[summary of the invention]
Based on this, be necessary to provide a kind of formation method of dispersion tensor fast.
A kind of dispersion tensor formation method may further comprise the steps: the calculating probability density function; Carry out the variable density sampling according to said probability density function and obtain not having disperse figure and the pairing data of a plurality of disperse figure; Said no disperse figure and the pairing data of a plurality of disperse figure are rebuild obtain reconstructed image.
Further, also comprise before the step of said calculating probability density function: sampling in advance obtains preparatory sampled data; The step of said calculating probability density function is: respectively said preparatory sampled data is calculated probability density function.
Further, said preparatory sampled data comprises no disperse figure and the corresponding preparatory sampled data of a plurality of disperse figure; The said step that respectively said preparatory sampled data is calculated probability density function is specially: respectively said no disperse figure and the corresponding preparatory sampled data of a plurality of disperse figure are calculated no disperse figure and the corresponding probability density function of a plurality of disperse figure.
Further, saidly also comprise after respectively said preparatory sampled data being calculated the step of probability density function: said no disperse figure and the corresponding probability density function of a plurality of disperse figure are carried out match obtain the first unified probability density function; Saidly carry out the step that variable density sampling obtains not having disperse figure and the pairing data of a plurality of disperse figure according to said probability density function and be specially: carry out the variable density sampling according to the said first unified probability density function, obtain and said no disperse figure and the corresponding data of a plurality of disperse figure.
Further, saidly also comprise after respectively said preparatory sampled data being calculated the step of probability density function: the probability density function corresponding to a plurality of disperse figure carries out match, obtains the step of the second unified probability density function; Saidly carry out the step that variable density sampling obtains not having disperse figure and the pairing data of a plurality of disperse figure according to said probability density function and be specially: carry out variable density according to the probability density function of no disperse figure and sample, obtain not having the data of disperse figure; Carry out the variable density sampling according to the said second unified probability density function, obtain the pairing data of a plurality of disperse figure.
Further, said preparatory sampled data comprises the preparatory sampled data of no disperse figure or any disperse figure; The said step that respectively said preparatory sampled data is calculated probability density function is specially: the preparatory sampled data to said no disperse figure or any disperse figure calculates no disperse figure or the corresponding probability density function of any disperse figure; Saidly carry out the step that variable density sampling obtains not having disperse diagram data and the pairing data of a plurality of disperse figure according to said probability density function and be specially: carry out the variable density sampling according to said no disperse figure or the corresponding probability density function of any disperse figure, obtain not having disperse diagram data and the pairing data of a plurality of disperse figure.
Further, said preparatory sampled data comprises the preparatory sampled data of no disperse figure and any disperse figure; The said step that respectively said preparatory sampled data is calculated probability density function is specially: the preparatory sampled data to said no disperse figure and any disperse figure calculates no disperse figure and the corresponding probability density function of any disperse figure; Saidly carry out the step that variable density sampling obtains not having disperse diagram data and the pairing data of a plurality of disperse figure according to said probability density function and be specially: the probability density function according to said no disperse figure carries out the variable density sampling, obtains not having the data of disperse figure; Carry out the variable density sampling according to the corresponding probability density function of said arbitrary disperse figure, obtain the data of a plurality of disperse figure.
In addition, also be necessary to provide a kind of imaging system of dispersion tensor fast.
A kind of dispersion tensor imaging system is characterized in that, comprises processing module, sampling module and rebuilding module; Processing module is used for the calculating probability density function; Sampling module is electrically connected with said processing module, and said sampling module is used for carrying out the variable density sampling according to said probability density function and obtains not having disperse diagram data and disperse diagram data; Rebuilding module is electrically connected with said sampling module, is used for said no disperse figure and the pairing data of a plurality of disperse figure are rebuild obtaining reconstructed image.
Further, said sampling module also is used to sample in advance and obtains preparatory sampled data.
In above-mentioned dispersion tensor formation method and the system; Through calculating probability density function, and the data of no disperse figure and each disperse figure are carried out the variable density sampling, significantly reduced the data of required sampling according to probability density function; Make acquisition time significantly reduce, realize quick dispersion tensor imaging.
[specific embodiment]
In order to solve the problem of image data overlong time total in traditional dispersion tensor imaging process, proposed the time that a kind of formation method of dispersion tensor fast shortens whole dispersion tensor imaging process.
CALCULATION OF PARAMETERS needs a no disperse figure and a plurality of disperse figure that has applied the disperse gradient to come compute tensor D in the dispersion tensor imaging (DTI), thereby carries out image reconstruction.D is the symmetrical matrix of a 3*3, so want to solve D, need apply six not disperse gradients of conllinear at least, promptly wants to calculate D, needs six disperse figure at least.Eigenvalue and characteristic vector through separating symmetrical matrix D just can obtain analyzing some parameters commonly used that dispersion tensor forms images; Like fractional anisotropy (fractional anisotropy; FA); Relatively anisotropy (relative anisotropy, RA) and apparent diffusion coefficient (apparent diffusion coefficient, ADC) etc.
See also Fig. 1, a kind of dispersion tensor formation method may further comprise the steps:
Step S10, the calculating probability density function.
Probability density function is used for confirming sampling density, with the control sampling process.In the sampling process of reality; Only partial data is sampled and to reduce consumed time effectively; And the information spinner of image will concentrate on low frequency region, and the contained quantity of information of high-frequency region is considerably less, therefore can sample through the mode of the employing variable density sampling controlled by probability density function; So both consumed time in the sampling process can be reduced, the generation of the pseudo-shadow of aliasing can also be reduced effectively.
Particularly, probability density function comprises the probability density function and the adaptive probability density function of non-self-adapting.Do not having to construct the probability density function of non-self-adapting, for example gauss of distribution function under the situation of prior information.The probability density function construction process of non-self-adapting is comparatively simple, can be applicable to the lower dispersion tensor imaging process of degree of accuracy.In a preferred embodiment, probability density function is adaptive probability density function, and adaptive probability density function is to construct according to the prior information that is looked like to form by a picture group, and its sample effect obviously is better than the probability density function of non-self-adapting.
Owing to more obtain better sample effect according to adaptive probability density function, in one embodiment, also comprise sampling in advance before the step S10 obtaining the step of preparatory sampled data.Simultaneously, step S10 is specially: respectively preparatory sampled data is calculated probability density function.Through sampling in advance, obtain prior information, make up adaptive probability density function.
Carry out a little square through the corresponding sampled data in advance of image institute in preparatory sampling institute on each row of phase code y direction, and add with, again through normalization, finally produce the corresponding probability density function of image.According to each image, make up corresponding adaptive probability density function, can obtain better sample effect.
Step S20 carries out the variable density sampling according to probability density function and obtains not having disperse figure and the pairing data of a plurality of disperse figure.
Because the energy of image mainly concentrates on low frequency region; The contained quantity of information of high-frequency region seldom; We just can adopt the mode of variable density sampling like this, mainly adopt low frequency signal, and the signal of high frequency region is adopted less or do not adopted as far as possible; So just can reduce the pseudo-shadow of aliasing effectively, improve the speed of sampling.According to adaptive probability density function the data of no disperse figure and a plurality of disperse gradient map are carried out the variable density sampling, only partial data is sampled, can significantly reduce sweep time.
Step S30 rebuilds no disperse figure and the pairing data of a plurality of disperse gradient map and to obtain reconstructed image.
The data of each image are carried out compressed sensing, and (Compressed Sensing CS) rebuilds and total variation is handled, and obtains reconstructed image.
In the above-mentioned dispersion tensor formation method; Through calculating probability density function, and the data of no disperse figure and each disperse figure are carried out the variable density sampling, significantly reduced the data of required sampling according to probability density function; Make acquisition time significantly reduce, realize quick dispersion tensor imaging.
As shown in Figure 2, B
0Be the probability density function of the no disperse figure that calculates according to real image, B
1~B
6Probability density function for each disperse figure of calculating according to real image.Can find out that by Fig. 2 each disperse figure and the correlation coefficient that does not have between the disperse figure approach 1.Therefore, as shown in Figure 3 in another embodiment in order further to reduce the time of whole dispersion tensor imaging process, sampled data is no disperse figure and the corresponding preparatory sampled data of a plurality of disperse figure in advance, and above-mentioned dispersion tensor formation method comprises the steps:
Step S301, sampling in advance obtains not having disperse figure and the corresponding preparatory sampled data of a plurality of disperse figure.
Step S303 calculates no disperse figure and the corresponding probability density function of a plurality of disperse figure to no disperse figure and the corresponding preparatory sampled data of a plurality of disperse figure respectively.
Step S305 carries out match to the corresponding probability density function of no disperse figure and a plurality of disperse figure and obtains the first unified probability density function.MeanB among Fig. 2
0-B
6The first unified probability density function that simulates for all probability density functions.
Step S307 carries out the variable density sampling according to the first unified probability density function, obtains and does not have disperse figure and the corresponding data of a plurality of disperse figure.
Probability density function through no disperse figure and a plurality of disperse figure; Simulate first a unified probability density function, and come the data of no disperse figure and a plurality of disperse figure are carried out the variable density sampling according to this first unified probability density function applicable to having or not disperse figure and a plurality of disperse figure.Based on the dependency between each image in the dispersion tensor imaging, adopt same probability density function that each image is carried out the variable density sampling, further saved acquisition time.
Step S309 rebuilds no disperse figure and the pairing data of a plurality of disperse figure and to obtain reconstructed image.
It is pointed out that through Fig. 3 and can find out that the probability density function of no disperse figure wants big with the diversity of the probability density function of a plurality of disperse figure than the diversity of the probability density function between each disperse figure.
In order to improve the precision of image, as shown in Figure 4 in another embodiment, step S305 and step S307 be replaceable to become following steps:
Step S401, the probability density function corresponding to a plurality of disperse figure carries out match, obtains the second unified probability density function.
Step S403 carries out the variable density sampling according to the probability density function of no disperse figure, obtains not having the data of disperse figure.
Step S405 carries out the variable density sampling according to the second unified probability density function, obtains the pairing data of a plurality of disperse figure.
See also Fig. 5, MeanB
1-B
6The second unified probability density function that simulates for the probability density function of all disperse figure.Above step is utilized the dependency between a plurality of disperse figure, makes up one second unified probability density function, comes the pairing data of disperse figure are sampled, and makes sampled result more accurate.
In another embodiment, as shown in Figure 6, sampled data is the preparatory sampled data of no disperse figure or any disperse figure in advance, and above-mentioned dispersion tensor formation method comprises the steps:
Step S601, sampling in advance obtains not having the preparatory sampled data of disperse figure or any disperse figure.
Step S603 calculates no disperse figure or the corresponding probability density function of any disperse figure to the preparatory sampled data of no disperse figure or any disperse figure.
Step S605 carries out the variable density sampling according to no disperse figure or the corresponding probability density function of any disperse figure, obtains not having disperse diagram data and the pairing data of a plurality of disperse figure.
Step S607 rebuilds no disperse figure and the pairing data of a plurality of disperse figure and to obtain reconstructed image.
Because each disperse figure and the correlation coefficient that does not have between the disperse figure approach 1; Here any one probability density function comes no disperse figure and a plurality of disperse figure are carried out the variable density sampling among the no disperse figure of employing or a plurality of disperse figure; Need not calculate the probability density function of other image, when guaranteeing certain accuracy, further reduce acquisition time.
In another embodiment, as shown in Figure 7 in order further to improve the speed of image data, sampled data is the preparatory sampled data of no disperse figure and any disperse figure in advance, and above-mentioned dispersion tensor formation method comprises the steps:
Step S701, sampling in advance obtains not having the preparatory sampled data of disperse figure and any disperse figure.
Step S703 calculates no disperse figure and the corresponding probability density function of any disperse figure to the preparatory sampled data of no disperse figure and any disperse figure.
Step S705 carries out the variable density sampling according to the probability density function of no disperse figure, obtains not having the data of disperse figure.
Step S707 carries out the variable density sampling according to the corresponding probability density function of arbitrary disperse figure, obtains the data of a plurality of disperse figure.
Step S709 rebuilds no disperse figure and the pairing data of a plurality of disperse figure and to obtain reconstructed image.
The probability density function of no disperse figure wants big with the diversity between the probability density function of a plurality of disperse figure than the diversity of the probability density function between each disperse figure, and the difference between the probability density function between each disperse figure is very little.Probability density function according to no disperse figure comes the data of no disperse figure are carried out the variable density sampling in the present embodiment; Probability density function according to arbitrary disperse figure comes the data of each disperse figure are carried out the variable density sampling, makes that the result of data acquisition is more accurate.
See also Fig. 8, a kind of dispersion tensor imaging system also is provided, comprise processing module 100, sampling module 200 and rebuilding module 300.
Processing module 100 is used for the calculating probability density function.Through sampling in advance, obtain prior information, make up adaptive probability density function; And carry out a little square through the corresponding sampled data in advance of image institute in preparatory sampling institute on each row of phase code y direction; And add with, again through normalization, finally produce the corresponding probability density function of image.According to each image, make up corresponding adaptive probability density function, can obtain better sample effect.It is pointed out that is not having under the situation of prior information, and processing module 100 also can be set up the probability density function of non-self-adapting, like gauss of distribution function.Specifically in the dispersion tensor imaging process, processing module 100 is used to calculate no disperse figure and the corresponding probability density function of a plurality of disperse figure.
Sampling module 200 is electrically connected with processing module 100, and sampling module 200 is used for carrying out the variable density sampling according to said probability density function and obtains not having disperse diagram data and disperse diagram data.Sampling module 200 also is used to sample in advance, obtains preparatory sampled data, and sampled data is used to supply processing module 100 calculating probability density functions in advance.According to adaptive probability density function no disperse figure and the pairing data of a plurality of disperse figure are carried out the variable density sampling, only partial data is sampled, can significantly reduce sweep time.
Rebuilding module 300 is electrically connected with sampling module 200, is used for said no disperse figure and the pairing data of a plurality of disperse figure are rebuild obtaining reconstructed image.The data of each image are carried out compressed sensing, and (Compressed Sensing CS) rebuilds and total variation is handled, and obtains reconstructed image.
In the above-mentioned dispersion tensor imaging system; Through calculating probability density function; And no disperse figure and the pairing data of a plurality of disperse figure are carried out the variable density sampling according to probability density function; Significantly reduce the data of required sampling, made acquisition time significantly reduce, realized quick dispersion tensor imaging.
It is to be noted; Because the dependency between each image; Processing module 100 can also be used for that the corresponding probability density function of a plurality of images is carried out match and obtain a probability density function; Only to use a probability density function that each image is sampled, make the speed of sampling further improve.
Simultaneously, processing module 100 also can only calculate the probability density function of any figure, and through this probability density function the data of no disperse figure and a plurality of disperse figure is carried out the variable density sampling.Because the dependency between each image is sampled to a plurality of images through same probability density function, when having guaranteed certain accuracy, has saved acquisition time.
The above embodiment has only expressed several kinds of embodiments of the present invention, and it describes comparatively concrete and detailed, but can not therefore be interpreted as the restriction to claim of the present invention.Should be pointed out that for the person of ordinary skill of the art under the prerequisite that does not break away from the present invention's design, can also make some distortion and improvement, these all belong to protection scope of the present invention.Therefore, the protection domain of patent of the present invention should be as the criterion with accompanying claims.