CN103040467A - Processing method of functional magnetic resonance imaging data for checking Chinese functional zones - Google Patents

Abstract

The invention relates to the technology of medical image aiding, in particular to a processing method of functional magnetic resonance imaging data for checking Chinese functional zones. The method includes the steps of firstly, acquiring blood oxygenation level dependent functional magnetic resonance imaging data and diffusion tensor imaging data of a to-be-checked object; secondly processing blood oxygenation level dependent functional magnetic resonance imaging data in task state; thirdly processing blood oxygenation level dependent functional magnetic resonance imaging data in resting state; and fourthly processing diffusion tensor imaging data.

Description

A kind of speaking Chinese language district checks and examine the processing method with fMRI data

Technical field

The present invention relates to the medical image ancillary technique, in particular a kind of speaking Chinese language district checks and examine the processing method with fMRI data.

Background technology

Language is human distinctive senior cognitive competence, and brain is as the material base of implementation language process, the course of processing of wide participation language.Brain diseases has become one of disease of serious threat China people's life health (such as the cerebrovascular disease annual morbidity 150-200/10 ten thousand people), there are every year a large amount of client need to carry out cerebral operations, although can successfully perform the operation to the patient who relates to multiple brain lesions or damage on the clinical medicine, the brain diseases patient who survives often leaves over the sequela such as hemiplegia, aphasis.Along with China's expanding economy, the people constantly promotes the pursuit of quality of life, and the patient more nourishes senior cognitive functions such as keeping language in the hope of the tight demand of more satisfied postoperative life quality except life is preserved in hope in cerebral surgery operation.Therefore, the impaired problem of needing clinically solution badly that become in the cognitive function brain districts such as language that effective efficiency causes in detecting to avoid performing the operation as far as possible in precise function location and the art before the art.Although, in recent years since, the armarium that China implements cerebral operations improves, medical level also steadily improves, but, clinical understanding to Brain ﹠ Language function Relations Among relatively lags behind, and the reference standard that relates to the department of cerebral surgery resection operation of language function still mainly restrains the classical conclusion in the brain districts such as (Wernicke) district based on the west to Bu Luoka (Broca) district and Weir Buddhist nun, and being difficult to provides accurately surgical guidance for domestic clinical doctor.

Nuclear magnetic resonance (magnetic resonance imaging, MRI) check the research that is used in large quantities patient's structure pathological changes, neuroimaging technology (the structure of different modalities in recent years, diffusion and functional imaging) be widely used in patient's research, wherein, blood oxygen level relies on functional mri (blood oxygen level dependent-functional MRI, BOLD-fMRI) be take deoxyhemoglobin as the endogenous contrast medium, the signal that utilizes the variation of regional flow's deoxyhemoglobin to cause strengthens to locate the image technology of brain domain.And diffusion tensor (diffusion tensor imaging, DTI) has reflected better that as the imaging technique of non-invasive reflection biological tissue water diffusion the microstructure of alba changes.It reflects in vivo water diffusion ability power and the direction of motion by utilizing the diffusion-sensitive gradient of a plurality of different sizes.Adopt the DTI technology can indirectly reflect the integrity of myelinization degree or fibre bundle, Ultrastructural change can pass through that the dispersion tensor index--the quantitative responses such as MD, FA and characteristic value λ 1, λ 2, λ 3 out behind the brain tissue impairment.Zoopery, pathology and iconography research show that all AD alba occurs destroys, the changes such as appearance demyelination, axon are lost, Reactive gliosis, the obstacle that this means limiting moisture motion is lost and is organized anisotropic change, and it is early stage to betide disease.

In recent years, along with the fMRI technology to patient's data processing method and analysis means day by day ripe (Price et al., 2006; Crinion et al., 2007; Suzanne et al., 2007; Marrelec et al., 2008), in developed country, the cerebral function imaging basic research is combined with clinical operation becomes an important trend with the precision that improves functional localization.For example use the front fMRI technology of art to the functional localization result of motor region cortex, reached 100% coincide (Roessler et al., 2005) with direct cortex electricity irritation monitoring result in the art.Compare simple motor function, although the laterality feature of language and functional localization performance thereof are more complicated, but along with the in the preoperative application checked and examined of language function of a large amount of language task normal forms, the functional image result of language task also becomes one of diagnostic criteria of language function laterality and functional localization before the art gradually, for the formulation of clinical operation scheme and rehabilitation programme provides reference.But, owing to there is the difference between language, research conclusion take western alphabetic writing as the basis can not directly apply to or simply by inference with it the Chinese that use Chinese written language, find that such as research the left side middle frontal gyrus is in western alphabetic writing different from the processing mechanism of Chinese character (Tan et al., 2005).Therefore, we are necessary to design the rationally effective Chinese language of a cover and check and examine system, and utilize Functional magnetic resonance imaging to obtain to be suitable for checking and examining the method in Chinese language function district, it is used for before the art of clinical cerebral surgery operation that the language function district checks and examine and during the art language function measures.

Summary of the invention

The purpose of this invention is to provide a kind of speaking Chinese language district and check and examine the processing method of using fMRI data.

Check and examine the processing method method of using fMRI data according to speaking Chinese language of the present invention district, may further comprise the steps:

(1) blood oxygen level dependence fMRI data and the diffusion tensor data of object are checked and examined in acquisition;

(2) blood oxygen level of Processing tasks attitude relies on fMRI data

(2-1) three-dimensional reconstruction of structure picture and function picture carries out the Fourier conversion with the performance data of the K space form preserved in the scanner;

(2-2) singular data is processed

Front 4 performance datas that removal collects are to avoid the magnetic field de-stabilising effect at the initial stage of scanning;

(2-3) time smoothing

Use blood oxygen level to rely on displacement information between aspect in the fMRI data, the voxel time series of input data set is shifted, make each aspect snap to the same time initial point;

(2-4) dynamic(al) correction and space smoothing

Take last seasonal effect in time series brain position as benchmark, other all seasonal effect in time series brains are with it alignment all, each time point obtains respectively the changing value of 6 directions like this, be moved beyond 1 voxel or rotate surpass 5 ° tested data and all reject, 6 column data are saved as file, carry out the space smoothing processing with getting data after the dynamic(al) correction, overall with half high value is set as 6;

(2-5) seasonal effect in time series standardization

With the performance data behind the space smoothing, doing head clearing processes, obtain the mask file, carry out the average of mask for a plurality of speaking Chinese language test assignments, be mask for a plurality of tested and tested Multiple-Scans again and merge processing, described merging processing mode is the time series of baseline with data normalization to 100 for the performance data in the mask is got the average analysis at last;

(2-6) regression analysis of performance data

Adopt top hat function, referring to following formula:

H(t)=∫ ₀ ^min(t，d)h(t-s)dt；

h(t)=(t/4) ⁴e ^4-t

Wherein, t is the independent variable of time point, and s is certain set time point, and d is stimulus duration, and h(t) representative is the intensity level of in time t variation, H(t) then represents time dependent integrated value,

Top hat function and stimulation sequence task time are carried out convolutional calculation, thus the shock response model of the task of generation, and specific algorithm is as follows:

$y\left(t\right)={\∫}_0^{t}f\left(T\right)h(t-T)\mathrm{dT}$

Wherein, t is the independent variable time point, and the time series of f (T) for stimulating h(t-T) is top hat function, y(t) is time dependent convolution value.

The blood oxygen level that analysis is obtained relies on fMRI data and model obtained above is done regression analysis, namely obtains the β value of complete each voxel of brain, and this numerical value can characterize this voxel and get dependency relation between time series and the model;

(3) blood oxygen level of processing tranquillization attitude relies on fMRI data

The pretreatment process with above-mentioned (2-1) to (2-5), calculate on this basis tranquillization attitude low-frequency oscillation amplitude, time series is converted into the frequency time domain by fast Fourier transform, calculate the square root of crest height, then the square root of each voxel in 0.01～0.08Hz scope averaged, square root after average is tranquillization attitude low-frequency oscillation amplitude A LFF, analyze ALFF value change situation, obtain corresponding Statistical Parametric Mapping, time series in the fMRI interesting image district is averaged, obtain reference time array, in the mode of voxel-voxel different reference time arrays are carried out correlation analysis, and use the Fisher conversion to improve the normality of correlation coefficient, the correlation coefficient that the relatively difference of the bonding strength between each crucial brain district, and measurement Different brain region connects;

(4) process the diffusion tensor data,

The two class coefficients that employing makes a variation based on the eigenvalue standard: relative anisotropy coefficient RA and Fractional anisotropy coefficient FA, algorithm is as follows:

$\mathrm{RA}=\frac{1}{\sqrt{2}}\frac{\sqrt{{({\mathrm{\λ}}_1-{\mathrm{\λ}}_2)}^2+{({\mathrm{\λ}}_2-{\mathrm{\λ}}_3)}^2+{({\mathrm{\λ}}_1-{\mathrm{\λ}}_3)}^2}}{({\mathrm{\λ}}_1+{\mathrm{\λ}}_2+{\mathrm{\λ}}_3)}$

$=\frac{\sqrt{3}}{\sqrt{2}}\frac{|D-\frac{1}{3}\mathrm{trace}\left(D\right)I|}{\mathrm{trace}\left(D\right)}$

$\mathrm{FA}=\frac{1}{\sqrt{2}}\frac{\sqrt{{({\mathrm{\λ}}_1-{\mathrm{\λ}}_2)}^2+{\left({\mathrm{\λ}}_2{-\mathrm{\λ}}_3\right)}^2+{({\mathrm{\λ}}_1-{\mathrm{\λ}}_3)}^2}}{\sqrt{{\mathrm{\λ}}_1^2+{\mathrm{\λ}}_2^2+{\mathrm{\λ}}_3^2}}$

$=\frac{\sqrt{3}}{\sqrt{2}}\frac{|D-\frac{1}{3}\mathrm{trace}\left(D\right)I|}{\left|D\right|}$

Wherein I is unit tensor, D is diffusion coefficient, three eigenvector λ, 1 λ, 2 λ 3 of dispersion tensor are mutually vertical, and the part that has made up each pixel is with reference to the fiber framework, in each voxel, characteristic value is arranged from big to small: the maximum dispersion coefficient of λ 1=, λ 2=middle rank dispersion coefficient, the minimum dispersion coefficient of λ 3=, λ 1 representative is parallel to the dispersion coefficient of machine direction, λ 2 and λ 3 represent lateral dispersion coefficient

Follow the trail of for white matter fiber, adopt tensor deviation algorithm to process the DTI data, as follows based on formula:

$v_o={\mathrm{\λ}}_1({\mathrm{\α}}_1{\hat{e}}_1+\frac{{\mathrm{\λ}}_2}{{\mathrm{\λ}}_1}{\mathrm{\α}}_2{\hat{e}}_2+\frac{{\mathrm{\λ}}_3}{{\mathrm{\λ}}_1}{\mathrm{\α}}_3{\hat{e}}_3)$

Wherein selecting the TEND algorithm is to consider the situation that the front patient's lesions position of Rhizoma Atractylodis Macrocephalae is had the physiology pathological changes, and the brain interested district that will choose is positioned at cerebral white matter and grey matter junction, from algorithm itself, adopt the TEND algorithm more to be conducive to track the fibre bundle that conforms to practical situation, selected parameter is as follows:

Processing parameter

The method according to this invention, also comprise (2-7) performance data result's Spatial normalization step, if it is tested to relate to one group or several groups, then need to carry out function result's space criteria fractional analysis, this process adopts the AFNI software of specialty to carry out, and may further comprise the steps:

At first, with the structure picture that obtains before and function as registration;

Secondly, structure picture (tested individual) is carried out the head clearing operation, gone to check behind the scalp, if go untotally, the head clearing that can try again operation is simultaneously by manually picking unnecessary zone;

At last, structure picture and the standard TT_N27 picture of head clearing carried out Spatial normalization, obtain simultaneously 12 correction parameters, i.e. the movement of three directions, rotation and Compression and Expansion value thereof.

The method according to this invention also comprises the functional localization of (2-8) language and the step that functional network is analyzed,

Functional image result to 8 tasks carries out the ANOVA analysis, calculates main effect and the reciprocal action of three factors (different input channels, different output task and different language complexity) that relate to; By the comparative analysis between task and conjoint analysis research brain different passage different task linguistic informations are separated functional character with integration process.

Utilize Independent Component Analysis that based on data drives that the original time series of 8 tasks is calculated separately functional network, such as input channel network, semantic network and output channel network etc., and analyze the similarities and differences feature of each network of each task.

The result who connects according to functional localization and function chooses region of interest, based on model-driven, to the coefficient of connection of original time series with each main brain district between effective connection analyzing method (structural equation model and dynamic causal model) computing network, determine closure, analyze the similarities and differences feature of each network of each task.

The analysis result that comparing function location, function connect (independent component analysis) and effectively connect is confirmed the relation in activation brain interval of each language task and the functional network feature in the un-activation brain district that may occur.

In the specific embodiment of the present invention, the method for utilizing Functional magnetic resonance imaging to carry out checking and examining in the speaking Chinese language district may further comprise the steps:

1) design relates to the language acknowledging task of a plurality of I/O channels and different language complexity, fMRI research by normal adult, accurately locate in the district to the speaking Chinese language brain, and definite functional network that is jointly formed by each function brain district, set up the data base of normal person's Chinese language processing brain domain, for clinical operation provides reference;

2) select the basic condition of classical language radioimmunoassay patient language loss, and select suitable language task according to checking and examining the result, choose suitable task with Rhizoma Atractylodis Macrocephalae before the language function district of (being generally the last week of performing the operation) check and examine;

3) for difference damage or diseased region, according to step 2) check and examine the result, with reference to step 1) normal person's Chinese language processing brain domain data base, scanning obtains patient's function result;

4) for difference damage or diseased region, Chinese language processing brain domain data base with reference to the step 1) normal person, and integrating step 2) chooses the analysis result of task in, choosing suitable task monitors with the language function district in the Rhizoma Atractylodis Macrocephalae, under the prerequisite that preliminary operation plan has been determined, open behind the cranium further affirmation scheme, on the basis of respecting operative Principle, to keep to greatest extent language function.

Description of drawings

Fig. 1 is the data analysis flow chart.

Fig. 2 is data analysis flow process detail drawing.

The specific embodiment

For understanding the crucial brain district of different language processing channel in depth, set up the data base of normal person's Chinese language processing brain domain, according to the language processing model of generally acknowledging in the world at present, carry out cerebral function imaging research.Finish different sensation input channels by the normal person, the fRMI experiment of the language task of different output channels, obtain the normal person and carried out the judgement of Chinese audition lexical semantic, the visual pattern Semantic judgement, the visual vocabulary Semantic judgement, audition vocabulary is repeated, the visual pattern name, the visual vocabulary name, audition Chinese sentence Semantic judgement, fMRI data during audition Chinese sentence reading task and corresponding behavioral data, and adopt brain function data analysis software and statistical software to carry out data analysis, obtain from Chinese different sensation input channels (vision and audition), the differing complexity language (comprises vocabulary, picture and statement) understanding and language produce relevant brain district, and in the different language processing tasks, different sensation input channels, the similarities and differences that the semantic Advanced Machining System Midbrain Area of voice activates, the language function district data base who has set up the above-mentioned Chinese language tasks of Chinese.

Wherein, linguistic data uses cover patients with brain injury language testing material (Snodgrass, 1980 commonly used; Shu Hua etc., 1989), and for to be applied to checking and examining of patient, the noun of all selecting to be familiar with in the daily life (animal is such as tiger, or instrument, such as umbrella) and sentence commonly used (as, N/A will be held at next spring).The tile designs of experiment employing standard, task is sequentially carried out balance, utilizes the MRI instrument, obtains simultaneously blood oxygen level functional imaging (BOLD-fMRI) tranquillization attitude, task attitude data and diffusion tensor (DTI) data.Tested task comprises:

1). picture name: vision presents a width of cloth figure (such as, tiger), requires tested speaking to read.

2). visual vocabulary is read: vision presents a Chinese two-character word (such as, tiger), requires tested reading aloud.

3). audition vocabulary is repeated: audition presents a Chinese two-character word (such as, tiger), requires the tested repetition of speaking.

4). the figure Semantic judgement: vision present a width of cloth figure (as, lion, rabbit), require the tested whether dangerous property of this thing of judging, the button reaction.

5). the visual vocabulary Semantic judgement: vision present a Chinese two-character word (as, tiger, sheep), require the tested whether dangerous property of this thing of judging, the button reaction.

6). the audition lexical semantic is judged: audition present a Chinese two-character word (as, pistol, umbrella), require tested judge whether dangerous property of this thing, button reaction.

7), the audition Chinese sentence is by interesting to listen to: audition presents a Chinese sentence, requires the tested button reaction of doing after hearing out sentence.

8), the audition Chinese sentence understands: audition present a Chinese sentence (as, N/A will be held at next spring; Or, will hold in the meeting in spring next year next time), require testedly to judge whether this sentence is normal sentence.

The data scanning parameter of selecting is as follows:

Use 3.0T superconduct magnetic resonance scanner (Siemens Trio, Germany) image data, the structure picture uses three-dimensional magnetization to prepare fast gradient echo (magnetization prepared rapid acquisition gradient echo, MPRAGE), obtain continuously sagittal plain T1 weighting picture (T1weight imaging, T1WI), be used for three-dimensional reconstruction and spatial registration, sweep parameter: TR=1900ms, TE=2.2ms, FA=9 ° of reversing time (inversion time, TI)=and 900ms, visual field 240mm * 240mm, matrix 256 * 256, bed thickness 1mm, continuously every.Tranquillization attitude and task attitude function are used gtadient echo-echo-planar imaging (gradient echo-echo planar imaging) sequence, sweep parameter: TR=2000ms, TE=40ms as fMRI, FA=90 °, visual field 240mm * 240mm, matrix 64 * 64, bed thickness 4mm, interlayer is every 1mm.DTI sweep parameter: TR=11000ms, TE=98ms, visual field 128mm * 128mm, matrix 256 * 256, bed thickness 2mm, continuously every, 30 directions, b=0 and 1000, sweep limits to calvarium section, covers full brain from Foramen magnum.

Select suitable language task to check and examine and analyze on the result in the language function district that obtains patient according to different diseased regions, main brain district according to excising or damage determines that suitable task is with monitoring in the Rhizoma Atractylodis Macrocephalae.Simultaneously, because of the timeliness of monitoring in the art, according to diseased region and radioimmunoassay result, task in the art is simplified, the time overall length was controlled within 5 minutes, and sequence is as follows:

1) involve Broca district (left hemisphere pars opercularis gyri frontalis inferioris and triangular part) pathological changes, select to comprise picture name, visual vocabulary are read, the audition lexical semantic is judged and the audition Chinese sentence by interesting to listen to comprehensive task;

2) involve Wernicke district (left hemisphere superior temporal gyrus middle temporal gyrus rear portion) pathological changes, select to comprise figure Semantic judgement, the comprehensive task that audition vocabulary is repeated, the audition lexical semantic is judged and the audition Chinese sentence is understood;

3) involve left hemisphere temporo top associated cortex pathological changes, select to comprise figure name, visual vocabulary are read, audition vocabulary is repeated and the audition Chinese sentence by interesting to listen to task;

4) involve left hemisphere middle temporal gyrus rear portion pathological changes, select to comprise that figure name, visual vocabulary semantic understanding, audition lexical semantic are understood and the comprehensive task of audition Chinese sentence understanding;

5) involve postmedian pathological changes at the bottom of the left hemisphere temporo, select to comprise that visual vocabulary is read, the comprehensive task of figure name and graphic semantic understanding;

6) involve the anterior pathological changes of left hemisphere temporo, select to comprise that audition vocabulary phonetic rehearsal, audition lexical semantic are understood, the audition Chinese sentence is by comprehensive task interesting to listen to and that the audition Chinese sentence is understood;

7) involve transverse temporal gyri and superior temporal gyrus middle part (one-sided or bilateral) pathological changes, select to comprise audition vocabulary is repeated and the audition Chinese sentence by interesting to listen to task;

8) if checking and examining display language before the art understands damage is arranged, select to comprise figure Semantic judgement, visual vocabulary semantic understanding, the task that the audition lexical semantic is judged and the audition Chinese sentence is understood;

9) if checking and examining display language before the art reads damage is arranged, the comprehensive task of select to comprise that visual vocabulary is read, figure name, visual vocabulary semantic understanding and audition Chinese sentence being understood;

10) if checking and examining display language before the art expresses damage is arranged, select to comprise that audition vocabulary is repeated, visual vocabulary is read, figure name and audition Chinese sentence be by interesting to listen to comprehensive task.

Wherein, the processing method checked and examined with fMRI data of speaking Chinese language district may further comprise the steps:

(1) blood oxygen level dependence fMRI data and the diffusion tensor data of object are checked and examined in acquisition;

(2) blood oxygen level of Processing tasks attitude relies on fMRI data

(2-1) three-dimensional reconstruction of structure picture and function picture carries out the Fourier conversion with the performance data of the K space form preserved in the scanner;

(2-2) singular data is processed

Front 4 performance datas that removal collects are to avoid the magnetic field de-stabilising effect at the initial stage of scanning;

(2-3) time smoothing

Use blood oxygen level to rely on displacement information between aspect in the fMRI data, the voxel time series of input data set is shifted, make each aspect snap to the same time initial point;

(2-4) dynamic(al) correction and space smoothing

Take last seasonal effect in time series brain position as benchmark, other all seasonal effect in time series brains are with it alignment all, each time point obtains respectively the changing value of 6 directions like this, be moved beyond 1 voxel or rotate surpass 5 ° tested data and all reject, 6 column data are saved as file, carry out the space smoothing processing with getting data after the dynamic(al) correction, overall with half high value is set as 6;

(2-5) seasonal effect in time series standardization

With the performance data behind the space smoothing, doing head clearing processes, obtain the mask file, carry out the average of mask for a plurality of speaking Chinese language test assignments, be mask for a plurality of tested and tested Multiple-Scans again and merge processing, described merging processing mode is the time series of baseline with data normalization to 100 for the performance data in the mask is got the average analysis at last;

(2-6) regression analysis of performance data

Adopt top hat function, referring to following formula:

H(t)=∫ ₀ ^min(t，d)h(t-s)dt；

h(t)=(t/4) ⁴e ^4-t

Wherein, t is the independent variable of time point, and s is certain set time point, and d is stimulus duration, and h(t) representative is the intensity level of in time t variation, H(t) then represents time dependent integrated value,

Top hat function and stimulation sequence task time are carried out convolutional calculation, thus the shock response model of the task of generation, and specific algorithm is as follows:

$y\left(t\right)={\∫}_0^{t}f\left(T\right)h(t-T)\mathrm{dT}$

Wherein, t is the independent variable time point, and the time series of f (T) for stimulating h(t-T) is top hat function, y(t) is time dependent convolution value.

The blood oxygen level that analysis is obtained relies on fMRI data and model obtained above is done regression analysis, namely obtains the β value of complete each voxel of brain, and this numerical value can characterize this voxel and get dependency relation between time series and the model;

(3) blood oxygen level of processing tranquillization attitude relies on fMRI data

The pretreatment process with above-mentioned (2-1) to (2-5), calculate on this basis tranquillization attitude low-frequency oscillation amplitude, time series is converted into the frequency time domain by fast Fourier transform, calculate the square root of crest height, then the square root of each voxel in 0.01～0.08Hz scope averaged, square root after average is tranquillization attitude low-frequency oscillation amplitude A LFF, analyze ALFF value change situation, obtain corresponding Statistical Parametric Mapping, time series in the fMRI interesting image district is averaged, obtain reference time array, in the mode of voxel-voxel different reference time arrays are carried out correlation analysis, and use the Fisher conversion to improve the normality of correlation coefficient, the correlation coefficient that the relatively difference of the bonding strength between each crucial brain district, and measurement Different brain region connects;

(4) process the diffusion tensor data,

The two class coefficients that employing makes a variation based on the eigenvalue standard: relative anisotropy coefficient RA and Fractional anisotropy coefficient FA, algorithm is as follows:

$\mathrm{RA}=\frac{1}{\sqrt{2}}\frac{\sqrt{{({\mathrm{\λ}}_1-{\mathrm{\λ}}_2)}^2+{({\mathrm{\λ}}_2-{\mathrm{\λ}}_3)}^2+{({\mathrm{\λ}}_1-{\mathrm{\λ}}_3)}^2}}{({\mathrm{\λ}}_1+{\mathrm{\λ}}_2+{\mathrm{\λ}}_3)}$

$=\frac{\sqrt{3}}{\sqrt{2}}\frac{|D-\frac{1}{3}\mathrm{trace}\left(D\right)I|}{\mathrm{trace}\left(D\right)}$

$\mathrm{FA}=\frac{1}{\sqrt{2}}\frac{\sqrt{{({\mathrm{\λ}}_1-{\mathrm{\λ}}_2)}^2+{\left({\mathrm{\λ}}_2{-\mathrm{\λ}}_3\right)}^2+{({\mathrm{\λ}}_1-{\mathrm{\λ}}_3)}^2}}{\sqrt{{\mathrm{\λ}}_1^2+{\mathrm{\λ}}_2^2+{\mathrm{\λ}}_3^2}}$

$=\frac{\sqrt{3}}{\sqrt{2}}\frac{|D-\frac{1}{3}\mathrm{trace}\left(D\right)I|}{\left|D\right|}$

Wherein I is unit tensor, D is diffusion coefficient, three eigenvector λ, 1 λ, 2 λ 3 of dispersion tensor are mutually vertical, and the part that has made up each pixel is with reference to the fiber framework, in each voxel, characteristic value is arranged from big to small: the maximum dispersion coefficient of λ 1=, λ 2=middle rank dispersion coefficient, the minimum dispersion coefficient of λ 3=, λ 1 representative is parallel to the dispersion coefficient of machine direction, λ 2 and λ 3 represent lateral dispersion coefficient

Follow the trail of for white matter fiber, adopt tensor deviation algorithm to process the DTI data, as follows based on formula:

$v_o={\mathrm{\λ}}_1({\mathrm{\α}}_1{\hat{e}}_1+\frac{{\mathrm{\λ}}_2}{{\mathrm{\λ}}_1}{\mathrm{\α}}_2{\hat{e}}_2+\frac{{\mathrm{\λ}}_3}{{\mathrm{\λ}}_1}{\mathrm{\α}}_3{\hat{e}}_3)$

Wherein selecting the TEND algorithm is to consider the situation that the front patient's lesions position of Rhizoma Atractylodis Macrocephalae is had the physiology pathological changes, and the brain interested district that will choose is positioned at cerebral white matter and grey matter junction, from algorithm itself, adopt the TEND algorithm more to be conducive to track the fibre bundle that conforms to practical situation, selected parameter is as follows:

Processing parameter

Wherein, " blood oxygen level of Processing tasks attitude relies on fMRI data " also comprises step: (2-7) performance data result's Spatial normalization

Above-mentioned steps is for ontoanalysis, if it is tested to relate to one group or several groups, then needs to carry out function result's space criteria fractional analysis, and this process adopts the AFNI software of specialty to carry out, and may further comprise the steps:

At first, with the structure picture that obtains before and function as registration;

Secondly, structure picture (tested individual) is carried out the head clearing operation, gone to check behind the scalp, if go untotally, the head clearing that can try again operation is simultaneously by manually picking unnecessary zone;

At last, structure picture and the standard TT_N27 picture of head clearing carried out Spatial normalization, obtain simultaneously 12 correction parameters, i.e. the movement of three directions, rotation and Compression and Expansion value thereof.

(2-8) functional localization of language and functional network analysis

Functional image result to 8 tasks carries out the ANOVA analysis, calculates main effect and the reciprocal action of three factors (different input channels, different output task and different language complexity) that relate to; By the comparative analysis between task and conjoint analysis research brain different passage different task linguistic informations are separated functional character with integration process.

Utilize Independent Component Analysis that based on data drives that the original time series of 8 tasks is calculated separately functional network, such as input channel network, semantic network and output channel network etc., and analyze the similarities and differences feature of each network of each task.

The result who connects according to functional localization and function chooses region of interest, based on model-driven, to the coefficient of connection of original time series with each main brain district between effective connection analyzing method (structural equation model and dynamic causal model) computing network, determine closure, analyze the similarities and differences feature of each network of each task.

The analysis result that comparing function location, function connect (independent component analysis) and effectively connect is confirmed the relation in activation brain interval of each language task and the functional network feature in the un-activation brain district that may occur.

Above specific embodiment only is used for technical scheme of the present invention is described and unrestricted; although describe the present invention in detail with reference to above-described embodiment; those of ordinary skill in the art is to be understood that; technical scheme of the present invention is made amendment or is equal to replacement; the essence and the protection domain that do not break away from technical scheme of the present invention, it all should be encompassed in the claim scope of the present invention.

Claims (2)

1. a speaking Chinese language district checks and examine the processing method of using fMRI data, it is characterized in that, said method comprising the steps of:

(1) blood oxygen level dependence fMRI data and the diffusion tensor data of object are checked and examined in acquisition;

(2) blood oxygen level of Processing tasks attitude relies on fMRI data

(2-1) three-dimensional reconstruction of structure picture and function picture carries out the Fourier conversion with the performance data of the K space form preserved in the scanner;

(2-2) singular data is processed

Front 4 performance datas that removal collects are to avoid the magnetic field de-stabilising effect at the initial stage of scanning;

(2-3) time smoothing

Use blood oxygen level to rely on displacement information between aspect in the fMRI data, the voxel time series of input data set is shifted, make each aspect snap to the same time initial point;

(2-4) dynamic(al) correction and space smoothing

Take last seasonal effect in time series brain position as benchmark, other all seasonal effect in time series brains are with it alignment all, each time point obtains respectively the changing value of 6 directions like this, be moved beyond 1 voxel or rotate surpass 5 ° tested data and all reject, 6 column data are saved as file, carry out the space smoothing processing with getting data after the dynamic(al) correction, overall with half high value is set as 6;

(2-5) seasonal effect in time series standardization

With the performance data behind the space smoothing, doing head clearing processes, obtain the mask file, carry out the average of mask for a plurality of speaking Chinese language test assignments, be mask for a plurality of tested and tested Multiple-Scans again and merge processing, described merging processing mode is the time series of baseline with data normalization to 100 for the performance data in the mask is got the average analysis at last;

(2-6) regression analysis of performance data

Adopt top hat function, referring to following formula:

H(t)=∫ ₀ ^min(t，d)h(t-s)dt；

h(t)=(t/4) ⁴e ^4-t

Wherein, t is the independent variable of time point, and s is certain set time point, and d is stimulus duration, and h(t) representative is the intensity level of in time t variation, H(t) then represents time dependent integrated value,

Top hat function and stimulation sequence task time are carried out convolutional calculation, thus the shock response model of the task of generation, and specific algorithm is as follows:

$y\left(t\right)={\∫}_0^{t}f\left(T\right)h(t-T)\mathrm{dT}$

Wherein, t is the independent variable time point, and the time series of f (T) for stimulating h(t-T) is top hat function, y(t) is time dependent convolution value.

The blood oxygen level that analysis is obtained relies on fMRI data and model obtained above is done regression analysis, namely obtains the β value of complete each voxel of brain, and this numerical value can characterize this voxel and get dependency relation between time series and the model;

(3) blood oxygen level of processing tranquillization attitude relies on fMRI data

The pretreatment process with above-mentioned (2-1) to (2-5), calculate on this basis tranquillization attitude low-frequency oscillation amplitude, time series is converted into the frequency time domain by fast Fourier transform, calculate the square root of crest height, then the square root of each voxel in 0.01～0.08Hz scope averaged, square root after average is tranquillization attitude low-frequency oscillation amplitude A LFF, analyze ALFF value change situation, obtain corresponding Statistical Parametric Mapping, time series in the fMRI interesting image district is averaged, obtain reference time array, in the mode of voxel-voxel different reference time arrays are carried out correlation analysis, and use the Fisher conversion to improve the normality of correlation coefficient, the correlation coefficient that the relatively difference of the bonding strength between each crucial brain district, and measurement Different brain region connects;

(4) process the diffusion tensor data,

The two class coefficients that employing makes a variation based on the eigenvalue standard: relative anisotropy coefficient RA and Fractional anisotropy coefficient FA, algorithm is as follows:

$\mathrm{RA}=\frac{1}{\sqrt{2}}\frac{\sqrt{{({\mathrm{\λ}}_1-{\mathrm{\λ}}_2)}^2+{({\mathrm{\λ}}_2-{\mathrm{\λ}}_3)}^2+{({\mathrm{\λ}}_1-{\mathrm{\λ}}_3)}^2}}{({\mathrm{\λ}}_1+{\mathrm{\λ}}_2+{\mathrm{\λ}}_3)}$

$=\frac{\sqrt{3}}{\sqrt{2}}\frac{|D-\frac{1}{3}\mathrm{trace}\left(D\right)I|}{\mathrm{trace}\left(D\right)}$

$\mathrm{FA}=\frac{1}{\sqrt{2}}\frac{\sqrt{{({\mathrm{\λ}}_1-{\mathrm{\λ}}_2)}^2+{\left({\mathrm{\λ}}_2{-\mathrm{\λ}}_3\right)}^2+{({\mathrm{\λ}}_1-{\mathrm{\λ}}_3)}^2}}{\sqrt{{\mathrm{\λ}}_1^2+{\mathrm{\λ}}_2^2+{\mathrm{\λ}}_3^2}}$

$=\frac{\sqrt{3}}{\sqrt{2}}\frac{|D-\frac{1}{3}\mathrm{trace}\left(D\right)I|}{\left|D\right|}$

Wherein I is unit tensor, D is diffusion coefficient, three eigenvector λ, 1 λ, 2 λ 3 of dispersion tensor are mutually vertical, and the part that has made up each pixel is with reference to the fiber framework, and in each voxel, characteristic value is arranged from big to small: the maximum dispersion coefficient of λ 1=, λ 2=middle rank dispersion coefficient, the minimum dispersion coefficient of λ 3=, λ 1 representative is parallel to the dispersion coefficient of machine direction, and λ 2 and λ 3 represent lateral dispersion coefficient.

2. speaking Chinese language according to claim 1 district checks and examine the processing method with fMRI data, it is characterized in that described method comprises the step of (2-7) performance data result's Spatial normalization:

Tested for one group or several groups, carry out function result's space criteria fractional analysis, may further comprise the steps:

At first, with the structure picture that obtains before and function as registration;

Secondly, structure is looked like to carry out the head clearing operation, gone to check behind the scalp, if go untotally, the head clearing that can try again operation is simultaneously by manually picking unnecessary zone;

At last, structure picture and the standard TT_N27 picture of head clearing carried out Spatial normalization, obtain simultaneously 12 correction parameters, i.e. the movement of three directions, rotation and Compression and Expansion value thereof.

Images