CN102293656A - Emotional stability evaluation system based on magnetic resonance imaging and evaluation method thereof - Google Patents

Abstract

The invention discloses an emotional stability evaluation system based on magnetic resonance imaging and an evaluation method thereof. The emotional stability evaluation system comprises an emotional stability characteristic sample database, a magnetic resonance imaging unit, a characteristic extraction unit, a pattern classifier and an emotional stability evaluation unit, wherein a magnetic resonance structure of emotional levels and emotional components with different emotional stabilities and resting state functional imaging attributes are stored in the emotional stability characteristic sample database; the magnetic resonance imaging unit acquires imaging data of a magnetic resonance structure phase and a resting state functional phase of the brain of a detected individual, and transmits the imaging data to the characteristic extraction unit; the characteristic extraction unit extracts the magnetic resonance structure and the resting state functional imaging attribute of the brain of the detected individual from the imaging data; the pattern classifier carries out classifier training on the magnetic resonance structure and the resting state functional imaging attribute; and the emotional stability evaluation unit evaluates the emotional stability level and/or emotional cause of the detected individual according to the classifier training result, the magnetic resonance structure and the resting state functional imaging attribute. Therefore, the system realizes accurate, objective and stable evaluation on the emotional level and cause of the detected individual.

Description

Emotional stability evaluation system and evaluation methodology thereof based on nuclear magnetic resonance

Technical field

The present invention relates to medical domain, relate in particular to a kind of emotional stability evaluation system and evaluation methodology thereof based on nuclear magnetic resonance.

Background technology

Emotion is a kind of comprehensive psychology and the physiological status of the various sensation of people, thought and act, is to stimulate the psychoreaction that is produced to external world, and incidental physiological reaction, as: pleasure, anger, sorrow, happiness etc.Emotion is that individual's subjectivity is experienced and impression, and is often relevant with disposition with mood, makings, personality.Emotion is the external phenotype of mental activity, and this external phenotype is in easier showing under the pressure sight at individuality.

Be emotionally stable and be meant the individual reaction that under the emotion situation, does not show overexcitement, the sign that is the embodiment of individual emotion capacity of self-regulation and maturity is referring to " Zhang Shi psychology voluminous dictionary " [M] 1, the Shanghai Lexicographic Publishing House, 1991.225, author: Zhang Chunxing.Individual emotional stability difference is metastable, and is showing under the different emotion situations: under in the face of complicated, urgent, fear or dangerous situation, the individuality of being emotionally stable shows takes calmly, and can bring into play the profile; The individuality of emotional instability shows is at a loss, hypertonicity, behavioral disorder, the ability that forfeiture has the situation under control.Emotional stability difference not only shows on the difference that changes with the variation cognitive activities level of situation, also shows on the difference of situation control ability.Referring to Claudia H, Michael K, Lilly B.The impact of personality andtaskcharacteristics on stress and strain during helicopter flight[J] .International JAviation Psychology, 1991,1 (4): 3012318. and Bertschy G.Relations between mooddisorders and personality[J] .Recent Data Encephale, 1992,18 (2): 187292.

Emotional lability is qualitative, and translating nervousness (Neuroticism) again is basic personality trait in the psychological study.It is a kind of lasting tendency that experiences the negative feeling state.The anxiety that the people that the qualitative score of emotional lability is high may run into than the easier experience of common people, indignation, guilty and depressed such emotion G.Matthewsand Ian J.Deary.Personality traits.Cambridge, UK:Cambridge University Press.1998..Their response environment pressure capability is relatively poor, more likely normal condition is interpreted as threatening, and regards little setback as desperate difficulty.They are self consciousness and shy normally, and might be difficult to control impulsion and deferred gratification.Qualitative and the low feeling quotrient of emotional lability is relevant, comprises mood regulation, and motivation and interpersonal skills are low referring to Daniel Goleman.Emotional Intelligence.New York:Bantam.1997..This also is " internalization " mental sickness such as phobia, depression, Panic disorder.A risk factor with other anxiety neurosis (being called as neurosis traditionally).Referring to this piece paper (Hettema, J.M., Neale, M.C., Myers, J.M., Prescott, C.A. ， ﹠amp; Kendler, the geminus cohort study of K.S. nervousness and Internalization Obstacle relation.American?journal?of?Psychiatry，163：857-864，2006)。

The people of emotional instability causes emotional response easily to the generation of incident, and is once causing anxious state of mind, relatively poor to the control of emotion.This stability of mood is relevant with individual's will intensity.The author thinks individual because the difference on the will causes the difference of emotion control for the psychology voluminous dictionary of Zhu Zhixian.Under stress situation, movable accuracy of the individual cognition that sight control energy deposit is high and motion action motility are impaired little, and resume speed is greater than the low individuality of energy reserve; Stress situation makes the low individuality of energy reserve show sympathetic hyperactivity, the movable loss of equilibrium of sympathetic-parasympathetic.

On the contrary, the higher people's emotion of emotional stability score is more stable, and they are normally tranquil, are not easy nervous or flurried.Though their negative feeling is low, positive emotion is not necessarily high.This is and extroversion's speciality independently mutually.Emotional instability is export-oriented, is a kind of front and the negative all very high type of emotional state level, i.e. so-called " emotion roller-coaster ".The people that the emotional stability score is higher, especially those extroversions are also high, common happy degree and higher to the satisfaction of life.

Emotional stability (emotional stability) is specific professional population such as aircrew (spacefarer), enter a higher school before the student contour stress, high cognitive load and the necessary basic psychology quality of high risk vocational activity, be prerequisite and the condition of finishing the high-tech operation under this special environment and the condition.The active disposal ability safe, the special feelings of flight of the achievement of entering a higher school quality, flight (space flight) is all closely related with the emotional stability level of this occupation individuality.Therefore, effectively emotional stability is estimated and predicted that be emotionally stable sexual relationship research of individual character influence factor and flying cadet just seems very important.

Emotional stability generally is considered to be made of three distinguishing but overlapped origin causes of formation: personality factor, emotional factor, cognitive factor, all these can be to emotion performance generation in various degree the influence of individuality under certain sight.Present subjective diagnostic techniques as estimating scale, emotional stability questionnaire table, can only be evaluated the grade of being emotionally stable singlely, can not draw the character of emotional stability; And this method is restricted for the subjective patient who expresses of forfeiture.And present method for objectively evaluating, the behavior determination method all is from different angles emotional stability to be carried out indirect evaluation with physiology's algoscopy, and is reliable inadequately and can only provide finite information, and the quantification of emotional stability also is restricted.

Therefore, a kind of energy is estimated emotional stability objective and accurately and is detected the system and method for various factors effect in the emotional stability, and very big social need is arranged.Depend on the statement and the test and appraisal of psychology scale of the subjectivity of object at present to a great extent for the evaluation of emotional stability, have stronger subjectivity, a kind of can reliable detection and the objective method of emotional stability associated change can improve assessment and prediction to emotional stability.If this technology can also detect the relevant origin cause of formation of difference in the emotional stability, as personality factor, emotional factor, cognitive factor, this will bring strong impetus to emotion relevant test and appraisal and diagnosis and treatment work.

The research of cerebral function imaging has in recent years made people that the understanding of the central mechanism of emotional state has been produced revolutionary variation.In fact, as if people have been in the edge that obtains the brand-new clinical information of tested emotional state by FMRI (Functional Magnetic ResonanceImaging, magnetic resonance functional imaging).Emotional stability is a kind of constant relatively state, the different subtype that functional imaging has redefined mental sickness such as depression, schizophrenia has specific neuroimaging mark, at this with reference to this piece paper (Gong Q, Wu Q, Scarpazza C, Lui S, Jia Z, Marquand A, Huang X, McGuire P, Mechelli A.. uses the prognosis prediction of the treating depression of High-Field means.Neuroimage，55：1497-503，2011)。And researched and developed the image feature index such as the order of severity of first depressed these diseases for indicating some complex diseases, at this with reference to this piece paper (Zhang JR, Wang JH, Wu QZ, Kuang WH, Huang XQ, HeY, Gong QY. unusual brain first, not medication depressive patient connects network.Biological?Psychiatry，2011)。Structure and tranquillization functional mri to tested and imaging operation personnel require less, repeatable high, can reflect the new functional imaging means of brain structure and spontaneous activity.Because the activity of brain is the concentrated expression to emotional state, include consciousness or be not intended to spontaneous, the utilization of functional imaging will make people be familiar with emotional state with a kind of objective method, and understand its potential neural loop better, distinguish the heterogeneity in its emotional state.

Yet, the subjective assessment method of existing evaluation emotional stability is subjected to many influences that adds factor easily, thereby the influence evaluation is objective and accurate, existing objective evaluation method only relies on single index to judge, the grade that can not accurately reflect emotional stability objectively, and existing assessment method mainly pays close attention to the emotional stability power, can not distinguish psychology, emotional factor composition in the emotional state.

In summary, existing emotion assessment technique obviously exists inconvenience and defective, so be necessary to be improved on reality is used.

Summary of the invention

At above-mentioned defective, the object of the present invention is to provide a kind of emotional stability evaluation system and evaluation methodology thereof, to realize to tested individuality be emotionally stable accurate, objective, the evaluation stably of the grade and the origin cause of formation based on nuclear magnetic resonance.

To achieve these goals, the invention provides a kind of emotional stability evaluation system, comprising based on nuclear magnetic resonance:

The feature samples storehouse of being emotionally stable is used to store the emotion grade with different emotional stabilitys and the magnetic resonance structure and the tranquillization attitude functional image attribute of emotion composition;

The nuclear magnetic resonance unit is used to obtain tested individual brain magnetic resonance structure phase and tranquillization attitude function imaging data mutually, and is transferred to feature extraction unit;

Feature extraction unit is used for magnetic resonance structure phase and tranquillization attitude function imaging data mutually from the transmission of described nuclear magnetic resonance unit, extracts the magnetic resonance structure and the tranquillization attitude functional image attribute of described tested individual brain;

Pattern classifier, the magnetic resonance structure and the tranquillization attitude functional image attribute that are used for described tested individual brain that described feature extraction unit is extracted carry out the grader training;

The emotional stability evaluation unit, be used for grader training result and the described emotion class information of the described feature samples library storage of being emotionally stable and the magnetic resonance structure and the tranquillization attitude functional image attribute of emotion composition, evaluate the emotional stability grade and/or the emotion origin cause of formation of described tested individuality according to described pattern classifier.

According to described emotional stability evaluation system based on nuclear magnetic resonance, described magnetic resonance structure and tranquillization attitude functional image attribute comprise: the subnet/component attribute of nodal community, reflection brain district's modularity or the different component of the function connection attribute of cooperative ability, reflection brain district power of influence between the full brain network attribute of reflection brain overall situation integration ability, the local attribute of the local integration ability of reflection brain, each brain district of reflection brain.

According to described emotional stability evaluation system based on nuclear magnetic resonance, the brain district of described subnet/component comprises: right frame volume cortex, abdomen outside prefrontal lobe cortex, bilateral temporo cortex, corpus amygdaloideum, caudatum, nucleus accumbens septi, anterior cingutate, Hippocampus, bottom right top, island leaf, posterior parietal cortex, prefrontal cortex, cingulum cortex.

According to described emotional stability evaluation system based on nuclear magnetic resonance, described emotional stability evaluation unit comprises:

The ranking subelement of being emotionally stable, be used for the emotion class information that the magnetic resonance structure of described tested individuality and tranquillization attitude functional image attribute is corresponding with the described feature samples storehouse of being emotionally stable and the magnetic resonance structure and the tranquillization attitude functional image attribute of emotion composition and compare, obtain the grade of being emotionally stable of described tested individuality; And/or

Emotion composition evaluation subelement, be used for according to the categorical attribute of the subnet/component of the different emotion compositions of the described feature samples library storage of being emotionally stable and the subnet/component attribute of described tested individual brain, carry out the composition identification of the emotional stability of described tested individuality, obtain the emotional stability origin cause of formation of described tested individuality and/or the weight of the described origin cause of formation.

According to described emotional stability evaluation system based on nuclear magnetic resonance, described emotional stability evaluation unit also comprises:

Obtain subelement, be used to receive the clinical emotion test and appraisal information of tested individuality, and, analyze the emotional stability grade and/or the emotion origin cause of formation of described tested individuality according to magnetic resonance structure and the tranquillization attitude functional image attribute and the described clinical Psychological Evaluation information of described tested individuality; And/or

The fine setting subelement is used to finely tune the relevant parameter of the be emotionally stable grade and/or the emotion origin cause of formation in the described feature samples storehouse of being emotionally stable.

According to described emotional stability evaluation system based on nuclear magnetic resonance, described feature extraction unit comprises:

The feature extraction subelement is used for isolating from the full brain network of the brain of described tested individuality the subnet/component of the different emotion compositions of reflection, and calculates described isolated subnet/component attribute, function connection attribute and nodal community;

Divide subelement, be used for full brain is divided into a plurality of different brain districts according to dissecting template, then according to described magnetic resonance structure and tranquillization attitude functional imaging data, with described brain district is that unit extracts the active time series in reflection tranquillization attitude brain district, adopts the time series/skin thickness in described brain district to make up described full brain network and subnet net again.

According to described emotional stability evaluation system based on nuclear magnetic resonance, when described emotional state feature samples storehouse also is used for storage and is used to make up described full brain network and subnet net, function connected node, by described magnetic resonance structure and tranquillization attitude functional imaging data trim values by the intermediate value, emotional stability grade/origin cause of formation evaluation result, clinical Psychological Evaluation information and the emotional stability grade/origin cause of formation that calculate.

According to described emotional stability evaluation system based on nuclear magnetic resonance, the described emotion origin cause of formation comprises: personality factor, emotional factor and cognitive factor.

According to described emotional stability evaluation system based on nuclear magnetic resonance, described pattern classifier is non-linear support vector machine classifier;

Described pattern classifier carries out cross validation to described grader training result.

The present invention also provides a kind of and has utilized above-mentioned any one system to realize the method that emotional stability is estimated, and described method comprises:

A, nuclear magnetic resonance unit obtain tested individual brain magnetic resonance structure phase and tranquillization attitude function imaging data mutually, and are transferred to feature extraction unit;

B, feature extraction unit from the magnetic resonance structure of described nuclear magnetic resonance unit transmission mutually with tranquillization attitude function imaging data mutually, extract the magnetic resonance structure and the tranquillization attitude functional image attribute of described tested individual brain;

C, pattern classifier carry out the grader training with the magnetic resonance structure and the tranquillization attitude functional image attribute of the described tested individual brain that described feature extraction unit is extracted;

D, emotional stability evaluation unit are according to the grader training result of described pattern classifier, and the magnetic resonance structure of the described emotion class information of the described feature samples library storage of being emotionally stable and emotion composition and tranquillization attitude functional image attribute, evaluate the emotional stability grade and/or the emotion origin cause of formation of described tested individuality.

The present invention is by gathering the objective image data of structure and tranquillization attitude magnetic resonance functional imaging gained, adopt the information processing technology therefrom to extract tested new individual configurations and tranquillization attitude functional image attribute, and will be somebody's turn to do with the feature samples storehouse of being emotionally stable in historical structure and the comparing of tranquillization attitude functional image attribute, preferably by pattern classification, methods such as self adaptation feedback learning are classified to image data eigenvalue and clinical emotional information, extract, optimize, the comprehensive assessment system that becomes to reflect objective the be emotionally stable grade and the origin cause of formation, and it is objective finally to obtain quantizating index, evaluate tested be emotionally stable grade and/or emotion composition exactly, this emotional stability evaluation result can be used for instructing medical personnel that the patient is carried out the science diagnosis, management and treatment.And the method for classifying modes based on support vector machine used in the present invention can obtain higher classification accuracy, uses cross validation method can obtain to classify more accurately at the method for classifying modes based on support vector machine.Therefore, compared with the prior art, the present invention has the noinvasive of emotional information, dynamic and instantaneous collection, by the sample size characteristic optimization, can accurate, objective, stably evaluate the be emotionally stable grade and the origin cause of formation situation of patient in a certain period.

Description of drawings

Fig. 1 is the composition sketch map of the emotional stability evaluation system based on nuclear magnetic resonance provided by the invention;

Fig. 2 is the composition sketch map of the emotional stability evaluation system based on nuclear magnetic resonance provided by the invention;

Fig. 3 is the method flow diagram of estimating based on the emotional stability of nuclear magnetic resonance provided by the invention;

Fig. 4 is the method flow diagram of emotional stability ranking provided by the invention;

Fig. 5 is the method flow diagram of emotion origin cause of formation evaluation provided by the invention;

Fig. 6 is the component identification process figure of the emotional stability evaluation system based on nuclear magnetic resonance provided by the invention.

The specific embodiment

In order to make purpose of the present invention, technical scheme and advantage clearer,, the present invention is further elaborated below in conjunction with drawings and Examples.Should be appreciated that specific embodiment described herein only in order to explanation the present invention, and be not used in qualification the present invention.

Referring to Fig. 1, the invention provides a kind of emotional stability evaluation system 10 based on nuclear magnetic resonance, comprising:

The feature samples storehouse 11 of being emotionally stable is used to store the emotion grade with different emotional stabilitys and the magnetic resonance structure and the tranquillization attitude functional image attribute of emotion composition;

Nuclear magnetic resonance unit 12 is used to obtain tested individual brain magnetic resonance structure phase and tranquillization attitude function imaging data mutually, and is transferred to feature extraction unit 13;

Feature extraction unit 13 is used for 12 magnetic resonance structures phase and the tranquillization attitude function imaging data of transmitting mutually from described nuclear magnetic resonance unit, extracts the magnetic resonance structure and the tranquillization attitude functional image attribute of described tested individual brain;

Pattern classifier 14, the magnetic resonance structure and the tranquillization attitude functional image attribute that are used for described tested individual brain that described feature extraction unit 13 is extracted carry out the grader training;

Emotional stability evaluation unit 15, be used for grader training result according to described pattern classifier 14, and the described described emotion class information of feature samples storehouse 11 storage and the magnetic resonance structure and the tranquillization attitude functional image attribute of emotion composition of being emotionally stable, evaluate the emotional stability grade and/or the emotion origin cause of formation of described tested individuality.

Emotional stability is meant the individual personal characteristics that variation is controlled and cognitive level shows for situation.To be the individual bearing the individual factors that a certain work institute with certain pressure must consideration for emotional state stable.Problem to be solved by this invention is the standard that lacks subjectivity at present for the evaluation of individual emotional stability, and subjectivity is strong, is subjected to individual variation and many influences that adds factor easily, thereby the influence evaluation is objective and accurate.In addition, the assessment method of existing emotional stability mainly relies on psychological scale to do qualitative judgement, can not reflect the psychological factor that is comprised in the classification of emotional state and the emotion objectively.Therefore, the present invention in advance with the magnetic resonance structure of the emotion grade of the historical different emotional stabilitys of a plurality of tested individualities and emotion composition and tranquillization attitude functional image property store in the feature samples storehouse 11 of being emotionally stable; When the emotion of new tested individuality is estimated, obtain tested individual brain magnetic resonance structure phase and tranquillization attitude function imaging data mutually by nuclear magnetic resonance unit 12, and be transferred to feature extraction unit 13 then; Feature extraction unit 13 the magnetic resonance structure mutually with tranquillization attitude function imaging data mutually in, extract the magnetic resonance structure and the tranquillization attitude functional image attribute of new described tested individual brain; Again magnetic resonance structure and tranquillization attitude functional image attribute are sent to pattern classifier 14 and carry out the grader training, last emotional stability evaluation unit 15 is evaluated the emotional stability grade and the emotion origin cause of formation of new tested individuality according to the grader training result and the described emotion class information of feature samples storehouse 11 storages and the magnetic resonance structure and the tranquillization attitude functional image attribute of emotion composition of being emotionally stable of described pattern classifier 14.This emotion evaluation to new tested individuality has brought accurately to be passed judgment on, for the work or the life of new tested individuality learn to give good guidance.

Provide in the present invention based on SVM (Suport Vector Machine, non-linear support vector machine) pattern classifier 14 is a kind of machine learning methods that are based upon on the Statistical Learning Theory basis, its maximum characteristics are according to structural risk minimization, improve the generalization ability of learning machine as far as possible, promptly obtain little error and still can guarantee test set is independently kept little error by limited training set sample.In addition, because the support vector algorithm is a protruding optimization problem, so locally optimal solution must be a globally optimal solution, this is that other sorting algorithms institute can not reach.Never carry out the SVM pattern classifier 14 that learning training obtains by this kind, use the score of this pattern classifier 14 can be used for differentiating the degree of certain unusual emotional state with emotion stability crowd's sample magnetic resonance structure and tranquillization attitude functional image.

In one embodiment of the invention, SVM is used as pattern classification algorithm, by using Gauss radially basic as kernel function, thereby at first initial selected characteristic attribute is mapped to higher dimensional space and calculates a hyperplane, the distance of this plane of maximization and each Attribute class example in this hyperplane, after finding this hyperplane, SVM can be individual to feature space and check which isolating planar prediction this individuality belongs to and go out a classification that does not indicate sample by shining upon; What input to pattern classifier 14 is that eigenmatrix and class indicate vector, low as 1 expression emotional stability height or-1 expression emotional stability.And emotional stability evaluation unit 15 is with corresponding historical magnetic resonance structure and tranquillization attitude functional image attribute in the magnetic resonance structure of tested individuality and tranquillization attitude functional image attribute and the emotional state feature samples storehouse 11, utilize pattern classifier 14 to discern, assess the emotional stability grade and/or the emotion origin cause of formation of described tested individuality.Preferably, what the method for using cross validations by emotional stability evaluation unit 15 can determine that body one by one belongs to certain emotional stability colony must score value, thereby judge the sexual state of being emotionally stable of this individuality, the method by cross validation can improve the emotional stability grade of confirming tested individuality and the accuracy of the origin cause of formation.

In one embodiment of the invention, described magnetic resonance structure and tranquillization attitude functional image attribute comprise: the subnet/component attribute of nodal community, reflection brain district's modularity or the different component of the function connection attribute of cooperative ability, reflection brain district power of influence between the full brain network attribute of reflection brain overall situation integration ability, the local attribute of the local integration ability of reflection brain, each brain district of reflection brain.And comprise for the brain district of described subnet/component: right frame volume cortex, abdomen outside prefrontal lobe cortex, bilateral temporo cortex, corpus amygdaloideum, caudatum, nucleus accumbens septi, anterior cingutate, Hippocampus, bottom right top, island leaf, posterior parietal cortex, prefrontal cortex, cingulum cortex.

Wherein, described magnetic resonance structure phase attribute comprises: the ectocinerea density that reflects the dense situation of neuron indirectly, ectocinerea volume, alba volume, the brain district deformation field of reflection brain district shape size cases, the skin thickness of complicated connection situation between the reflection neurocyte.Described magnetic resonance function phase attribute comprises: function connection degree, dependency, mutual information etc.Described nodal community comprises: the degree (degree) in some brains district (node), efficient (Effciency) and connectedness (Betweenness), the amplitude index of physiology frequency range between the brain district voxel.Described subnet/component attribute comprises: the importance ranking of the origin cause of formation (ranking value), each stability of dividing, the significance level of the origin cause of formation etc.

By magnetic resonance functional imaging technology studies have shown that human brain: many brain area (being called for short brain district or node) are relevant with the emotion process, and these brain districts comprise prefrontal lobe cortex outside time right frame volume cortex, the abdomen, bilateral temporo cortex, corpus amygdaloideum, caudatum, nucleus accumbens septi, anterior cingutate, Hippocampus, bottom right top, island leaf, posterior parietal cortex, prefrontal cortex, cingulum cortex etc.Emotional stability may change relevant with function connection, structure and the biochemistry in these brain districts.Human brain magnetic resonance imaging data and can detect brain plot structure situations of change such as tested intensity of activation, function bonding strength, density/volume, skin thickness in these brain districts based on the subsequent analysis of dissecting template; Point out that the function relevant be emotionally stable forward or negative sense connects variation, possible structural change and biochemistry changes; Central nervous system's mechanism that identification is relevant with the emotional state origin cause of formation with emotional stability.

Referring to Fig. 2, in one embodiment of the invention, described emotional stability evaluation unit 15 comprises:

The ranking subelement 151 of being emotionally stable, be used for the emotion grade that the magnetic resonance structure of described tested individuality and tranquillization attitude functional image attribute is corresponding with the described feature samples storehouse 11 of being emotionally stable and the magnetic resonance structure and the tranquillization attitude functional image attribute of emotion composition and compare, obtain the grade of being emotionally stable of described tested individuality; And/or

Emotion composition evaluation subelement 152, be used for according to the categorical attribute of the subnet/component of the different emotion compositions of described feature samples storehouse 11 storage of being emotionally stable and the subnet/component attribute of described tested individual brain, carry out described tested individuality the composition identification of emotional stability, obtain the emotional stability origin cause of formation of described tested individuality and the weight of the described origin cause of formation.

The ranking subelement 151 of being emotionally stable is compared the magnetic resonance structure of tested individuality and tranquillization attitude feature capability image attribute and historical magnetic resonance structure and tranquillization attitude functional image attribute, to utilize with generic attribute pattern classifier 14 carry out pattern classification analyze the tested individual magnetic resonance structure and the tranquillization attitude functional image attribute that make new advances, with the historical structure of different emotional stability grade colony and the classification value of tranquillization attitude functional image attribute in the described feature samples storehouse 11 of being emotionally stable, and unite and use pattern classifier 14 to improve classification accuracies magnetic resonance structure attribute and tranquillization attitude functional attributes.Wherein, unite use magnetic resonance structure phase attribute and tranquillization attitude classification value that functional attributes draws belong to a certain colony as individuality degree of membership; Multiple attribute is carried out the confidence level of the degree of variation of repetition Classification and Identification gained classification afterwards as this degree of membership; The ranking value of being emotionally stable=a certain grade degree of membership+confidence level of being emotionally stable of the patient that new described tested individuality is new in other words.The different emotion composition colony that emotion composition evaluation subelement 152 usefulness have been stored based on the feature samples storehouse 11 of being emotionally stable the categorical attribute of historical subnet/component, described new tested individual subnet/component attribute is carried out composition identification, and the principal component analysis technology by having priori is carried out composition with tested individual subnet/component attribute and is discerned to such an extent that make new advances the emotion composition tested.By must the make new advances weight of tested various emotion compositions of the identification of the composition between the same subnet/constituent classification attribute of tested individual subnet/component attribute and different historical grade samples grade.

In one embodiment of the invention, emotional stability evaluation unit 15 also comprises:

Obtain subelement 153, be used to receive the clinical emotion test and appraisal information of tested individuality, and, analyze the emotional stability grade and/or the emotion origin cause of formation of described tested individuality according to magnetic resonance structure and the tranquillization attitude functional image attribute and the described clinical Psychological Evaluation information of described tested individuality; And/or

Fine setting subelement 154 is used to finely tune the relevant parameter of the be emotionally stable grade and/or the emotion origin cause of formation in the described feature samples storehouse 11 of being emotionally stable.

Obtain subelement 153 and obtain tested clinical emotional information, for example receive patient's electronic health record recorded information, perhaps the doctor is by obtaining patient's electronic health record recorded information of subelement 153 inputs.The correction weight of the emotion composition that analyzes the patient in conjunction with the benchmark weight and the clinical emotional information of described emotion composition.And fine setting subelement 154 is by the self adaptation feedback learning method in the feature samples storehouse 11 of being emotionally stable, and for example artificial neural network is finely tuned the emotional stability grade and/or the emotion composition of new tested individuality.Fine setting subelement 154 is finely tuned the fine setting weight of described emotion composition, the weight after being optimized by the self adaptation feedback learning method in the feature samples storehouse 11 of being emotionally stable.Above adjustment mode also is applicable to correction and the fine setting when grade is emotionally stable in ranking subelement 151 evaluations of being emotionally stable.

Emotion composition evaluation subelement 152 is isolated the subnet/component that reflects different emotion compositions by the mode identification technology with priori from full brain network, reflect the weight and the dominance relation of the subnet/component attribute of certain emotion composition respectively.The component that emotion composition evaluation subelement 152 identifies by the component recognition technology is determined the emotion origin cause of formation of tested individuality, and the component recognition technology draws the subnet/component attribute ordering of different brackets between subnet/component attribute of representing this this kind of individuality emotion origin cause of formation, draws the weight of the various emotion origin causes of formation.In the present invention, the described emotion origin cause of formation comprises: personality factor, emotional factor and cognitive factor.

In one embodiment of the invention, described feature extraction unit 13 comprises:

Feature extraction subelement 131 is used for isolating from the full brain network of the brain of described tested individuality the subnet/component of the different emotion compositions of reflection, and calculates described isolated subnet/component attribute, function connection attribute and nodal community;

Divide subelement 132, be used for full brain is divided into a plurality of different brain districts according to dissecting template, then according to described magnetic resonance structure and tranquillization attitude functional imaging data, with described brain district is that unit extracts the active time series in reflection tranquillization attitude brain district, adopts the time series/skin thickness in described brain district to make up full brain network and subnet net again.

Feature extraction unit 13 respectively to structure mutually and tranquillization attitude function picture extract the attribute of different aspects such as network attribute, function connection attribute and nodal community, unite the federation properties that structure phase and tranquillization attitude function is looked like to extract full brain, subnet, three aspects of node simultaneously.Wherein, feature extraction subelement 131 is isolated the subnet/component that reflects different emotion compositions by the component recognition technology with priori from full brain network, calculate isolated subnet/component attribute, function connection attribute and nodal community respectively.Divide 132 on subelement and full brain is divided into several different brain districts according to dissecting template, utilize described dissection template to cut apart resulting described magnetic resonance structure in described functional imaging unit and tranquillization attitude functional imaging data then, with described brain district is the time series that unit extracts reflection tranquillization attitude cerebration, makes up full brain network and subnet net with described brain district and time series/skin thickness.

In one embodiment of the invention, the division subelement 132 of feature extraction unit 13 at first is divided into several different brain districts to full brain according to dissecting template, utilize the dissection template to cut apart nuclear magnetic resonance unit 12 resulting structures and tranquillization attitude functional imaging data then, the Yi Nao district is for unit extracts the time series that reflects cerebration, and require mental skill district and time series make up full brain network and subnet net.Its mode comprises with different brain districts that as node the seasonal effect in time series relation is built with to connection layout as the limit between the Yi Nao district, promptly concern orderliness and sensing between the brain district, and undirected connection layout is the out of order and sensing of relation between the brain district; The various magnetic resonance 26S Proteasome Structure and Functions of the full brain of calculation chart, subnet and node aspect are attribute mutually; The calculating that function connects comprises: determine brain interested district to dissect template or self-defined mode, extract the time series in brain interested district, calculate the dependency relation between the brain interested district.These two pieces of papers of emotional stability network main reference (Moissset et al., 2007) (Mobbs D, Hagan CC, Azim E, Menon V, Reiss AL, the personality foresight activity in award relevant and emotion zone with humour, PNAS, 45:16502-16506,2005; Reske M etal., the dysfunction of being emotionally stable, JOURNAL OF PSYCHIATRIC RESEARCH, 41:918-927,2007).The brain district that feature extraction unit 13 makes up subnet/component comprises: right frame volume cortex, abdomen outside prefrontal lobe cortex, bilateral temporo cortex, corpus amygdaloideum, caudatum, nucleus accumbens septi, anterior cingutate, Hippocampus, bottom right top, island leaf, posterior parietal cortex, prefrontal cortex, cingulum cortex.

In of the present invention other are preferably implemented, when described emotional state feature samples storehouse 11 also is used for storage and is used to make up full brain network and subnet net, function connected node, by described magnetic resonance structure and tranquillization attitude functional imaging data trim values by the intermediate value, emotional stability grade/origin cause of formation evaluation result, clinical Psychological Evaluation information and the emotional stability grade/origin cause of formation that calculate.Carry out when emotional stability is evaluated with reference to comparison and necessary information feedback in this feature samples storehouse 11 of being emotionally stable.Described subnet is meant the combination in several brains district, but its function is indeterminate; And described component also is the combination in several brains district, but has clear and definite function.To the sample in the feature samples storehouse 11 of being emotionally stable through after the study repeatedly, finally make each evaluation result all reach steady statue, thereby guarantee the stability of emotional stability evaluation result.

Referring to Fig. 3, the present invention also provides and has adopted said system to realize the method for estimating based on the emotional stability of nuclear magnetic resonance, and described method comprises:

Among the step S301, nuclear magnetic resonance unit 12 obtains tested individual brain magnetic resonance structure phase and tranquillization attitude function imaging data mutually, and is transferred to feature extraction unit 13;

Among the step S302, feature extraction unit 13 from the magnetic resonance structure of described nuclear magnetic resonance unit 12 transmission mutually with tranquillization attitude function imaging data mutually, extract the magnetic resonance structure and the tranquillization attitude functional image attribute of described tested individual brain;

Among the step S303, pattern classifier 14 carries out the grader training with the magnetic resonance structure and the tranquillization attitude functional image attribute of the described tested individual brain that feature extraction unit 13 is extracted;

Among the step S304, emotional stability evaluation unit 15 is according to the grader training result of pattern classifier 14, and the be emotionally stable described emotion class information of feature samples storehouse 11 storage and the magnetic resonance structure and the tranquillization attitude functional image attribute of emotion composition, evaluate the emotional stability grade and/or the emotion origin cause of formation of described tested individuality.

In the present invention, nuclear magnetic resonance unit 12 can adopt existing MRI (Magnetic ResonanceImaging, nuclear magnetic resonance) equipment to realize.Nuclear magnetic resonance unit 12 from the brain of tested individuality obtain different magnetic resonance imaging datas comprise structure mutually with function mutually, also can be used for obtaining different aspects tested individuality out of phase nuclear magnetic resonance image data and be transferred to feature extraction unit 13.Feature extraction unit 13 is from the magnetic resonance structure of new tested individuality and the magnetic resonance structure and the tranquillization attitude functional image attribute of the tested individuality of tranquillization attitude functional imaging extracting data.For historical structure in the feature samples storehouse 11 of being emotionally stable and tranquillization attitude functional image attribute, also obtain and preserve by feature extraction unit 13.Feature extraction unit 13 extracts the structure and the tranquillization attitude functional image attribute of tested individuality, input to the pattern classifier 14 of support vector machine again, pattern classifier 14 uses the nuclear correspondence course that attribute is carried out nonlinear mapping, attribute is mapped to higher dimensional space, associating utilization structure phase and tranquillization attitude functional attributes trainable pattern classifier 14; Based on the distance in the higher dimensional space hyperplane, pattern classifier 14 can calculate the probability that the individuality that makes new advances belongs to corresponding colony in the feature samples storehouse 11 of being emotionally stable, if the probability score that this new individuality corresponds to this colony more than or equal to 90%, then this new tested individuality just belongs to this colony; 1-probabilistic classification score has been represented this individual abnormal level; Unite and use grader to improve classification accuracy structure attribute and tranquillization attitude functional attributes; Repeatedly use grader to draw the classification confidence level of the degree of variation of classification value as this individuality.

In the present invention, adopt method for classifying modes to use cross validation method to improve the classification accuracy of pattern classifier 14 based on support vector machine.Cross validation is to stay a cross validation method can calculate the unusual probability distribution of each individuality by use, in staying a cross validation, a sample is selected to test, and remaining sample is used to trainable pattern classifier 14, and the quality of pattern classifier 14 is assessed with the individual classification results of test; Individual by the multiple Different Individual of taking away in the colony as test, obtain to belong to its unusual score, and the average score of all individualities of taking away, obtain an average classification rate; Checking is used simultaneously draws experimenter's performance curve and carries out, experimenter's performance curve be one when distinguish threshold value not simultaneously sensitivity as abscissa, specificity grader curve chart as vertical coordinate.

The ranking subelement 151 of being emotionally stable is compared the magnetic resonance structure of tested individuality and tranquillization attitude functional image attribute and historical magnetic resonance structure and tranquillization attitude functional image attribute, to utilize pattern classifier 14 must the make new advances historical magnetic resonance structure of the different grade colonies that are emotionally stable and the classification value of tranquillization attitude function tranquillization attitude functional image attribute in individual magnetic resonance structure and tranquillization attitude functional image attribute and the feature samples storehouse 11 of being emotionally stable with generic attribute, at first unite the attribute training classifier in the pattern classifier 14 that uses magnetic resonance structure phase attribute and tranquillization attitude function, use associating tagsort classification value that device draws in the pattern classifier 14 to belong to the degree of membership of a certain colony, as final classification value D (KP) as individuality; To carrying out the confidence level C (C%) of the degree of variation of repetition Classification and Identification gained classification afterwards with generic attribute in the multiple feature samples storehouse 11 of being emotionally stable as this degree of membership D (KP).The final new tested ranking value of the being emotionally stable=a certain grade separation value D (KP) that is emotionally stable+confidence level (C%).

Studies show that, different emotion composition may be corresponding different subnet/components, detects different subnets/component activation degree and can determine composition factor in the emotion.At tested structure and tranquillization attitude functional imaging data, adopt PCA (Principal component analysis, principal component analysis) isotype recognition technology to detect different sub-network/component in conjunction with priori and activate degree; By having the component that principal component analysis drew of priori classificating knowledge, can be defined as the heterogeneity factor in the emotion; And, can determine the shared proportion of various emotion compositions by difference between more same individual different sub-network/component and registration ordering.

Referring to Fig. 4, in one embodiment of the invention, the assessment system 100 of being emotionally stable carries out the flow process of emotional stability ranking:

Among the step S401, the brain of the 12 pairs of tested individuals in nuclear magnetic resonance unit carries out the magnetic resonance structure and tranquillization attitude functional imaging data are gathered;

Among the step S402,12 pairs of magnetic resonance structures of being gathered in nuclear magnetic resonance unit and tranquillization attitude functional imaging data are carried out pretreatment;

Among the step S403, feature extraction unit 13 is extracted the attribute character of a plurality of aspects such as the full brain network attribute of tested individuality and the full brain network attribute of normal person from magnetic resonance structure and tranquillization attitude functional imaging data; Choose emotional characteristics network (and subnet) according to the emotion composition, be aided with clinical emotion scale and other information, calculate the classification value of emotional characteristics network and node image feature.

Among the step S404, pattern classifier 14 is determined the emotional stability calibration algorithm according to emotion composition, clinical emotional information, basic image feature by algorithms library, and utilizes described algorithm to calculate, and obtains the class parameter value of being emotionally stable.

Among the step S405, demarcate the emotional stability grade point according to image feature.The feature samples storehouse 11 establishment stage emotional stability measured values of being emotionally stable.

Among the step S406, compare according to be emotionally stable feature samples storehouse 11 establishment stage evaluation values and clinical assays value, both match condition and the ranking value of being emotionally stable all deposit the feature samples storehouse 11 of being emotionally stable in.

Among the step S407, pattern classifier 14 carries out cross validation, to the sample process multi-level artificial neuron learning network correction and the fine setting in the feature samples storehouse 11 of being emotionally stable, the emotional stability ranking value of finally being measured.

In one embodiment of the invention, provide the instantiation of the grade classification of being emotionally stable, be described below:

Support vector machine (SVM) algorithm uses the modes such as weighted linear combination of expression as the arbiter between two classes.This linear combination can maximize the poorest sample of classification performance and differentiate interplanar border or distance.All data base's sample individualities are used to train the SVM algorithm under the initial situation.Judge emotional stability grade process, at first the SVM algorithm can remove the less tested individual of weight absolute value in the linear combination, and constructs new SVM pattern classifier 14 on the residue individuality.After finding a new linear discriminant plane, remove the less tested individual of weight absolute value in the linear combination again, and on the residue individuality, construct another new SVM pattern classifier 14.This process repeats all to be used to structure SVM pattern classifier 14 up to all samples.

The SVM algorithm is that the optimal classification face (Optimal Hyperplane) under the linear separability situation proposes.So-called optimal classification face requires classifying face not only two class sample points can be separated error-free exactly, and will make the classification space maximum of two classes.The general type of d dimension space neutral line discriminant function is g (x)=w ^TX+b, the classifying face equation is w ^TX+b=0, we carry out normalization with discriminant function, and all samples of two classes are all satisfied | g (x) | 〉=1, this moment is from the nearest sample of classifying face | g (x) |=1, and require classifying face that all samples can both correctly be classified, require it to satisfy exactly

y _i(w ^Tx _i+b)-1≥0，i＝1，2，...，n。(1)

Those samples that equal sign is set up are called support vector (Support Vectors).The gap size in the classification space (Margin) of two class samples:

Margin＝2/||w||(2)

Therefore, optimal classification face problem can be expressed as following constrained optimization problems, promptly under the constraint of condition (4), asks function

$\mathrm{\φ}\left(w\right)=\frac{1}{2}{\left|\right|w\left|\right|}^2=\frac{1}{2}\left(w^{T}w\right)---\left(3\right)$

Minima.

After the training aids of pattern classifier 14 has been constructed, the feature screening is all independently carried out on the training set in circulation each time, (Leave-one-out Cross-Validation LOOCV) is used for judging at support vector machine classification prediction algorithm which emotional stability grade given sample belongs to actually to use the leaving-one method cross validation.The mistake branch rate of the cross validation of pattern classifier 14 is calculated and output as a result of, in LOOCV, cross-validation process is taken out a sample at every turn, carries out once complete analysis with remaining data set then, comprises which tested individual of decision is remarkable in training set.The pattern classifier 14 of support vector machine makes up and is used to the sample of that taking-up in the tested tabulation that obtains.Whether program can write down prediction this time correct, and this process is carried out repeatedly then, and is all predicted up to all samples.The output form has shown which sample is by classification of correct/error ground and the overall wrong branch rate of cross validation.

Final pattern classifier 14 obtains the classification value that this individuality emotional characteristics vector belongs to certain sample criterion group of being emotionally stable, equally, carry out the component efficiency attribute again and the calculating of the classification value of this attribute of sample criterion group of being emotionally stable, the classification value that draws belongs to the confidence level of certain emotion sample criterion group as this individuality.Be input as example with two dimension, the model input is respectively " the activation degree of feature brain district K " and the grade of being emotionally stable of clinical emotion evaluation after handling, and the model of emotional stability evaluation unit 15 is output as the grade of being emotionally stable.Overstable, stable, basicly stable, unstable, the extremely unstable or score value scope of being emotionally stable that the class set of being emotionally stable can be, the classification function of this subclass of various kinds is a Gauss distribution, according to the activated situation in brain district, its grade of being emotionally stable can be from " super steady " until change between " extremely unstable ".By learning the emotion clinical and image information adjustment function in feature samples storehouse quietly, make the sensitivity of being emotionally stable between the grade pitch different, extremely unstable the highest, make early diagnosis and intervene more pregnancy to its sensitivity between the instability.Through categorizing process, can obtain the grade of being emotionally stable of a non-zero at last, the grade of being emotionally stable this moment mapping is output as: the grade of being emotionally stable D (KP)+confidence level (C%).

Suppose to exist in the full brain network attribute n separate component x (t)=[x ₁(t), x ₂(t) ..., x _n(t)] ^T, the component identification of full brain in fact also is categorizing process, purpose is labeled as a new individual D predefined N kind component type C={C exactly ₁, C ₂.。。C _NIn one or more, training algorithm is exactly to form in part set in one of the good type of labelling to obtain an effective classification function F:D → C, and new individuality is mapped to type of coding.Choose activation degree attributes such as (spontaneous activity levels) as the recognition feature attribute, according to independence assumption in the model-naive Bayesian, each characteristic attribute is independently to the contribution of classification, does not have influence to each other.At first in full brain network attribute, find P characteristic point representing the emotion general character in the identifying, (following the example of of P can exert an influence to recognition effect) according to priori.Certainly, the big more selected characteristic of P value is many more, and recognition accuracy is also high more, but the also corresponding increase of computation complexity.In actual identifying, P comes value according to priori (the brain district number of emotion network, characteristic attribute number) and empirical equation, (from 10 to 30 incremental variations are observed the influence of P value to recognition result stability, employed P value scope when selected stability is the highest).

Can obtaining by the random function that structure is dissected in the template brain district number at random of P characteristic point, construct relatively difficulty of a good random function, the more difficult true random of accomplishing when especially being confined to dissect the masterplate number, so the recognition strategy that P is ordered in the actual moving process is to choose non-conterminous P point as characteristic point in full brain scope and subnet/component, such P point is dispersed in each position, all the probability of the same feature of representative is very little, and because the distribution probability difference of different characteristic in various emotion compositions, so have the difference degree for different emotion components.

Referring to Fig. 5, in one embodiment of the invention, emotional stability assessment system 10 carries out the flow process of emotion composition evaluation:

Among the step S501, nuclear magnetic resonance unit 12 obtains the magnetic resonance structure and the tranquillization attitude functional imaging data of tested individuality;

Among the step S502,12 pairs of described acquisition magnetic resonance structures in nuclear magnetic resonance unit and tranquillization attitude functional imaging data are carried out pretreatment, and the described magnetic resonance structure after will handling and tranquillization attitude functional imaging transfer of data are to feature extraction unit 13;

Among the step S503, feature extraction unit 13 is extracted image features such as full brain network attribute according to described magnetic resonance structure and tranquillization attitude functional imaging data;

Among the step S504, pattern classifier 14 extracts injury personality factor component, emotional factor component, cognitive factor component respectively in conjunction with priori, subnet/component attribute that calculating image feature is worth going out compares through after the normalization, and is aided with clinical emotional information;

Among the step S505, emotional stability evaluation unit 15 draws the component of the emotion origin cause of formation (type) that can represent this individual patient according to the training result of pattern classifier 14.Consider difference of specific gravity between the different emotion compositions, the difference of the more a certain component and the feature samples storehouse 11 similar component different brackets samples of being emotionally stable draws the weight of every kind of origin cause of formation at last.Comprise personality composition and proportion, emotion composition and proportion, cognitive component and proportion.

Concrete, can carry out that subnet/component is divided and node calculates according to the emotion composition, same individuality is carried out lateral comparison between different sub-network/component attribute, carry out longitudinal comparison between same subnet/component attribute for Different Individual; Determine the leading composition (personality, emotionality or awareness) and the different origins proportion of this individuality emotion more respectively twice.

In one embodiment of the invention, provide the instantiation of an emotion origin cause of formation identification, be described below:

Feature extraction unit 13 at first defines the characteristic attribute X1 that extracts, X2 ..., Xp, its sample average is designated as

Sample standard deviation is designated as S1, S2 ..., Sp.At first do the standardization conversion

$x_i=\frac{X_i-{\stackrel{\‾}{X}}_i}{S_i}$

Following definition is arranged:

(1) as if Y1=a11x1+a12x2+...+a1pxp,

And make Var (Y1) maximum, claim that then Y1 is a first principal component;

(2) as if Y2=a21x1+a22x2+...+a2pxp,

(a21, a22 ..., a2p) perpendicular to (a11, a12 ..., a1p), and make Var (Y2) maximum, claim that then Y2 is a Second principal component;

(3) similarly, can have third and fourth, five ... main constituent, we have 3 main constituents in this definition.After as above defining, emotion composition identifying is as follows:

1, establishing the ATTRIBUTE INDEX of observing individuality is x1, x2 ..., xp, their component target---main constituent is zl, z2 ..., zm (m≤p), then

$\left\{\begin{array}{c}{z}_1=l_{11}{x}_1+l_{12}{x}_2+\·\·\·+l_{1p}{x}_p\\ \·\·\·\\ z_m=l_{m1}{x}_1+l_{m2}{x}_2+\·\·\·+l_{\mathrm{mp}}{x}_p\end{array}\right.---\left(1\right)$

z ₁, z ₂..., z _mBe called former attribute index x ₁, x ₂..., x ₆The first, the second ..., the m main constituent.

2. initial data is collected in nuclear magnetic resonance unit 12, gets following tables of data:

Be provided with stochastic variable x ₁, x ₂..., x _p, its sample average is designated as

Sample standard deviation is designated as S ₁, S ₂..., S _pAt first do the standardization conversion

$x_i=\frac{X_i-{\stackrel{\‾}{X}}_i}{S_i}---\left(2\right)$

3, pattern classifier 14 calculates correlation matrix, characteristic of correspondence value λ ₁λ _p(by arranging from big to small) and characteristic of correspondence vector thereof

$R=\left(\begin{array}{ccc}{r}_{11}& {\mathrm{<figure-callout\; id="12"\; label="r"\; filenames="HSA00000505150300011.png,HSA00000505150300012.png"\; state="\{\{state\}\}">r</figure-callout>}}_{12}\&CenterDot;\&CenterDot;\&CenterDot;& r_{1p}\\ r_{21}& r_{22}\&CenterDot;\&CenterDot;\&CenterDot;& r_{2p}\\ M& M& M\\ r_{p1}& r_{p2}\&CenterDot;\&CenterDot;\&CenterDot;& r_{\mathrm{pp}}\end{array}\right)---\left(3\right)$

R wherein _Ij(i, j=1,2 ..., p) be primal variable x _iWith x _jCorrelation coefficient, its computing formula is

$r_{\mathrm{ij}}=\frac{\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}(x_{\mathrm{ki}}-{\stackrel{\&OverBar;}{x}}_i)(x_{\mathrm{kj}}-{\stackrel{\&OverBar;}{x}}_j)}{\sqrt{\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}{(x_{\mathrm{ki}}-{\stackrel{\&OverBar;}{x}}_i)}^2\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}{(x_{\mathrm{kj}}-{\stackrel{\&OverBar;}{x}}_j)}^2}}---\left(4\right)$

Because R is a real symmetric matrix (is r _Ij=r _Ji), so only need calculate on it triangle element or down the triangle element get final product.

4, computation of characteristic values and characteristic vector

At first separating characteristic equation | λ I-R|=0 obtains eigenvalue _i(i=1,2 ..., p), and make its sequence arrangement by size, i.e. λ ₁〉=λ ₂〉=..., 〉=λ _p〉=0; Obtain corresponding to eigenvalue respectively then _iCharacteristic vector e _i(i=1,2 ..., p).

(3) calculate main constituent contribution rate and contribution rate of accumulative total

Main constituent z _iContribution rate:

Contribution rate of accumulative total:

Generally get contribution rate of accumulative total and reach the eigenvalue of 85-95% ₁, λ ₂..., λ _mIt is pairing the first, the second ..., m (m≤p) individual main constituent.As: by the correlation matrix computation of characteristic values, and the contribution rate of each main constituent and contribution rate of accumulative total (seeing the following form).By following table as can be known, the contribution rate of accumulative total of the first, the second, the 3rd main constituent is up to 86.5%, so a demand goes out the first, the second, and the 3rd main constituent z ₁, z ₂, z ₃Get final product.Following table is eigenvalue and main constituent contribution rate:

5, emotional stability evaluation unit 15 is calculated main origin cause of formation load

$p(z_k,x_i)=\sqrt{{\mathrm{\&gamma;}}_k}{e}_{\mathrm{ki}}(i,k=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;,p)---\left(5\right)$

Can further calculate principal component scores thus:

$Z\left(\begin{array}{ccc}{z}_{11}& {\mathrm{<figure-callout\; id="12"\; label="z"\; filenames="HSA00000505150300011.png,HSA00000505150300012.png"\; state="\{\{state\}\}">z</figure-callout>}}_{12}\&CenterDot;\&CenterDot;\&CenterDot;& z_{1m}\\ z_{21}& z_{22}\&CenterDot;\&CenterDot;\&CenterDot;& z_{2m}\\ M& M& M\\ z_{n1}& z_{n2}\&CenterDot;\&CenterDot;\&CenterDot;& z_{\mathrm{nm}}\end{array}\right)---\left(6\right)$

Following table is the main constituent loading matrix:

6. emotional stability evaluation unit 15 is carried out the interpretation of result similar type:

First principal component z1 and x1, x3, x4, x5, x8, x9 have bigger positive correlation, can be regarded as the emotion composition representative in the emotional stability;

Second principal component, z2 and x2 have bigger positive correlation, with x7 bigger negative correlation are arranged, and can be regarded as the personality composition representative in the emotional stability;

The 3rd main constituent z3 and x6 have bigger positive correlation, can be regarded as the cognitive component representative in the emotional stability.

Referring to Fig. 6, based in the method for classifying modes of support vector machine from training to the flow process of discerning whole process, be described below:

Among the step S601, feature extraction unit 13 is chosen P characteristic point from individuality to be identified;

Among the step S602, the characteristic frequency table that pattern classifier 14 is concentrated according to feature database component training sample calculates the different component product of probability;

Among the step S603, emotional stability evaluation unit 15 is selected component of candidate;

Among the step S604, emotional stability evaluation unit 15 is according to each component threshold values that feature database component training sample set provides, and judges component whether in the threshold values scope, if execution in step S605 then, if not execution in step S606 then.

Among the step S605, emotional stability evaluation unit 15 output components;

Among the step S606, emotional stability evaluation unit 15 is thought unknown component with it.

By repeatedly component identification, finally obtain the importance ranking of (1) origin cause of formation: utilize the dependency repeatedly move between the same origin cause of formation that component analysis obtains, try to achieve the stability indicator of the variance of every kind of origin cause of formation with the tolerance different origins, utilize this index and the distribution on Different Individual, calculate the general stability index of every kind of origin cause of formation, draw the ordering curve chart of all origin causes of formation on Different Individual and overall ordering curve chart; (2) each stability of dividing: take turns component identification more and obtain a plurality of origin causes of formation, utilize the correlation coefficient between the origin cause of formation and carry out the multi-dimentional scale analysis and show similarity between the same origin cause of formation that repeatedly operation obtains; (3) significance level of the origin cause of formation: obtain all origin causes of formation by cartogram (mean chart, T check figure, standard deviation figure) or individual origin cause of formation scattergram, calculate the significant indexes in the population sample composition of repeatedly dividing certain individual origin cause of formation of back.

Utilize emotional stability evaluation system 10 provided by the invention to carry out two researchs, first research is about using the support vector machine sorting algorithm whether can discern the patient colony of different depressed grades, research contents comprises to be analyzed the intractable depressive patient (RDD) and the cerebral cortex MRI feature of the non-intractable depressive patient of pairing (NDD), select the RDD (23 example) and NDD (23 example) age of two groups, sex, the patient that the order of severity of pretreat and depression is complementary, utilize the support vector machine sorting algorithm that two groups ectocinerea characteristic is classified, reached 67.39% accuracy rate of diagnosis; Further, integrated structure and tranquillization attitude functional image index and brain structure index utilize the support vector machine sorting algorithm that RDD and NDD patient are classified equally, and its accuracy rate is up to 92.77%.

Have the function MR investigation to show: different emotion composition may be corresponding different function subnets, detects different function subnet state of activation (specifically threshold value) and may determine the shared proportion of this kind composition in the composition of emotion or the emotional state.These researchs are about judging whether the imaging of magnetic resonance tranquillization attitude can be used to discern different emotion compositions, and research contents comprises tested emotion loop/network (emotion matrix), corpus amygdaloideum, anterior cingutate, posterior cingutate, prefrontal cortex, cortex of temporal lobe and the thalamus that can react emotional state.Result of study is to have set up the pattern that can distinguish different emotion compositions.

Second research is about judging the image degree of each image index to dysthymic disorder's disease with fuzzy comprehensive evaluation method, three image index: 1-Cost, 2-Eglobal as shown in the table, 3-Elocal.

(2) calculate every group average Ridit value

Result of calculation,

Be worth littler, big more to the influence degree of disease; Otherwise,

Value is bigger, and is more little to the influence of disease.Research conclusion:

Illustrate that image index 3-Elocal is bigger to the influence degree of disease.

In sum, the present invention is by gathering the objective image data of structure and tranquillization attitude magnetic resonance functional imaging gained, adopt the information processing technology therefrom to extract tested new individual configurations and tranquillization attitude functional image attribute, and will be somebody's turn to do with the feature samples storehouse of being emotionally stable in historical structure and the comparing of tranquillization attitude functional image attribute, preferably by pattern classification, methods such as self adaptation feedback learning are classified to image data eigenvalue and clinical emotional information, extract, optimize, the comprehensive assessment system that becomes to reflect objective the be emotionally stable grade and the origin cause of formation, and it is objective finally to obtain quantizating index, evaluate tested be emotionally stable grade and/or emotion composition exactly, this emotional stability evaluation result can be used for instructing medical personnel that the patient is carried out the science diagnosis, management and treatment.And the method for classifying modes based on non-linear support vector machine used in the present invention can obtain higher classification accuracy, uses cross validation method can obtain to classify more accurately at the method for classifying modes based on support vector machine.Therefore, compared with the prior art, the present invention has the noinvasive of emotional information, dynamic and instantaneous collection, by the sample size characteristic optimization, can accurate, objective, stably evaluate the be emotionally stable grade and the origin cause of formation situation of patient in a certain period.

Certainly; the present invention also can have other various embodiments; under the situation that does not deviate from spirit of the present invention and essence thereof; those of ordinary skill in the art work as can make various corresponding changes and distortion according to the present invention, but these corresponding changes and distortion all should belong to the protection domain of the appended claim of the present invention.

Claims (10)

1. the emotional stability evaluation system based on nuclear magnetic resonance is characterized in that, comprising:

The feature samples storehouse of being emotionally stable is used to store the emotion grade with different emotional stabilitys and the magnetic resonance structure and the tranquillization attitude functional image attribute of emotion composition;

The nuclear magnetic resonance unit is used to obtain tested individual brain magnetic resonance structure phase and tranquillization attitude function imaging data mutually, and is transferred to feature extraction unit;

Feature extraction unit is used for magnetic resonance structure phase and tranquillization attitude function imaging data mutually from the transmission of described nuclear magnetic resonance unit, extracts the magnetic resonance structure and the tranquillization attitude functional image attribute of described tested individual brain;

Pattern classifier, the magnetic resonance structure and the tranquillization attitude functional image attribute that are used for described tested individual brain that described feature extraction unit is extracted carry out the grader training;

The emotional stability evaluation unit, be used for grader training result and the described emotion class information of the described feature samples library storage of being emotionally stable and the magnetic resonance structure and the tranquillization attitude functional image attribute of emotion composition, evaluate the emotional stability grade and/or the emotion origin cause of formation of described tested individuality according to described pattern classifier.

2. the emotional stability evaluation system based on nuclear magnetic resonance according to claim 1, it is characterized in that described magnetic resonance structure and tranquillization attitude functional image attribute comprise: the subnet/component attribute of nodal community, reflection brain district's modularity or the different component of the function connection attribute of cooperative ability, reflection brain district power of influence between the full brain network attribute of reflection brain overall situation integration ability, the local attribute of the local integration ability of reflection brain, each brain district of reflection brain.

3. the emotional stability evaluation system based on nuclear magnetic resonance according to claim 2, it is characterized in that the brain district of described subnet/component comprises: right frame volume cortex, abdomen outside prefrontal lobe cortex, bilateral temporo cortex, corpus amygdaloideum, caudatum, nucleus accumbens septi, anterior cingutate, Hippocampus, bottom right top, island leaf, posterior parietal cortex, prefrontal cortex, cingulum cortex.

4. the emotional stability evaluation system based on nuclear magnetic resonance according to claim 2 is characterized in that, described emotional stability evaluation unit comprises:

The ranking subelement of being emotionally stable, be used for the emotion class information that the magnetic resonance structure of described tested individuality and tranquillization attitude functional image attribute is corresponding with the described feature samples storehouse of being emotionally stable and the magnetic resonance structure and the tranquillization attitude functional image attribute of emotion composition and compare, obtain the grade of being emotionally stable of described tested individuality; And/or

Emotion composition evaluation subelement, be used for according to the categorical attribute of the subnet/component of the different emotion compositions of the described feature samples library storage of being emotionally stable and the subnet/component attribute of described tested individual brain, carry out the composition identification of the emotional stability of described tested individuality, obtain the emotional stability origin cause of formation of described tested individuality and/or the weight of the described origin cause of formation.

5. the emotional stability evaluation system based on nuclear magnetic resonance according to claim 4 is characterized in that, described emotional stability evaluation unit also comprises:

Obtain subelement, be used to receive the clinical emotion test and appraisal information of tested individuality, and, analyze the emotional stability grade and/or the emotion origin cause of formation of described tested individuality according to magnetic resonance structure and the tranquillization attitude functional image attribute and the described clinical Psychological Evaluation information of described tested individuality; And/or

The fine setting subelement is used to finely tune the relevant parameter of the be emotionally stable grade and/or the emotion origin cause of formation in the described feature samples storehouse of being emotionally stable.

6. the emotional stability evaluation system based on nuclear magnetic resonance according to claim 2 is characterized in that described feature extraction unit comprises:

The feature extraction subelement is used for isolating from the full brain network of the brain of described tested individuality the subnet/component of the different emotion compositions of reflection, and calculates described isolated subnet/component attribute, function connection attribute and nodal community;

Divide subelement, be used for full brain is divided into a plurality of different brain districts according to dissecting template, then according to described magnetic resonance structure and tranquillization attitude functional imaging data, with described brain district is that unit extracts the active time series in reflection tranquillization attitude brain district, adopts the time series/skin thickness in described brain district to make up described full brain network and subnet net again.

7. the emotional stability evaluation system based on nuclear magnetic resonance according to claim 6, it is characterized in that, when described emotional state feature samples storehouse also is used for storage and is used to make up described full brain network and subnet net, function connected node, by described magnetic resonance structure and tranquillization attitude functional imaging data trim values by the intermediate value, emotional stability grade/origin cause of formation evaluation result, clinical Psychological Evaluation information and the emotional stability grade/origin cause of formation that calculate.

8. the emotional stability evaluation system based on nuclear magnetic resonance according to claim 3 is characterized in that, the described emotion origin cause of formation comprises: personality factor, emotional factor and cognitive factor.

9. the emotional stability evaluation system based on nuclear magnetic resonance according to claim 1 is characterized in that described pattern classifier is non-linear support vector machine classifier;

Described pattern classifier carries out cross validation to described grader training result.

10. one kind is utilized any one system of claim 1～9 to realize the method that emotional stability is estimated, and it is characterized in that described method comprises:

A, nuclear magnetic resonance unit obtain tested individual brain magnetic resonance structure phase and tranquillization attitude function imaging data mutually, and are transferred to feature extraction unit;

B, feature extraction unit from the magnetic resonance structure of described nuclear magnetic resonance unit transmission mutually with tranquillization attitude function imaging data mutually, extract the magnetic resonance structure and the tranquillization attitude functional image attribute of described tested individual brain;

C, pattern classifier carry out the grader training with the magnetic resonance structure and the tranquillization attitude functional image attribute of the described tested individual brain that described feature extraction unit is extracted;

D, emotional stability evaluation unit are according to the grader training result of described pattern classifier, and the magnetic resonance structure of the described emotion class information of the described feature samples library storage of being emotionally stable and emotion composition and tranquillization attitude functional image attribute, evaluate the emotional stability grade and/or the emotion origin cause of formation of described tested individuality.

Images