several hrs, to obtain the correct set of classifying in 79 energy 100% ground.Although the accurate ratio of good set in can't calculating data (because of the limit search infeasible), yet this ratio has surpassed expection (z=8.78, P＜10 at random ^-50).In many cases, successful subclass not only obtains 100% correct classification, but also obtains each experimenter correctly is classified to the high posterior probability (for example＞0.98) of its group.

Discuss

The one aspect of the present invention on the basis of classifying as succeeing in this example is the paired zero lag part of a MEG seasonal effect in time series intercorrelation.Characterize dynamic synchronization in the big neural network by the relevant set that forms of all these kinds between the signal of these 248 MEG pick offs.The result shows, can adjust these synchronicitys because itself in addition at the morbid state that also can differentiate brain than the change in the smaller subset very doughtily.This idea is very similarly to observe between the health volunteer and according to the wider notion of synchronicity as the basis of higher brain function according to the part correlation brain connection pattern.

For with the differentiation of electrophysiology brain connection pattern and classification application in aspect health and disease lay foundation, utilize quantitative EEG early stage work experience very long road.From conceptive, the method for certain embodiments of the invention has followed those early stage leading methods, but in fact it is substantially also inequality, because (i) it utilizes measuring technique (MEG is with respect to EEG) more accurately; (ii) it is based on single test (with respect to average test of many times); (iii) its basic tool is the relation measurement between the sensor signal (intercorrelation), rather than (for example spectrum power in the special frequency band) of primary (for example signal amplitude) in the separated sensor or derivation measured; And (iv) this kind intercorrelation is to calculate (after with original nerve signal prewhitening) from immobilized time series, so its reflection is real interacts because of constantly different nerve.In addition, (v) the intensity of synchronicity is to measure with the high of short duration resolution that is about (1 millisecond); And (vi) given right interaction is separated by the remainder with neutral net, thereby resulting part zero lag intercorrelation is not subjected to the pollution of synteny.

Example 5

In this example, in 142 experimenters altogether, test the classification factor of the interaction of neural colony, and estimate its probability and check several disease of brain as biological marker as the brain state.For this reason, at first, study 52 experimenters to derive the identification and classification function.Then, externally in the validation-cross program, these functions are applied to the new group that forms by 46 experimenters.At last, incorporate the full sample of other 44 experimenters into to obtain forming by 142 experimenters.Neural synchronously the interaction successfully carried out classification and provided excellent external certificate result the brain state.

Material and method

142 experimenters have participated in research as the volunteer who charges altogether.7 groups are arranged, comprise healthy collator (HC), Alzheimer's disease (AD) patient, schizophrenia (SZ) patient, chronic alcoholism (CA) patient, dry syndrome (SS) patient, multiple sclerosis (MS) patient and face ache (FP) patient.Shown in each group composed as follows: HC (N=89[48 male, 41 women], age [meansigma methods ± SEM] 43.7 ± 1.7, scope 10-82 year); AD (N=9 male, 74.0 ± 2.1 years old age, average trickle mental status detected [MMSE] score 21.13 ± 1.5); SZ (N=16[13 male, 3 women], 45.8 ± 2.5 years old age); CA (N=3 male, 57.3 ± 0.9 years old age); SS (N=10[1 male, 9 women], 54.8 ± 3.2 years old age); MS (N=12[4 male, 8 women], 40.7 ± 3.3 years old age, secondary progress type or recurrence remission form); FP (N=3 women, 47.3 ± 6.5 years old age, arthralgia).Each experimenter who belongs to a patient group has the functional brain disease, and its diagnosis is to be made as follows by the expert of medical domain separately.AD patient diagnoses and is determined according to cross-cutting common recognition diagnosis meeting to satisfy following standard: (i) according to DSM-IV[7] diagnosis of dementias and (ii) according to the possible or AD patient likely of NINCDS-ARDA standard.SZ patient diagnoses according to the DSM-IV standard, there is not the electric shock treatment history, there is not head trauma (be in hospital all night or lose consciousness) greater than 5 minutes, be difficult toward substance depilatory, do not have current material/alcohol dependence or abuse, and can not influence central nervous system's medical conditions (for example epilepsy).CA patient did not drink in 24 hours before the research beginning and uses the breath analyzer test soft.SS patient diagnoses according to the U.S.-Europe common recognition dry syndrome criteria for classification that team formulated.They claim by its doctor with by neural psychometry to be suffered from cognitive dysfunction by clinical confirmation.MS patient satisfies through the McDonald standard of revising [10], has the cerebral lesion more than or equal to 10T2, worsen once again or the steroid burst after through at least 30 days, and have tangible MS subtype.FP patient is suffered from temporomandibular arthralgia and masticatory muscles muscular fasciae pain (arthralgia) by diagnosis.At last, contrast group comprises experimenter and other the healthy experimenters who is complementary in years with each patient group.All experimenters except belonging to those that contrast group, all take the medicine relevant with its disease of brain; In these medicines some be used for the treatment of psychotic.

For ease of carrying out outside cross validation, according to o'clock two successive experimenter subsamples being analyzed with the irrelevant random time of data analysis.First sample comprises 52 experimenters (6 groups) and is made up of following group: HC (N=25[17 male, 8 women], age 47.0 ± 3.6, scope 23-82 year); AD (N=6 male, 76.8 ± 1.8 years old age); SZ (N=10[7 male, 3 women], 48.2 ± 2.9 years old age); CA (N=3 male, 57.3 ± 0.9 years old age); SS (N=4 women, 56.3 ± 5.2 years old age); MS (N=4[2 male, 2 women], 42.5 ± 6.4 years old age).Second sample comprises 46 experimenters (5 groups), and its data are to handle after first sample.This sample is made up of following group: HC (N=33[15 male, 18 women], age 36.8 ± 2.8, scope 11-67 year); AD (N=2 male, 76.0 ± 3.0 years old age [73,79]); SZ (N=2 male, 30.0 ± 2.0 years old age [27,33]); SS (N=5[1 male, 4 women], 51.4 ± 4.5 years old age); MS (N=4[2 male, 2 women], 36.8 ± 5.2 years old age).

Task-data collection

Target herein is to make brain be in steady statue and do not participate in any specific tasks.For this reason, the experimenter lies on the back on the MEG instrument and a point at its front of eye gaze～62 centimeters places reaches 45-60 second (for different experimenters), and meanwhile, (frequency with 1017Hz is taken a sample, and 0.1-400Hz filters from 248 axial gradient meters; Magnes 3600WH, 4-D Neuroimaging, San Diego CA) collects the MEG data.Obtain each experimenter's data acquisition system thus, described data acquisition system is by having 45,000-60,248 time serieses compositions of 000 time point.Utilize the event synchronization subtraction to remove heart illusion electric wave in each sequence.

Conceptual data is analyzed

All of the following stated analyze all be to single test without level and smooth and carry out without average data.For calculating the zero lag intercorrelation between the MEG sensor time sequence, make each independent sequence become static by " prewhitening ", because the relatedness that the nonstatic in the sequence can lead to errors.Therefore, the first step of described analysis be residual that time series is carried out modeling and derived static (or accurate static) to measure in order to calculate paired relatedness, intercorrelation for example.Previous work shows that the ARIMA model with 25 AR rank, the 1st jump branch and the 1st rank MA is enough to obtain with respect to the in fact immobilized residual of meansigma methods, variance and autocorrelation structure.Use SPSS statistical package (SPSS that is used for Windows, the 15th edition, SPSS Inc., Chicago, IL, 2006) to estimate residual.Utilize the DCCF routine (Compaq Visual FortranProfessional 6.6B version) of IMSL staqtistical data base to calculate each to the zero lag intercorrelation between the static residual.In view of the above, calculate part zero lag intercorrelation PCC between i and j the pick off for all the sensors _Ij ^oAnd statistical significance.For calculating descriptive statistic and other statistics, utilize Fisher ' s z conversion with PCC _Ij ^oBe transformed into z _Ij ^o, so that its distribution normalization:

$Z_{\mathrm{ij}}^o=0.5[\ln (1+{\mathrm{PCC}}_{\mathrm{ij}}^o)-\ln (1-{\mathrm{PCC}}_{\mathrm{ij}}^o)]$

At each sample, the right data of each separated sensor are carried out covariance univariate analysis (univariateanalyses of covariance; ANCOVAs), z wherein _Ij ^oBe dependent variable, sex (binary variable) and age are covariants.For assess each pick off in first and second sample between the congruence of distribution of group's effect, will be to given pick off to existing or not existing remarkable effect to be encoded into 1 and 0 respectively, and calculate χ ²Test statistics.

Linear discrimination classification is analyzed

This analysis is used to judge whether to exist the experimenter correctly to be divided to the z of group separately _Ij ^oSubclass.In this is analyzed, in typical discriminant function (CDF) space, carry out the differentiation of Fisher ' s group and experimenter's classification.CDF be the predictor variable weighting linearity and, and obtain by statistical analysis.For L group (supposition L is less than the predictor quantity k in the subclass), there be j=L-1 discriminant function CDF _i, i=1 wherein, 2 ..., j.In this multidimensional CDF space, implement group and differentiate (and each independent experimenter's classification).The posterior probability that this analysis obtains group's classification function and each experimenter is divided to particular demographic.In addition, to the whole sample application forward substep linear discriminant analysis (the program 7M of BMDP Dynamic, the 7th edition, statistical package, Los Angeles, CA, 1992) of 142 experimenters' formation, to derive single predictor subclass.Use this program default F value (F-to-add-a-predictor=4.0, F-to-remove-a-predictor=3.996).The input predictor of this analysis is from 271 z that pick off is right _Ij ^oValue, it shows the group's effect (P＜0.001, F-test) that has highly significant in ANOVA.Doing like this is the bigger predictor space of being made up of 30,628 values in order to reduce.

Genetic algorithm (GA)

The main purpose of this example is to identify the predictor subclass of success from very large space.In the size of predictor set very under the situation of big (N=30,628), even for for several predictor, a kind of brute force method that is used to discern all these subset also is to make us unacceptable on calculating.For this reason, utilize genetic algorithm to shorten computation time and make the search optimization by the following stated.Choose z randomly by a certain quantity _Ij ^oThe initial subclass that predictor (from obtainable 30,628) forms, and make group size=5 and the even little GA that intersects move 24 hours.If at searching period, the separating of fitting function for 2 * 10 ⁵In individual generation, is invariable, then chooses new random subset and repeats this computing.In fact, GA follows the natural selection rule of " survival of the fittest ".Thus, maintenance can improve the characteristic of Function of Evaluation (being classification in this kind situation), and the characteristic that can not improve Function of Evaluation then is rejected.

The statistical analysis of first sample

These analyses have two purposes, and promptly the quantity of (i) test good predict factor subclass surpasses this hypothesis of quantity of expection at random, and (ii) produces classification function so that the 2nd sample carried out validation-cross.

For result at random, suppose that M experimenter belongs to L group and can exist by N (=30,628) subclass of the identical size of k predictor formation in the individual predictor then will make each experimenter by 100% s of anticipated number at random that correctly is divided to the subclass of its group be:

$s=Q{\left(\frac{1}{L}\right)}^M,$ $Q=\frac{N!}{k!(N-k)!}$

For the analysis of using the limit search, (6 experimenters suffer from Alzheimer's disease to select k=2, L=3 and M=17,3 experimenters suffer from chronic alcoholism, 8 contrast experimenters for being complementary), carry out linear discriminant analysis to utilize the powerful crosscheck algorithm that stays.For keeping concrete subclass, each experimenter 100% correctly need be divided to its corresponding group.

Linear discrimination classification is analyzed: outside validation-cross

Linear discriminant analysis obtains the classification function of each group, uses these classification functions that each independent experimenter is divided to a certain group then.For carrying out outside validation-cross, use from the classification function of first sample derivation the experimenter from second sample is classified.For this reason, use can be carried out the classification function of 100% correct classification and be applied to second sample first sample.

The result

First sample (52 experimenters, 6 groups): initial analysis

At first, found z 18% pick off centering _Ij ^oStatistically evident group effect (P＜0.05, F test, ANCOVA).Then, to z _Ij ^oSubclass implement linear discrimination classification analysis (utilizing GA), whether each independent experimenter successfully can be divided to its group separately to find out.In fact, found many (thousands of) this kind energy 100% correctly to divide each z among these 52 experimenters _Ij ^oThe predictor subclass.(in fact can't determine the definite quantity of all these subset).The one example is shown among Figure 10.In many situations, successful subclass not only obtains 100% correct classification, but also obtains each experimenter correctly is divided to high (for example＞0.98) posterior probability of its group.

Owing to may have many z _Ij ^oSubclass, thereby the quantity whether quantity of learning successfully subclass surpass expection at random will be rather useful.This quantity is very big usually, and depends on the possible z of quantity k, institute of the actual prediction factor in the subclass _Ij ^oThe whole size (N=30,628) of predictor, group's quantity and with the experimenter's that is classified quantity.We study this problem in two kinds of traceable modes.In first kind of analysis, we only use 2 predictor that the more boy sample of data has been carried out limit search (referring to " method " part).With expection at random～4 subclass (s=3.63) compare, the assessment of this limit obtains the subclass (utilizing the crosscheck algorithm that stays of brute force) that 560 energy 100% correctly classifys.These ratio difference are (binomial theorem, normal deviate z=23.4, P＜10 clearly ^-50), this shows the good set that has excessive (being higher than at random).In second analyzed, we had used whole first sample that is formed by 52 experimenters, use 10 predictor, and this will be infeasible for the limit search.The expection number of subsets quality entity that can obtain 100% correct classification at random is zero (saying to be 0.0069 definitely).Yet our identification and classification program obtains the set of the correct classification in 79 energy 100% ground after the meeting in operation one.Although can't calculate good set in our data accurate ratio (because of the limit search infeasible), yet this ratio obviously surpasses expection (z=8.78, P＜10 at random ^-50).

Second continuous sample (46 experimenters, 5 groups): outside validation-cross

For the brute force of estimating this analysis with and whether can be used as suitable clinical trial, in the time period subsequently after first sample to handling from 46 experimenters' data.Particularly, the whether consistent in the following areas answer of the analysis result of seeking second sample: (a) by ANCOVA assessed for each right independent z of pick off with the result of first sample _Ij ^oThe distribution of group's effect of (N=30,628), and (b) second sample carried out sorting result (outside validation-cross) according to the classification function of deriving from first sample.For the former, to compare with 18% in first sample, the pick off centering 11% is found z _Ij ^oStatistically evident group effect (P＜0.05, F test, ANCOVA); Each this kind of pick off centering effect be distributed in these two samples very consistent (P＜10 ^-11, χ ²Test).For the latter, in the time will being applied to second sample from the classification function that first sample calculation goes out, many z that in first sample, provide 100% classification _Ij ^oSubclass still provides excellent classification score (＞90%, in thousands of).These results have stressed the similarity of these two samples and have proved and had excellent outside validation-cross result.

3.3 total current sample (142 experimenters, 7 groups)

To analyzing, to obtain 142 experimenters' current sample (7 groups) altogether from other 44 experimenters' data.Identified and manyly each experimenter 100% correctly can be divided to its z of group separately _Ij ^oThe predictor subclass.The quantity of these subclass is thousands of (under 20 predictor situations), and even has only 16 z _Ij ^oPredictor can provide 100% correct classification results.Also it should be noted that in most of situations posterior probability＞0.95 of experimenter's classification, this outstanding ability that this method has been described.At last, obtained by 12 z by substep discriminant analysis (referring to " method " part) _Ij ^oThe subclass that predictor forms, described subclass correctly is divided to its group separately with 86.6% among these 142 experimenters.For this set, also obtain two validation-cross results.At first, provide 78.9% correct classification by jack knife (Jackknifed) classification of staying a crosscheck algorithm to be obtained.The second, utilize 80% in the described data (selecting at random) to make this program operation 10 times, with the calculating classification function, and use described classification function to predict all the other 20% experimenters' group allocation.Average correctly be categorized as 86.4% (scope: 79.3-93.8%), and average correct jack knife classification is 77% (scope: 72.1-83.6%).These discoveries show the correct classification that can obtain high percentage ratio powerfully.

Discuss

Be used for the weak synchronized mechanism of local cortex and can be dependent on the return property side shoot (recurrent collateral) of pyramid bundle cell (tract cell) and the concrete thalamacortical neuron of parvalbumin immunoreactivity, and the thalamacortical neuron of calbindin immunoreactivity can be responsible for large-scale many focuses cortex synchronization.These discoveries show that the synchronicity of fine particle size can be the basic sides of cortical functional, and described cortical functional can be subjected to the difference of various disease process and destroy, thereby produce the feature corresponding to disease specific.

One different problem relates to the concrete subclass of the zero lag part correlation that can obtain the high-class rate.Under the condition of the combinatorial property of large space with 30,628 values and subclass problem, these subclass can't find by the limit search.But, adopt several diverse ways to discern and estimate this kind " good " subclass.At first, we use the classical theory of statistics, use the substep linear discriminant analysis and discern single predictor subclass.Because whole set is very big, therefore we reduce this set in the following manner: at first with disease (" group ") as immobilisation factor, implement variance analysis, then the individual predictor of 271 (among 30,628) of group's effect of demonstrating highly significant is carried out the substep linear discriminant analysis.This analysis obtains the set be made up of 12 predictor, and these 12 predictor are stayed in the classification (jackknifed leave-one-outclassification) and have high classification rate in 80/20% random division in standard classification analysis again and at jack knife.Yet, as any stepwise procedure, this substep linear discriminant analysis relies on concrete standard to import and/or remove predictor from equation in each step, and these standards also have main influence to the result except the direction (forward or backward) to stepping has the influence.Although this analysis is useful, may not be suitable for our concrete application best.For this reason, at first attempting the predictor subclass of powerful identification desirable (100% classification), promptly discerning by in whole associativity predictor subclass space, searching for.We have focused on little sub-set size (＜20) and have gone up to avoid overfitting.Make this program operation after several days, the initial analysis that we utilize random search to carry out does not produce any interesting result.Therefore, carry out genetic search algorithm and promptly locate ideal set.In fact, this algorithm has obtained a large amount of these subset in one day.This quantity has surpassed the quantity of expection at random, because find that (a) is from the limit search of the data of several diseases and several predictor, and (b) therein the quantity of perfect forecast factor subclass surpass in the larger samples of random amount, although can't determine the definite quantity of these subclass.Next step is to estimate the ability that this kind perfect forecast factor subclass is externally classified to new experimenter in the validation-cross scheme.By genetic algorithm being focused on this problem, identify thousands of the subclass (＞90% that can obtain excellent validation-cross rate; Hundreds of＞95%).Because the search volume is bigger, thereby can't learn the definite quantity of these subclass.

In a word, these results prove that (a) exists enough information to be used to distinguish various disease of brain states in the zero lag part correlation, (b) can utilize linear discriminant analysis successfully to extract this information, (c) these results have only exceeded the quantity that expection at random obtains, and (d) these results are powerful and pass through validation-cross to a great extent.Should point out clearly no matter these researchs are to be in which kind of stage, all in evolution always, because will need the classification function that upgrades the predictor subclass and be associated when adding new research experimenter and new disease group inevitably.In addition, can attempt and/or develop other sorting technique (for example according to support vector machine (support vector machine)), to improve classification results.At last, should point out, although above-mentioned analysis is to be applied to 7 groups, it is arbitrary to group that yet it also can usually be applied to, with for example as filler test (healthy collator still be all patients) or as ancillary method more specifically between specific disease of brain, to carry out differential diagnosis (for example being MS or disease of similar MS or the like).

Other example

Neural activity in the brain can produce magnetic signal and two kinds of signals of the signal of telecommunication.Magnetic signal corresponding to brain can utilize magneticencephalogram (MEG) pick off to survey, and the signal of telecommunication then can utilize electroencephalogram (EEG) pick off to survey.Electromagnetic transducer as herein described can be used for surveying the signal of telecommunication or magnetic signal.

Except that MEG and EEG pick off, also can use other form to collect of short duration data from brain.For example, Functional MRI (fMRI) is a kind of form that is used to provide corresponding to the data of the electron spin behavior during the inherent specific activities of body.Positron emission tomography (positron emission tomography (PET)) is that another kind is used to survey the form from the gamma ray radiation of introducing intravital radioactive substance emission.Computed tomography (computed tomography; CT) be the another kind of form that produces data according to the X ray that is scanned.Data from these forms can be used for improveing the estimated value that is produced by subject matter as herein described.

In an example, subject matter of the present invention comprises the system and the business method of the data that are used to produce dynamic function.Figure 11 display system 1000 comprises central server 1100 and communication network 1200.Central server 1100 comprises the server 1110 that is coupled to data base 1105 and terminating machine 1120.Server 1110 comes execution algorithm according to the instruction that is stored in memorizer or other storage facility (for example the data base 1105).Data base 1105 can comprise magnetic, optics or other data storage device.Terminating machine 1120 provides input equipment and output device, to allow operation and control system 1000.

In Figure 11, client computer platform 1310,1320 and 1330 is represented clinic or health care place, and it produces data according to subject matter of the present invention.Show three this kind client computer platforms among the figure, yet also contain more or less client computer platform.Pick off 1314 produces for example data at client computer platform 1310 places under the control of native processor 1312.Described data comprise the time series corresponding to cerebration.Described time series is to utilize the native processor 1312 that communicates with pick off 1314 to obtain, and pick off 1314 can comprise superconducting quantum interference device (SQUIDS) array.The time series data that is stored in the native processor 1312 utilizes communication network 1200 and central server 1100 to communicate.Communication network 1200 can comprise wired or wireless network, and the example comprises Ethernet, LAN (local area network; LAN), wide area network (wide area network such as for example the Internet; WAN) and PSTN (public switched telephone network PSTN).

Central server can comprise processor, and described processor is coupled to memorizer and stores the instruction that is used to carry out algorithm as herein described.Central server can comprise a more than processor, and described processor can be distributed on a plurality of positions.The those skilled in the art will easily know, the processor of central server can be implemented by any suitable processor, for example include but not limited to RISC or CISC microprocessor, microcontroller, pico computer, FPGA, ASIC, analog processor unit, quantum computer or biological processor, and can comprise single or multiple processing units.Processor also can be the type with batch processing mode or real-time mode operation.

In an example, on the basis of subscribing to, authorize or the registered client platform.On the basis of paying, central server is carried out a kind of algorithm, to produce the estimated value of dynamic cerebration according to time series.In an example, central server provides the report that comprises described estimated value.Described estimated value can alphanumeric or graphical format provide.

Figure 12 A shows the method for being carried out by an example of subject matter of the present invention 2000.In 2010, the time of reception sequence data.Described time series data is to carry out the task of opening eyes that only relates to nominal stimulation and motor activity (for example, visually fixation object thing) time the experimenter to produce.The task of opening eyes of this type makes experimenter's brain remain in idle state substantially.

Behind pick off or sensor array generation time sequence data through certain hour after, processor can receive and store described data.In 2020, remove the illusion electric wave in the data.The illusion electric wave can comprise illusion electric wave, heart illusion electric wave, physical motion or other illusion electric wave that produces by breathing.In 2030, by for example converting the MEG time series to immobilized white noise sequence, with described data prewhitening.In 2040,, produce synchronous coupled estimated value by the calculating section intercorrelation.Then, in 2050, described estimated value and template are compared.

In an example, the data of storing according to the particular subject of checking produce described template.In an example, the data of storing according to deriving from a plurality of different experimenters produce described template.Can implement to analyze by the number of subjects certificate is compared with template, and in an example, use the result of this particular subject that template is made amendment.In another example, after in a period of time, collecting and editing a plurality of experimenters, revise template with batch processing mode.

Figure 12 B shows the suitable method 2500 of utilizing network for example shown in Figure 11 to carry out.In 2510, connect reception number of subjects certificate by the Internet.In an example, number of subjects is according to comprising the MEG time series data.In 2520, for example utilize server 1100 to analyze.In 2530, use and come more new database 1105 corresponding to the information of this particular subject.In 2540, utilize network to report the result to the client computer platform, described result can comprise the analysis to data.

In an example, central server provides the screening report, so that the normality indication to be provided.This kind binary system report shows it is normally or undesired, and can be by the client computer platform as the threshold determination about the brain state.

In an example, central server can provide diagnosis, and described diagnosis comprises according to comparing the classification of carrying out with the data base.Described data base comprises corresponding to several seasonal effect in time series of being analyzed before this storage data.In addition, when receiving the client computer time series data, can use more new database of new data.In an example, trend data can be asked and receive to the client computer platform, and described trend data comprises the time series data early of specific brain regions and the comparison of later time series data.When forming diagnosis, subject matter of the present invention is distinguished various disease states.

Described data base can be provided for producing the data of template or model, to be used to analyze specific experimenter.For example, template can with specified disease or other neural state be corresponding or corresponding with normal brain.

In an example, central server provides feedback, so that can monitor experimenter's progress.Particularly, can come monitoring of diseases progress and therapeutic advance by in a time period, producing a plurality of estimated values.In addition, can during medicine examination clothes, produce the estimated value of neural synchronicity.Can estimate the safety and the effect of treatment rule of life by utilizing subject matter monitoring medicine examination clothes of the present invention.

The algorithm of being carried out by computer can be built into saved software instruction in memorizer.Some part of this software can be carried out in client computer platform and central server.

In an example, partly estimated value is specified to the function of subject age.The data of adjusting through the age can be stored among the data base.Other data also can be stored among the data base and be used to differentiate, and comprise for example known medical conditions or treatment rule of life.

Along with data base's evolution, expect specific variable will with specific morbid state height correlation.Therefore, these specific variablees can have different weights, with faster, distinguish different states more accurately.In an example, can use the relevant subclass of calculating to come the experimenter is classified as predictor.For example, can utilize the linear discrimination classification analysis of the use method of " staying a crosscheck ".In an example, six relevant being enough to 1.0 posterior probability to experimenter correctly classify (100% is correct).

In addition, it is believed that embodiments of the invention can have the utility program of the honesty that is used to distinguish the experimenter.In the form of a lie detector, collect the data that match with allegation to be measured from the experimenter.In addition, it is believed that other embodiments of the invention can have the utility program that is used to analyze or test intelligence.Therefore, specific markers can be identified as with specific intelligent grade and match.

Some embodiment of the present invention can provide objective test, with the accuracy that strengthens diagnosis, will be advanced into the presymptomatic stage to the identification of AD (and other state), and as the treatment monitor.

The adjustable number of pick off that is used for the capture time sequence is whole to arbitrary value, and in an example, and this quantity is reduced to the value that is enough to obtain the conclusion be concerned about.For example, example uses the set of sensors that reduces (promptly six or still less), is enough to obtain significant time series about the conclusion of specific neural state with generation.

Subject matter of the present invention can be used for having the experimenter of various brain states.Can utilize brain state that subject matter of the present invention discerns, diagnoses or monitor or some example of disease to comprise: the people who is subjected to ethanol or drug influence, sacred disease or state, multiple sclerosis, bipolarity emotion disease, traumatic brain injury, Parkinson's disease, depression, autoimmune function disease, neurodegenerative disease or disease, pain, and the disease that can influence central nervous system (CNS).Except that the influence of identification ethanol or other medicines, subject matter of the present invention can be used for diagnosing chronic alcoholism or fetal alcohol symdrome.Monitor the variation of every day of brain state when for example, embodiments of the invention are used in experimenter's potable spirit or use medicine.

By and large, embodiments of the invention can be used for using the template of being stored to come diagnostic state or disease, some kinds of different states of differentiation or disease and monitor the experimenter in time periods.

In an example, subject matter of the present invention comprises magneticencephalogram (MEG) data is carried out with different levels trooping.Can watch attentively the experimenter of 10 health and collect the collection data from 248 axial gradients when certain point reaches 45 seconds.Described data are carried out pretreatment, to remove heart illusion electric wave or the illusion electric wave of blinking.

Use static MEG data, acardia or nictation the illusion electric wave synchronous dynamic brain Network Layering time troop can be by visual.(0.1-400Hz takes a sample with the frequency of 1017Hz, Magnes 3600WH, 4-D Neuroimaging, San Diego, CA) collection data from 248 axial gradient meters when 10 health volunteers watch certain luminous point attentively to reach 45 seconds.By match autoregression integration rolling average (ARIMA) model and get residual and after with the time series prewhitening, calculate all paired zero lag part intercorrelation (N=30,628), thus provide between each neural colony the directly estimated value of synchronous coupled intensity and symbol (positive and negative) with 1 millisecond of short duration resolution.Use additivity woodlot set analysis by different level, and according to the relevant distance that derives of the right average portion of each pick off-pick off.Division to tree has shown the powerful pattern that the experimenter troops.Utilize each paired average portion of trooping between the pick off that makes up relevant, the interaction between estimating respectively to troop.The curve of trooping can show to be formed and interactional abundant complexity.Subject matter of the present invention can be used for the interactional neural colony that can assess various diseases group or state is carried out the function grouping.Some embodiment of the present invention can comprise the interactional linear discrimination classification analysis of synchronous nerve that magneticencephalogram (MEG) is assessed.

AD is the representative example of being considered.MEG can be used for assessing the dynamical state of 3 old experimenter group midbrain: the normal subjects (N=6,72.3+/-2.4 years old, average+/-SEM), experimenter (the N=6 that suffers from mild cognitive ability obstacle, 76.9+/-2.5 years old), and the experimenter (N=6,76.8+/-1.6 years old) who suffers from AD.(0.1-400Hz takes a sample with the frequency of 1017Hz, Magnes 3600WH, 4-D Neuroimaging, San Diego, CA) collection data from 248 axial gradient meters when the experimenter watches certain point attentively to reach 45 seconds.Data are carried out pretreatment, to remove heart or the illusion electric wave of blinking.By match autoregression integration rolling average (ARIMA) model and get residual and after with the time series prewhitening, calculate all paired zero lag part intercorrelation (N=30,628), thus provide between each neural colony the directly estimated value of synchronous coupled intensity and symbol (positive and negative) with 1 millisecond of short duration resolution.Can use these relevant little subclass as predictor, these experimenters are classified in these 3 groups.For example, can utilize the linear discrimination classification analysis of staying a crosscheck method of using brute force.The posterior probability that for example having 6 relevant predictor subclass is enough to 1.0 correctly is divided to its group's (i.e. 100% correctly classification) separately with all experimenters.Subject matter of the present invention can be used as the dynamic test of brain function.

Figure 13 shows the skeleton diagram according to the process of an embodiment.In 2610, (for example MEG) implements non-invasive test to the experimenter by electromagnetic measurement equipment.As indicated above, in one embodiment, the indication experimenter carries out the fixedly visual stimulus task that eyes are opened, so that experimenter's brain is in the idle condition that eyes are opened.In 2620, electromagnetic measurement device is collected the time series data of patient's brain.In one embodiment, data are taken a sample corresponding to the minimum sampling frequency of the 1kHz of 1 millisecond of of short duration resolution or better frequency.Sampling rate that this is fast relatively and of short duration resolution are substantially corresponding to carrying out neururgic speed in experimenter's brain.The many pick offs that distributed by the brain that spatially centers on the experimenter carry out data collection.

In 2630, one group of time series (wherein each time series is collected by corresponding sensor) is transmitted or otherwise is delivered to the data center that has data processing facility and optionally have data storage facility.In 2640, receive data in data center.The processing carried out in 2650 produces represent between each neural colony of experimenter on adding up the independently dynamic model of of short duration measurement.Described of short duration measurement can be for example be that detected by different sensors and institute time correlation sensing signal.These signals can overlap according to the sampling interval, and ground overlaps (be simultaneously, or nonsynchronous less than detectable amount) so that it does not have to lag behind.Perhaps, described of short duration measurement can be based on asynchronous but temporary transient relevant signal, for example interactional signal in window (for example, 50 milliseconds window) sometime.

Of short duration measurement between each neural colony can be with how relevant to pick off, and is perhaps relevant with other grouping (for example by showing the group that of short duration interactional 3 or more a plurality of pick off form each other).

The statistical independence of of short duration measurement with each between or consider that the obvious interaction between other sensor group of other variable is relevant.Can being implemented on the statistics independently, a kind of compute type of of short duration measurement is in the part intercorrelation described in the above-mentioned example.Yet in scope of the present invention and spirit, other method is also applicable to some application.For example, use residual can produce the statistical independence of the grouping of of short duration measurement.

As indicated above, the dynamic property of model means that the model of of short duration measurement is represented as the function of time, thereby makes it all different for each sampling period.Obviously, the dynamic model of of short duration measurement can be regarded as the network of interactional space nodes on a kind of meaning, and is not only the network that only has node in the structure configuration.Although above-mentioned example provides the space representation form of " brain map (brain map) ", described data also can be expressed as any suitable form in scope of the present invention and spirit.

As mentioned above, handling raw measurement data, can realize some advantage to remove the illusion electric wave and/or each time series prewhitening to be had with generation aspect the signal of meansigma methods, variance and autocorrelative stationary characteristic.This prewhitening step further improves the statistical independence of the of short duration measurement that will calculate.

In case calculate dynamic model, just can further handle, to simplify or to filter this model to it.A kind of filtration types is to use threshold function table to remove to have the of short duration measurement of more weak relatively value, only stays the brain that strong of short duration measurement is used to analyze the experimenter.

In one embodiment, analyze of short duration measurement to obtain covariance with one or more external attributes (for example age, race or neural psychological capacity) of experimenter.

In 2650, data center compares the dynamic model of of short duration measurement with one or more templates of classifying according to various brain states.Template can be regarded as the nervous physiology state model through checking on a kind of meaning.In one type embodiment, template is based on the experimenter group with common nervous physiology feature (for example disease or deformity) that is before estimated respectively.In this embodiment, these templates are through checking because and the corresponding index of state of the experimenter group of template institute foundation between have strong statistical correlation.

Each template can perhaps be the subclass of this kind dynamic model from the dynamic model as of short duration measurement.Template can be stored as data record, perhaps can be represented as a kind of algorithm or function, when this algorithm or function are carried out " comparison " at the dynamic model with the experimenter, revises this dynamic model to obtain comparative result.On a kind of meaning, template is exactly a classification function.In an exemplary embodiment, template is the form of the data shade (data mask) of the branch (weighted tap) with weighting.

As in the above-mentioned example, template can only limit to the selected subclass of the grouping (for example of short duration measurement to) of of short duration measurement, and all the other of short duration measurements are then because of being left in the basket with the pairing state of this template is irrelevant.In this kind mode, different templates can have the different grouping with the of short duration measurement of corresponding state or disease association.

When the dynamic model (or its subclass) of patient data when comparing with one or more templates, can be compared the different subclass of this dynamic model with each different templates.Therefore, for the template to A, B and E (discerning according to its space orientation) of representing associated sensor data, the data of of short duration measurement dynamic model that only need relatively take from the experimenter are to A, B and E.To C, D and the relevant different templates of E, can only use those that take from dynamic model right for data wherein.Resulting comparison can be scored, and perhaps otherwise represents degree of correlation.Perhaps, described comparison can produce binary system (being/deny) result.

In one aspect of the invention, the dynamic model of storage experimenter's of short duration measurement, and use this dynamic model to compare subsequently with same experimenter's more recent measurement.This method is applicable to following the tracks of progression of disease or estimating the effectiveness of specific therapy.In related embodiment, make template according to different pieces of information set, and use this template to follow the tracks of patient's state within a certain period of time from same experimenter.

In 2660, described system produces report, and this can comprise the pictorial representation form of patient's dynamic model, this pictorial representation form be mapped to 2 dimensions or 3 dimension spaces visual to realize, this is similar to the output shown in Fig. 3 or 10.

Figure 14 shows graphic according to the flow of information 3000 of one aspect of the invention.Clinic 3010 comprises experimenter's gauge 3012, reaches doctor or Laboratory Technician 3014.Network node 3016 helps realizing and the communicating by letter of remote node.In one embodiment, network node 3016 comprises the computer system with network interface, for example PC.Network node 3016 also can help realizing the operator interface between doctor 3014 and the instrument 3012.

In one embodiment, produce measured value by instrument 3012, these measured values were stored in the network node 3016 before transmission on the spot.Then, the indication network node transfers to external system analysis with instrument output 3018.Patient's summary 3020 that this system creation is associated with patient ID corresponding to instrument output 3018.This system comprises and will compare with diagnostic cast 3022 based on the information of instrument output according to the information (for example instrument output 3018) of above-mentioned any analytical technology processing from patient's summary 3020.In one embodiment, diagnostic cast 3022 is similar to above-mentioned template.

Can utilize the result 3024 of comparison to produce report 3026, to be delivered to clinic 3010 by network node 3016.Report 3026 can comprise the result 3024 of comparison, and the output of the figure of discussion that produces automatically and rendering results 3024.In addition, 3024 can be associated as a result, also be used for being delivered to clinic 3010 by network node 3016 with questionnaire 3028.Questionnaire 3028 can be filled in by doctor 3014, so that other care information about patient, test environment, therapy, artificial diagnosis or the like to be provided.Then, provide the questionnaire of being filled in, be stored in explicitly in the data storage 3030 to export 3018 with patient ID, report 3026, result 3024 and instrument with as feedback/tracking 3032.

Should be understood that above that explanation is intended to as illustrative and non-limiting explanation.For example, the various embodiments described above (and/or its various aspects) use that can mutually combine.After reading above explanation, many other embodiment for the those skilled in the art with obviously.Therefore, the scope of subject matter of the present invention should be determined according to the whole equivalent scope of enclose claim and these claim.In the claim of enclosing, word " comprises (including) " and reaches that " wherein (in which) " " comprises (comprising) " and reach the equivalent term of English for no reason of " wherein (wherein) " as corresponding word.In addition, in the claims, it is open that word " comprises (including and comprising) ", promptly lists in this kind word element afterwards to comprise that also system, device, object or the process of other element still are regarded as belonging in the scope of this claim in claim.And in claim above, word " first ", " second " reach " 3rd " etc. only with marking, and are not intended to its object is applied the numerical value requirement.

It is in order to meet some country for the requirement that summary is provided, so that the reader can determine the character of technology disclosure apace that summary of the present invention is provided.Should be understood that when submitting summary to that it will be not used in scope or the implication of explaining or limiting claims.In addition, in " specific embodiment " part in front, different characteristic can be combined so that the disclosure pipelining.The open method of this kind is not really wanted to be interpreted as reacting following intention: the embodiment that is advocated need have than the more feature of clearly being addressed in each claim of feature.But each claim reflects as mentioned, and subject matter of the present invention can be in the feature that is less than single all features that disclose embodiment.Therefore, above claims are incorporated in " specific embodiment " part now, and wherein each claim is all represented an independent embodiment alone.

Claims (109)

1. one kind is used for system that experimenter's nervous physiology activity is analyzed and classified, and described system comprises:

Data input pin, be used to receive experimenter's data acquisition system, described experimenter's data acquisition system representative is by the active time series of the collected nervous physiology of the pick off of many spatial distributions, and described pick off is arranged in order to survey experimenter's neural signaling during the idle condition of opening at eyes;

Data storage, be used to store a plurality of templates of classifying according to various brain states, each in the wherein said template is all represented from each neural colony that known at least one other experimenters that present set brain state measure on statistics the independently selected subclass of of short duration measurement;

Processor is coupled to described data input pin and described data storage with communication mode, and be programmed in order to:

Handle described experimenter's data acquisition system, with the dynamic model of the of short duration measurement in each the neural colony that obtains the described experimenter of representative; And

At least a portion of described dynamic model is compared with described a plurality of templates,, produce described experimenter's nervous physiology classification of activities with when described dynamic model is corresponding with in described a plurality of templates at least one.

2. the system as claimed in claim 1 is characterized in that described processor is programmed in order to handle described experimenter's data acquisition system to produce the prewhitening time series, and described prewhitening time series has meansigma methods, variance, reaches autocorrelative stationarity feature.

3. system as claimed in claim 2 is characterized in that the algorithm that described processor is programmed in order to use based on the rolling average of autoregression integration produces described prewhitening time series.

4. the system as claimed in claim 1, it is characterized in that described processor is programmed in order to calculate the part intercorrelation of described experimenter's data acquisition system, with the intensity of signaling and the estimated value of symbol between each group that produces described many pick offs, described estimated value is represented the interaction of each neural colony.

5. system as claimed in claim 4, each group that it is characterized in that described many pick offs is that in described many pick offs each is to pick off, and wherein calculate described part intercorrelation, to produce in described many pick offs each to the intensity of the direct signaling between at least one subclass of pick off and the estimated value of symbol.

6. system as claimed in claim 4 is characterized in that calculating described part intercorrelation, to produce in described many pick offs each to the intensity of the direct short-term signaling that occurs in the window between the pick off and the estimated value of symbol when about 50 milliseconds.

7. system as claimed in claim 4 is characterized in that calculating described part intercorrelation, with produce in described many pick offs each between the pick off directly and the intensity of the synchronous signaling of essence and the estimated value of symbol.

8. system as claimed in claim 4, it is characterized in that described processor is further programmed in order to analyze described part intercorrelation, to obtain the covariance about at least one parameter, described at least one parameter is to be selected from the group that is made up of following: age, race, and nervous physiology capacity or its combination in any.

9. system as claimed in claim 4 is characterized in that described part intercorrelation comprises at least a type that is selected from by the following group that forms: (a) positive part intercorrelation, and (b) negative part intercorrelation.

10. system as claimed in claim 4 is characterized in that described processor is further programmed in order to described part intercorrelation is carried out linear discriminant analysis, and to produce the classification function set, described classification function is used to produce described classification.

11. system as claimed in claim 10 is characterized in that described processor is programmed in order to determine the associated subset of described part intercorrelation, described associated subset is made up of some part intercorrelation relevant with carrying out described classification.

12. system as claimed in claim 11, it is characterized in that described processor is programmed in order to utilization stays a crosscheck algorithm to carry out described linear discriminant analysis, wherein, all need to carry out 100% classification in the described part intercorrelation that is regarded as being correlated with any one.

13. system as claimed in claim 11 is characterized in that described processor is programmed in order to utilize genetic algorithm to carry out described linear discriminant analysis.

14. system as claimed in claim 4 is characterized in that described processor is further programmed in order to the intensity of the signaling between each group of described many pick offs and the estimated value of symbol are used the threshold function, to produce described dynamic model.

15. system as claimed in claim 14, the nervous physiology classification of activities that it is characterized in that described experimenter comprises: to the intensity of the signaling between each group of described many pick offs and the estimated value application class function of symbol, described classification function is corresponding to one of them template in described a plurality of templates.

16. the system as claimed in claim 1 is characterized in that described processor is programmed in order to described dynamic model and at least one record of representing the experimenter to treat the administration history are stored in the described data storage explicitly.

17. the system as claimed in claim 1, the nervous physiology classification of activities that it is characterized in that described experimenter comprise about with at least one designator described in described a plurality of templates to response.

18. the system as claimed in claim 1 is characterized in that in the described template each only represents in each neural colony relevant with described set brain state in statistically evident of short duration measurement.

19. the system as claimed in claim 1 is characterized in that in the described template each represents the combination of the selected subclass of of short duration measurement in each neural colony of a plurality of different experimenters.

20. the system as claimed in claim 1, second of second template is corresponding in the first corresponding and described dynamic model that the nervous physiology classification of activities that it is characterized in that described experimenter is represented first template in described dynamic model and the described a plurality of templates and the described a plurality of templates, and described second template is different from described first template.

21. the system as claimed in claim 1 is characterized in that described dynamic model is based on dynamic neural network, described dynamic neural network is represented the grouping of the synchronous neural signaling that is detected by described many pick offs.

22. system as claimed in claim 21, the grouping that it is characterized in that described synchronous neural signaling comprises the paired grouping of the neural signaling that is defined by the synapse activity of the different brains position that takes place in about 1 millisecond.

23. the system as claimed in claim 1 is characterized in that described processor is programmed in order to compare with at least one template in described a plurality of templates by the information subset that will be selected from described dynamic model to obtain described classification.

24. the system as claimed in claim 1 is characterized in that described processor is programmed in order to optionally to upgrade at least one in described a plurality of template according to described classification.

25. the system as claimed in claim 1 is characterized in that described a plurality of template comprises separately corresponding to the neururgic template of at least one class that is selected from by the following group that forms: normal condition, Alzheimer's disease, lose intelligence syndrome in early stage, mild cognitive ability obstacle, schizophrenia, dry syndrome, alcoholism, alcohol damaged, fetal alcohol symdrome, multiple sclerosis, Parkinson's disease, bipolarity emotion disease, traumatic brain injury, depression, the autoimmune function disease, neurodegenerative disease, pain, influence central nervous system's disease, or its combination in any.

26. the system as claimed in claim 1 is characterized in that described data input pin is coupled to magneticencephalogram (MEG) instrument of the pick off with described many spatial distributions with communication mode.

27. system as claimed in claim 26 is characterized in that described pick off comprises superconducting quantum interference device (SQUID) pick off.

28. the system as claimed in claim 1 is characterized in that described experimenter's data acquisition system comprises that of short duration resolution is about 1 millisecond data.

29. the system as claimed in claim 1 is characterized in that the idle condition that described eyes are opened is to realize by send the instruction of carrying out the eye gaze task to described experimenter.

30. system as claimed in claim 29, it is characterized in that first experimenter's data acquisition system comprise less than or equal the idle condition activity that about one minute eyes are opened.

31. the system as claimed in claim 1 is characterized in that described data input pin comprises network communication device.

32. the system as claimed in claim 1 is characterized in that described data storage comprises at least one data base.

33. the system as claimed in claim 1, it is characterized in that described processor is programmed in order to handle described experimenter's data acquisition system, to remove the illusion electric wave that is selected from by at least a type of the following group that forms: the illusion of blinking electric wave, heart illusion electric wave, skeletal muscle illusion electric wave or its combination in any.

34. one kind is used for system that experimenter's nervous physiology activity is analyzed, described system comprises:

Data input pin, be used to receive experimenter's data acquisition system, described experimenter's data acquisition system representative is by the active time series of each collected nervous physiology in the pick off of many spatial distributions, and described pick off is arranged in order to survey described experimenter's neural signaling; And

Processor, with communication mode be coupled to described data input pin and be programmed in order to:

Handle described experimenter's data acquisition system, obtaining dynamic brain model, described dynamic brain model is represented independently of short duration measurement on statistics in each neural colony of brain of described experimenter; And

Analyze described dynamic brain model, to estimate described experimenter's nervous physiology state.

35. system as claimed in claim 34, it is characterized in that described on statistics independently of short duration measurement comprise in described experimenter's data acquisition system the set of the part intercorrelation between the different time sequence of packets, synchronous substantially interaction between each neural colony in the described experimenter's of described part intercorrelation set representative the brain.

36. system as claimed in claim 34 is characterized in that described processor is programmed in order at least a portion of described dynamic brain model is compared with different neural activity models, so that the result of described comparison indicates described experimenter's nervous physiology state.

37. system as claimed in claim 36, it is characterized in that described dynamic brain model is based on the described experimenter's of first described experimenter's data acquisition system of constantly obtaining the first dynamic brain model, and wherein said different neural activity model is based on from the described experimenter's who obtains at second different experimenter's data acquisition systems of constantly obtaining the second dynamic brain model, and described second is different from described first constantly constantly; And

Wherein said processor is programmed the potential variation of nervous physiology between described first moment and described second moment of representing described experimenter in order to the described experimenter's who analyzes nervous physiology state.

38. system as claimed in claim 37, the result who it is characterized in that described comparison indicate the degree of potential variation of described experimenter's nervous physiology.

39. system as claimed in claim 36 is characterized in that the dynamic brain model of described different neural activity model corresponding at least one different experimenter; And

Wherein said processor is programmed the potential difference of representing the nervous physiology between described experimenter and described at least one different experimenter in order to the described experimenter's who analyzes nervous physiology state.

40. system as claimed in claim 39 is characterized in that described degree of relatively indicating the described potential difference of the nervous physiology between described experimenter and described at least one experimenter different with described experimenter.

41. system as claimed in claim 36, it is characterized in that described different neural activity model is the nervous physiology state template, represent selected dynamic brain model subclass, described selected dynamic brain model subclass be based on and each of the corresponding a plurality of experimenter's data acquisition systems of a plurality of experimenter in part intercorrelation set between each grouping of different time sequence, known described a plurality of experimenters present set brain state; And

On behalf of described experimenter, wherein said processor is programmed in order to the described experimenter's who analyzes nervous physiology state present the degree of described set brain state.

42. system as claimed in claim 36 is characterized in that described different neural activity model comprises a plurality of different templates, described a plurality of different templates comprise:

First template, represent selected dynamic brain model subclass, described selected dynamic brain model subclass is based on and first experimenter gathers part intercorrelation set between the grouping of different time sequence in each of corresponding a plurality of experimenter's data acquisition systems, and known described first experimenter set presents the first brain state; And

Second template, represent selected dynamic brain model subclass, described selected dynamic brain model subclass is based on and second experimenter gathers part intercorrelation set between the grouping of different time sequence in each of corresponding a plurality of experimenter's data acquisition systems, and known described second experimenter set presents the second brain state that is different from the described first brain state;

The result of wherein said comparison indicates the experimenter to present the degree of described first brain state and the described second brain state.

43. system as claimed in claim 34, the nervous physiology state that it is characterized in that described experimenter is to be selected from the group that is made up of following: normal condition, Alzheimer's disease, mistake intelligence syndrome in early stage, mild cognitive ability obstacle, schizophrenia, dry syndrome, alcoholism, alcohol damaged, fetal alcohol symdrome, multiple sclerosis, Parkinson's disease, bipolarity emotion disease, traumatic brain injury, depression, autoimmune disease, neurodegenerative disease, pain, the disease that influences the central nervous system or its combination in any.

44. system as claimed in claim 34, the grouping that it is characterized in that described different time sequence be with described pick off in all each the corresponding described seasonal effect in time series of pick off is divided into groups in pairs, make in described experimenter's brain that described among each neural colony independently of short duration measurement on statistics is that each is to the intensity of the signaling between the pick off and the estimated value of symbol in described many pick offs, described estimated value is represented the paired interaction between the described neural colony.

45. system as claimed in claim 34 further comprises:

Data storage can mode of operation be coupled to described processor; And

Wherein said processor is programmed in order to described dynamic brain model at least one data storage record with the administration of the described experimenter's of representative at least a treatment is stored in the described data storage explicitly.

46. system as claimed in claim 34 is characterized in that described number of subjects is according to comprising that of short duration resolution is 1 millisecond data.

47. system as claimed in claim 34 is characterized in that described number of subjects is according to comprising the active data of the representative nervous physiology of described first experimenter during the idle condition of opening eyes.

48. system as claimed in claim 47, it is characterized in that representing the described number of subjects of the active seasonal effect in time series of described nervous physiology according to comprise less than or equal about one minute eye gaze activity.

49. system as claimed in claim 34, it is characterized in that described data input pin is coupled to magneticencephalogram (MEG) instrument of the pick off with described many spatial distributions with communication mode, described magneticencephalogram (MEG) instrument is selected from the group that is made up of magnetometer and axial gradient meter.

50. system as claimed in claim 34 is characterized in that described processor is programmed in order to handle described number of subjects certificate, to produce the prewhitening time series, described prewhitening time series has meansigma methods, variance, reaches autocorrelative stationarity feature.

51. system as claimed in claim 50 is characterized in that the algorithm that described processor is used based on the rolling average of autoregression integration produces described prewhitening time series.

52. system as claimed in claim 34, it is characterized in that described processor be programmed in order to described on statistics independently the set of of short duration measurement carry out linear discriminant analysis.

53. the active system of nervous physiology that is used to analyze first experimenter, described system comprises:

Data input pin, be used to receive a plurality of brain activity data set corresponding to the eye gaze task, each set representative in described a plurality of brain activity data set is by the active time series of the collected nervous physiology of the pick off of many spatial distributions, and described pick off is arranged in order to survey corresponding experimenter's neural signaling; And

Processor, with communication mode be coupled to described data input pin and be programmed in order to:

Handle the set of each brain activity data, producing corresponding neural activity dynamic model, described neural activity dynamic model is represented between each neural colony of described first experimenter coupling with time correlation, comprising:

Handle described brain activity data, to produce the prewhitening time series, described prewhitening time series has meansigma methods, variance, reaches autocorrelative stationarity feature;

Calculate the paired part intercorrelation of described prewhitening seasonal effect in time series, each is to the intensity of the signaling between the pick off and the estimated value of symbol in described many pick offs to produce, and described estimated value is represented the paired interaction of neural colony;

Described part intercorrelation is carried out classification, and with the measurement of the dependency that produces described brain activity data and verified reference data, described verified reference data is corresponding to multiple different nervous physiology states.

54. system as claimed in claim 53 is characterized in that described multiple different nervous physiology states comprise the two states at least that is selected from by the following group that forms: normal condition, Alzheimer's disease, lose intelligence syndrome in early stage, mild cognitive ability obstacle, schizophrenia, dry syndrome, alcoholism, alcohol damaged, fetal alcohol symdrome, multiple sclerosis, Parkinson's disease, bipolarity emotion disease, traumatic brain injury, depression, autoimmune disease, neurodegenerative disease, pain, influence central nervous system's disease, or its combination in any.

55. system as claimed in claim 53 is characterized in that the algorithm that described processor is programmed in order to use based on the rolling average of autoregression integration produces described prewhitening time series.

56. system as claimed in claim 53 is characterized in that calculating described part intercorrelation, to produce in described many pick offs each to the intensity of the direct short-term signaling that occurs in the window between the pick off and the estimated value of symbol when about 50 milliseconds.

57. system as claimed in claim 53 is characterized in that calculating described part intercorrelation, to produce in described many pick offs each to the intensity of the synchronous signaling of essence between the pick off and the estimated value of symbol.

58. system as claimed in claim 53, it is characterized in that described processor is selected from least one class part intercorrelation by the following group that forms by further programming in order to analysis, to obtain covariance: (a) the positive part intercorrelation of the described part intercorrelation of described prewhitening seasonal effect in time series, and (b) the negative part intercorrelation of the described part intercorrelation of described prewhitening seasonal effect in time series.

59. system as claimed in claim 53, it is characterized in that described processor by further programming in order to described many pick offs each between the intensity of signaling and the estimated value of symbol use the threshold function, to produce the active dynamic model of first nerves.

60. system as claimed in claim 53 is characterized in that described processor is programmed in order to determine the associated subset of described part intercorrelation, described associated subset is made up of some part intercorrelation relevant with carrying out described classification.

61. system as claimed in claim 60, it is characterized in that described processor is programmed in order to utilization stays a crosscheck algorithm to carry out described linear discriminant analysis, wherein, all need to carry out 100% classification in the described part intercorrelation that is regarded as being correlated with any one.

62. system as claimed in claim 60 is characterized in that described processor is programmed in order to utilize genetic algorithm to carry out described linear discriminant analysis.

63. system as claimed in claim 53 is characterized in that described processor is programmed in order to form the nervous physiology template according to described classification.

64. one kind is used for method that experimenter's nervous physiology cerebration is classified automatically, described method comprises:

Receive experimenter's data acquisition system by data handling system, described experimenter's data acquisition system representative is by the active time series of the collected nervous physiology of the pick off of many spatial distributions, and described pick off is arranged in order to survey experimenter's neural signaling during the idle condition of opening at eyes;

Handle described experimenter's data acquisition system by described data handling system, obtaining dynamic model, described dynamic model is represented independently of short duration measurement on statistics in each neural colony of described experimenter;

Safeguard that by described data handling system the set of the template of classifying according to various brain states, each in the wherein said template all represent from each neural colony that known at least one other experimenter who presents set brain state measure on statistics the independently selected subclass of of short duration measurement;

By described data handling system at least a portion of described dynamic model is compared with described a plurality of templates,, produce described experimenter's nervous physiology classification of activities with when described dynamic model is corresponding with in described a plurality of templates at least one.

65., further comprise as the described method of claim 64:

Handle described experimenter's data acquisition system to produce the prewhitening time series, described prewhitening time series has meansigma methods, variance, reaches autocorrelative stationarity feature.

66. as the described method of claim 65, it is characterized in that handling described experimenter's data acquisition system and comprise to produce the prewhitening time series: the algorithm of using based on the rolling average of autoregression integration produces described prewhitening time series.

67. as the described method of claim 64, it is characterized in that handling described experimenter's data acquisition system comprises to obtain dynamic model: the part intercorrelation of calculating described experimenter's data acquisition system, with the intensity of signaling and the estimated value of symbol between each group that produces described many pick offs, described estimated value is represented the interaction of each neural colony.

68., it is characterized in that the part intercorrelation of calculating described experimenter's data acquisition system comprises as the described method of claim 67: produce each pick off in described many pick offs between the intensity of signaling and the estimated value of symbol.

69. as the described method of claim 67, the part intercorrelation that it is characterized in that calculating described experimenter's data acquisition system can produce in described many pick offs each to the intensity of the direct short-term signaling that occurs in the window between the pick off and the estimated value of symbol when about 50 milliseconds.

70. as the described method of claim 67, the part intercorrelation that it is characterized in that calculating described experimenter's data acquisition system can produce in described many pick offs each between the pick off directly and the intensity of the synchronous signaling of essence and the estimated value of symbol.

71., further comprise as the described method of claim 67:

Analyze described part intercorrelation, to obtain the covariance about at least one parameter, described at least one parameter is to be selected from the group that is made up of following: age, race, and nervous physiology capacity or its combination in any.

72. as the described method of claim 67, it is characterized in that the part intercorrelation of calculating described experimenter's data acquisition system can produce at least a part intercorrelation type that is selected from by the following group that forms: (a) positive part intercorrelation, and (b) negative part intercorrelation.

73., further comprise as the described method of claim 67:

Described part intercorrelation is carried out linear discriminant analysis, to produce the classification function set; And

Utilize described classification function to produce described classification.

74., further comprise as the described method of claim 73:

Determine the associated subset of described part intercorrelation, described associated subset is made up of some part intercorrelation relevant with carrying out described classification.

75. as the described method of claim 74, it is characterized in that carrying out described linear discriminant analysis and comprise utilizing and stay a crosscheck algorithm, wherein, all need to carry out 100% classification in the described part intercorrelation that is regarded as being correlated with any one.

76., it is characterized in that carrying out described linear discriminant analysis and comprise the described subclass of utilizing genetic algorithm to select described part intercorrelation as the described method of claim 74.

77., further comprise as the described method of claim 67:

The intensity of the signaling between each group of described many pick offs and the described estimated value of symbol are used the threshold function, to produce described dynamic model.

78. as the described method of claim 67, it is characterized in that comparing to produce described classification and comprise: to the intensity of the signaling between each group of described many pick offs and the described estimated value application class function of symbol, described classification function is corresponding to one of them template in described a plurality of templates.

79., further comprise as the described method of claim 64:

By described data handling system described dynamic model is associated with at least one record of representing the experimenter to treat the administration history.

80. as the described method of claim 64, it is characterized in that with the described at least part of described dynamic model compare with described a plurality of templates can produce about with described a plurality of templates at least one designator to response.

81., it is characterized in that in the described template each only represents in each neural colony relevant with described set brain state in statistically evident of short duration measurement as the described method of claim 64; And

Wherein the described at least part of described dynamic model is compared with described a plurality of templates and comprise: the different subclass of described dynamic model are compared with each template.

82., further comprise as the described method of claim 64:

Set up each in the described template, wherein each template is represented the combination of the selected subclass of the of short duration measurement in each neural colony of a plurality of different experimenters.

83. as the described method of claim 64, it is characterized in that the described at least part of described dynamic model compared with described a plurality of templates and can produce described experimenter's the active described classification of nervous physiology, so that second template in the first corresponding and described dynamic model that first template in described dynamic model and the described a plurality of templates is represented in described classification and the described a plurality of templates is second corresponding, described second template is different from described first template.

84. as the described method of claim 64, it is characterized in that handling described experimenter's data acquisition system comprises to obtain described dynamic model: produce dynamic neural network, described dynamic neural network is represented the grouping of the synchronous neural signaling that is detected by described many pick offs.

85. as the described method of claim 84, it is characterized in that producing described dynamic neural network and comprise the grouping of representing synchronous neural signaling, the grouping of described synchronous neural signaling comprises the paired grouping of the neural signaling that is defined by the synapse activity of the different brains position that takes place in about 1 millisecond.

86., further comprise as the described method of claim 64:

According to described classification, upgrade at least one in described a plurality of template.

87., it is characterized in that safeguarding that described template set comprises that maintenance is separately corresponding to the neururgic template of at least one class that is selected from by the following group that forms: normal condition as the described method of claim 64, Alzheimer's disease, lose intelligence syndrome in early stage, mild cognitive ability obstacle, schizophrenia, dry syndrome, alcoholism, alcohol damaged, fetal alcohol symdrome, multiple sclerosis, Parkinson's disease, bipolarity emotion disease, traumatic brain injury, depression, autoimmune disease, neurodegenerative disease, pain, influence central nervous system's disease, or its combination in any.

88., further comprise as the described method of claim 64:

Acquisition comprises the magneticencephalogram of described experimenter's data acquisition system.

89., it is characterized in that obtaining described magneticencephalogram and comprise: 1 millisecond sampling interval at the most data are taken a sample as the described method of claim 88.

90. one kind is utilized data handling system to obtain automatically method structural or the nervous physiology assessment that the neuro chemistry encephalopathy (HIE) becomes, described method comprises:

Obtain the brain data acquisition system, described brain data acquisition system representative is by the active time series of each collected nervous physiology in the pick off of many spatial distributions, and described pick off is arranged the neural signaling in order to the brain of surveying the experimenter;

Handle described brain data acquisition system, obtaining dynamic brain model, described dynamic brain model is represented independently of short duration measurement on statistics in each neural colony of brain of described experimenter; And

Analyze described dynamic brain model, to obtain nervous physiology assessment to brain.

91. as the described method of claim 90, it is characterized in that handling described brain data acquisition system with obtain described on statistics independently of short duration measurement comprise: calculate described brain data acquisition system in the part intercorrelation between the different time sequence of packets and gather, so that described dynamic brain model is represented the interaction that is close to generation in the brain between each neural colony in time.

92. as the described method of claim 91, it is characterized in that handling described brain data acquisition system comprises: obtain described dynamic brain model based on described part intercorrelation set, the described grouping of the different time sequence of described part intercorrelation set be with described pick off in all each the corresponding described seasonal effect in time series of pick off is divided into groups in pairs, make that described part intercorrelation set is that each is to the intensity of the signaling between the pick off and the estimated value of symbol in described many pick offs, described estimated value is represented the paired interaction between the described neural colony.

93., it is characterized in that analyzing described dynamic brain model and comprise as the described method of claim 90:

The neural activity models that at least a portion of described dynamic brain model is different with at least one are compared, so that the result of described comparison indicates the nervous physiology state of brain.

94., it is characterized in that described brain data acquisition system is constantly to obtain first as the described method of claim 93; And

Wherein said method further comprises:

Obtain another brain data acquisition system in second moment that is different from described first moment;

Based on processing to described another brain data acquisition system, obtain described at least one different neural activity model, wherein said another brain data acquisition system is corresponding to same brain; And

Wherein analyze described dynamic brain model and comprise, survey the nervous physiology potential variation of described brain between described first moment and described second moment based on described comparison.

95., it is characterized in that the analysis meeting of described dynamic brain model is produced the result of the degree of the potential variation of nervous physiology of indicating described brain as the described method of claim 94.

96., further comprise as the described method of claim 93:

Obtain another brain data acquisition system from least one other brain;

Based on processing, obtain described at least one different neural activity model to described another brain data acquisition system; And

Wherein analyze described dynamic brain model and comprise the potential difference of surveying between described brain and described at least one other brain of nervous physiology.

97., it is characterized in that the analysis meeting of described dynamic brain model is produced the result of the degree of the potential difference of nervous physiology between the described brain of indication and described at least one other brain as the described method of claim 96.

98., further comprise as the described method of claim 93:

Based on described at least one different neural activity model, produce the nervous physiology state template, wherein based on and each of the corresponding a plurality of brain data acquisition systems of set brain state in part intercorrelation set between each grouping of different time sequence, present selected dynamic brain model subclass; And

Wherein the analysis to described dynamic brain model comprises the degree of described brain corresponding to described set brain state of calculating.

99. as the described method of claim 93, it is characterized in that the described at least part of described dynamic brain model and described at least one different model compared and comprise described dynamic brain model is compared with a plurality of different templates that described a plurality of different templates comprise:

First template is represented selected dynamic brain model subclass, described selected dynamic brain model subclass be based on and each of the corresponding a plurality of brain data acquisition systems of the first brain state in part intercorrelation set between the grouping of different time sequence; And

Second template, represent selected dynamic brain model subclass, described selected dynamic brain model subclass be based on and each of the corresponding a plurality of brain data acquisition systems of the second brain state in part intercorrelation set between the grouping of different time sequence, the described second brain state is different from the described first brain state; And

The result of wherein said comparison indicates the degree of correspondence of described brain and described first brain state and the described second brain state.

100., it is characterized in that the nervous physiology assessment that described dynamic brain model is analyzed to obtain described brain is comprised the assessment that obtains to be selected from by the following group that forms: normal condition as the described method of claim 90, Alzheimer's disease, lose intelligence syndrome in early stage, mild cognitive ability obstacle, schizophrenia, dry syndrome, alcoholism, alcohol damaged, fetal alcohol symdrome, multiple sclerosis, Parkinson's disease, bipolarity emotion disease, traumatic brain injury, depression, autoimmune disease, neurodegenerative disease, pain, influence central nervous system's disease, or its combination in any.

101., further comprise as the described method of claim 90:

Described dynamic brain model is associated with at least one record, and described at least one record representative conference influences the administration of at least a therapy of described brain potentially.

102., it is characterized in that obtaining described brain data and comprise: when described brain is in idle condition when being subjected to the static vision stimulation, obtain the data of taking a sample with about 1 millisecond of short duration resolution as the described method of claim 90.

103., further comprise as the described method of claim 90:

Handle described brain data, to produce the prewhitening time series, described prewhitening time series has meansigma methods, variance, reaches autocorrelative stationarity feature.

104., further comprise as the described method of claim 103:

To the algorithm of described brain data application, to produce described prewhitening time series based on the rolling average of autoregression integration.

105., further comprise as the described method of claim 90:

To the described independently of short duration measurement execution linear discriminant analysis on statistics in each neural colony in described experimenter's brain.

106. one kind helps utilizing data handling system to obtain automatically method structural or the nervous physiology assessment that the neuro chemistry encephalopathy (HIE) becomes, described method comprises:

Provide instruction with:

Obtain the brain data acquisition system, described brain data acquisition system representative is by the active time series of each collected nervous physiology in the pick off of many spatial distributions, and described pick off is arranged the neural signaling in order to the brain of surveying the experimenter; And

Described brain data acquisition system is offered described data handling system; And

Reception is to the nervous physiology assessment of described experimenter's brain, described nervous physiology assessment is based on by described data handling system to be handled described brain data acquisition system, to form dynamic brain model by described data handling system, described dynamic brain model is based on independently of short duration measurement on statistics in each neural colony in described experimenter's brain, wherein said dynamic brain model is represented between each neural colony of described brain and is close to the interaction of appearance in time, and analyzes described dynamic brain model to obtain the described nervous physiology assessment to described brain.

107. one kind helps utilizing data handling system to obtain automatically method structural or the nervous physiology assessment that the neuro chemistry encephalopathy (HIE) becomes, described method comprises:

Provide instruction to obtain the brain data acquisition system, described brain data acquisition system representative is by the active time series of each collected nervous physiology in the pick off of many spatial distributions, and described pick off is arranged the neural signaling in order to the brain of surveying the experimenter; And

Use described data handling system with:

Handle described brain data acquisition system, to obtain dynamic brain model, described dynamic brain model is based in described experimenter's brain independently of short duration measurement on statistics in each neural colony, and wherein said dynamic brain model is represented the interaction that is close to appearance between each neural colony of described brain in time; And

Analyze described dynamic brain model to obtain described nervous physiology assessment to described brain.

108. a computer-readable media comprises the instruction that is suitable for making the following operation of computer system execution:

Receive experimenter's data acquisition system, described experimenter's data acquisition system representative is by the active time series of each collected nervous physiology in the pick off of many spatial distributions, and described pick off is arranged the neural signaling in order to the brain of surveying the experimenter;

Handle described experimenter's data acquisition system, to obtain dynamic brain model, described dynamic brain model is represented independently of short duration measurement on statistics in each neural colony in described experimenter's brain; And

Analyze described dynamic brain model to estimate described experimenter's nervous physiology state.

109. the active system of nervous physiology that is used to analyze the experimenter, described system comprises:

Data source, be used to obtain experimenter's data acquisition system, described experimenter's data acquisition system representative is by the active time series of each collected nervous physiology in the pick off of many spatial distributions, and described pick off is arranged in order to survey described experimenter's neural signaling; And

Processor, with communication mode be coupled to described data source and be programmed in order to:

Handle described experimenter's data acquisition system, to obtain dynamic brain model, described dynamic brain model is represented independently of short duration measurement on statistics in each neural colony in described experimenter's brain; And

Analyze described dynamic brain model to estimate described experimenter's nervous physiology state.

Images