TWI345461B - Q-space sampling method and diffusion spectrum imaging method employing the same - Google Patents

Description

1345461 IX. Description of the Invention: [Technical Field] The present invention relates to a magnetic vibration technique, and more particularly to a Q-space sampling method and a diffusion spectrum contrast method using the same. [Prior Art] Diffusion spectrum imaging (DSI) is an extended application of magnetic resonance (MR) technology, which can non-invasively obtain information on tissue fibers.

The conventional DSI technology mainly includes the following steps: (a) First stage sampling: using a series of radio frequency (RF) pulses and magnetic field gradients, the spatial information of the organization can be distinguished by a spatial coding procedure. This first stage sampling is considered Sampling of the structural image. (b) Second stage sampling: In addition to spatial coding, a diffusion-weighted magnetic field gradient (hereinafter referred to as a diffusion gradient) is also applied. Since information on the movement of water molecules can be added to the imaged tissue at this stage, the applied diffusion gradient is considered to be a diffusion code. Under these two stages of coding, structural images can be enhanced by diffusion-weighted contrast, and such images are called diffusion weighted images (DWI). The contrast of DWIs will be different under different diffusion gradients. This second stage sampling is also referred to as Q space sampling. (c) Signal processing: Since the diffusion gradients of different directions and intensities represent different coordinates in the three-dimensional Q space, different DWIs with corresponding spatial positions are arranged in a Q space (corresponding to the Q-space position of the original design). After arranging all the DWIs in the Q space, the inverse Fourier transform is performed on the sampling data of these Q spaces to reconstruct the probability density function (PDF) of the water molecule motion, and further the probability density function is performed. The orientation distribution function is obtained for the radial integration, and the information of the tissue fibers can be obtained by analyzing these direction distribution functions. The big drawback of DSI technology is that the sampling speed is too slow. According to the literature Yixxo specializes in "Proseding of the 13th Annual Meeting of Miami Beach, Florida, USA, 2005"

Track Similarity to Determine Optimum Sequence Parameters for Diffusion Spectrum Imaging", which indicates that the best clinical parameter for DSI technology is to set 515 diffusion coding gradients, and set the radius b of the sampling range of Q space to be maximum bmax=6000 s/mm2 However, under this optimal parameter setting, it takes 1 hour to obtain the sampling data in the Q space, and it takes a long time to be very disadvantageous for clinical application.

Conventional methods for improving the sampling speed of DSI are generally divided into two types: one is to improve the sampling speed of the first stage, that is, to perform in the image space. The other is to improve the sampling speed of the second stage, that is, to execute in the Q space. The method for improving the sampling speed of the first stage in the image space can be referred to K. Arfanakis, pp. 675, (3), pp. 675-83, March, 2005. Space undersamping in PROPELLER imaging", this method is mainly used in the propeller magnetic image (PROPELLER MRI) using k-space undersampling, so it can reduce the 50% when sampling DWIs Sampling time. However, this conventional method requires an additional mutual cross-linking (co-registration). 习 Another method for improving the sampling speed of the first stage in the image space can be referred to D.A. Feinberg et al. at Magweii'c

Vol.48, (1), ρρ.1-5, 2002, "Simultaneous echo refocusing in EPI" and TG Reese et al. /Vocee Hill 'wg 〇/ the 14th Annual Meeting of ISMRM, Seattle, Washington, 2006 The proposed "Halving imaging time of whole brain diffusion spectrum imaging (DSI) using simultaneous echo refocusing (SER) ΕΡΓ, this method uses simultaneous echo refocusing (SER), so it can be in one echo sequence (echo train) In the case of obtaining more than two DWIs, the data acquisition time is reduced. However, this method can reduce the image quality of the obtained DWIs. Another method for improving the sampling speed of the first stage in the image space can be referred to T. Theussl on the corpse rocee hill.wg <?/ ί/ze

San Diego, CA, USA, ρρ.91-98, 2001. proposed "Optimal Regular Volume Sampling", which states that in the image space, the body-center cubic is sampled by the body-center cubic. (face-center cubic) sampling lattice to sample Computed Tomography (CT) or magnetic resonance image (MRI), which can be compared with rectangular sampling lattice (also known as Cartesian sampling lattice) (Cartesian sampling lattice)) can avoid the aliasing of structural images, so that the approximate image quality can be obtained with a small number of samples, thus increasing the sampling efficiency. The rectangular sampling lattice, the body-centered stacked lattice and the centroid stacked sampling lattice and the respective generation matrix (generating 7 1345461 matrix) can be referred to FIG. However, the above-mentioned three methods are purely selected to perform the improvement in the structural image space. Therefore, only the first stage sampling speed of the 4 DSI technique can be changed, and the second stage sampling speed performed under the Q space is not entered. Improvement. It is worth noting that the image sampling of the L4 in the existing fast imaging technology only accounts for a small part of the DSI sampling process, so the speed of sampling the DSI is accelerated depending on the second stage. The sampling speed, it is known that these three methods still can not effectively improve the sampling speed of DSI. However, the time required to create the contrast of the DWI by the diffusion gradient magnetic field is limited by the diffusion rate of the physical water molecules, which makes the second-stage Q-space sampling speed difficult to increase. It is conventional to improve the sampling speed of the second stage in the Q space by reducing the number of samples of the Q space in a fixed Q-space sampling range (L-W. Kuo 専 feet in Proceeding 〇f the 13th Annual

"Using Track Similarity to Determine Optimum Sequence Parameters for Diffusion Spectrum Imaging" as proposed in Miami Beach, Florida, USA, 2005. However, the conventional method of sampling in the Q space is based on the rectangular sampling lattice, and there is no optimized sampling lattice, so it is easy to cause signal identification error due to aliasing, so the sampling speed is increased. Limited promotion. SUMMARY OF THE INVENTION Accordingly, it is an object of the present invention to provide a diffusion spectrum contrast method that improves sampling efficiency. Thus, the diffusion spectrum imaging method of the present invention comprises the following steps: (A) obtaining a position of a plurality of sampling data in the q-space 8 1345461 by a regular-non-rectangular sampling lattice in the Q space, and obtaining a series of diffusion weights accordingly Magnetic resonance image; and (B) arranging the series of diffusion-weighted magnetic resonance images of step (A) in a Q space according to a regular_non-rectangular sampling lattice, and processing to obtain a probability density function; wherein 'step (A) The following sub-steps are included: (A-1) receiving a required number of samples set by the user in the Q space and a range of the sampling area; (A-2) in a regular-non-rectangular sampling lattice, in an iterative manner Finding a sampling interval when an actual number of samples converges to the required number of samples; and (A-3) obtaining a position of the sampled data in the Q space based on the sampling interval and the regular_non-rectangular sampling lattice.

Yet another object of the present invention is to provide a Q-space sampling method that improves sampling efficiency. The Q spatial sampling method of the present invention comprises the following steps: (a) receiving a required number of samples set by the user in the Q space and a range of the sampling area; (b) under the regular-non-rectangular sampling lattice, And determining, in an iterative manner, a sampling interval when an actual number of samples converges to the required number of samples; and (c) obtaining a position of the sampled data in the 卩 space based on the sampling interval and the regular _ non-rectangular sampling lattice. [Embodiment] The foregoing and other technical contents, features, and advantages of the present invention will be apparent from the following detailed description of the preferred embodiments of the accompanying drawings. Before the present invention is described in detail, it is to be noted that in the following description, similar processes are denoted by the same reference numerals. The first preferred embodiment of the present invention comprises the following steps: Step 1 is to obtain a regular non-rectangular sampling lattice in the Q space. The position of the sampled data in the Q space, and according to the series of DWIs^ and the "rule" of the "rule"-non-rectangular sampling lattice means that the position of the sampling point on the crystal lattice can be represented by the generator matrix of the integer period . The generator matrix is a matrix, and its row vector (c〇lUmn vector) represents the repeating direction and distance of the sample point in three-dimensional space. The length of this vector is called the unit unit sampling interval (unit sanipHng jntervai) e "non-rectangular" Means non-Cartesian sampling of the crystal lattice. The rule_non-rectangular sampling lattice of this embodiment includes a body-centered stacked sampling lattice and a centroid stacked sampling lattice, but is not limited thereto.攸..., now-non-rectangular sampling day &day; arranged in the Q space and performing inverse Fourier transform to obtain the probability density function of the water molecule. Step 3 is a radial integration of the probability density function obtained in step 2 to obtain a direction distribution function. Referring to Figure 3, the β value worthy of the phonetic note is 忍, 丨, the sampling method of 4, and the steps! Included Sub-Steps Sub-step: Receives the number of required samples 10 1345461 and a range of sampling areas set by the user in the . In the present embodiment, the sampling area is a spherical sampling area, so in this step, the radius of the spherical sampling area set by the user is received.

Substep 12· Under the regular-non-rectangular sampling lattice, the basic unit sampling interval is approximately inversely proportional to the cubic root of the number of samples in the spherical sampling region, so the iteration is repeated according to the following equation to find that the actual sampling number converges. a coarse sampling interval to the desired number of samples/near, and the coarse sampling interval is 1 basic unit sampling spacing of the regular-non-rectangular sampling lattice: d(n + l) = d(n) x (4) l ^desired ) where ' + factory is the sampling interval for the nth, n+1th iterations respectively, and the heart (η) is the actual number of samples for the nth iteration.

Sub-step 13: progressively increasing the value of the coarse sampling interval until the distance between the outermost sampling data in the spherical sampling region and the center point of the spherical sampling region is the same as the radius set in step ii, The sampling interval is the fine sampling pitch. Sub-step 14: Calculate the exact coordinates of each sampled material in the Q-slot in the Q-slot based on this fine-sampling spacing and regular-non-rectangular sampling lattice. Sub-step 15: These precise coordinates are input to a MR scanner, which is based on these precise coordinates to set the corresponding diffusion code gradient for each precise block to obtain the series of DWIs. Since this step is a conventional technique and is not an improvement focus of the present invention, it will not be repeated here. The second preferred embodiment of the diffusion spectrum imaging method of the present invention is similar to the first preferred embodiment, except that sub-step 13 and sub-step 14 of the first embodiment become: sub-step 13' : Calculate the coarse coordinates of each sample data of the spherical sampling area in the ^ space according to the rough sampling interval of the sub-step 12 and the rule_# rectangular sampling lattice'. Sub-step 14': Sub-step 13, the obtained coarse coordinates are multiplied by a ratio φ (=a/〇 to obtain a precise coordinate. And a is sub-step 13, in the outermost sampling in the spherical sampling area The distance between the data position and the center point of the spherical sampling area, and !* is the radius set by substep u.

In order to verify that the regular-non-rectangular sampling lattice used in the Q space of the embodiment of the present invention is indeed more efficient than the rectangular sampling lattice, the experimental group is compared (ie, the body core is stacked under the first embodiment). The sampling efficiency of the sampling lattice and the control group (ie: using a rectangular sampling lattice). H In order to correctly determine the sampling efficiency, a 3 Tesla magnetic vibration scanner (Siemens # Tri〇, Germany) and a double-focus echo-balanced diffusion-weighted k-wave plane pulse sequence (twice ref〇cused baiance(j ech) 〇diffusion EPI) 515 DWIs were collected as the standard group, and the parameters were set as follows. 515 isotropic diffusion coding gradients, the radius of the spherical sampling area was 6〇〇〇s/mm2, and ΤΚ/ΊΈ was 29 〇〇/15〇ms and a rectangular sampling lattice is used. The image unit volume (el) uses the uniform structure 12 1345461 'and the image resolution is 2.7 mm, and 40 layers of images are collected in the middle portion of the brain. Then the experimental group and the control group reduced the number of samples in the Q space by about 34% compared with the standard group, and the number of samples in the experimental group and the control group were as close as possible to compare the errors between the two groups and the standard group. The number of samples in the control group is 341, and after obtaining the sampling coordinates of the control group, 'the DWIs of the control group are obtained by the difference between the DWIs of the standard group. And the radius of the spherical sampling area of the experimental group is 6 〇〇〇s/mm2. And The number of samples to be sampled is close to 341, so it can be set to 341, 340, 342, etc. Then, under the flow of the first embodiment, the heart rate is 339, which is 2 samples less than the control group. After the sampling coordinates of the embodiment are obtained, the DWIs of the experimental group are obtained by the difference between the DWIs of the standard group. After obtaining the DWIs of the control group and the experimental group, the Hamming filter (Hammingfilte〇) is performed on the DWIs of the two groups respectively. And spherical windowing functi〇n to reduce the overlap caused by signal folding into the observation area (fieid 〇f view, FOV). Then, the processed signal is inversely Fourier transformed to obtain water molecules. The probability density function of the motion is further processed to obtain the = distribution function. After that, the control group and the standard group are judged by respectively obtaining the average angular error of the orientation between the control group and the standard group, the experimental group and the standard group. The error of the direction distribution function between the experimental group and the standard group. Referring to Figure 4, it can be found from the figure that only diffusion anisotropy is considered. 13 1345461 diffusion aniso Troppy,DA) When the brain position is above 0 035, the average angular error between the control group and the standard group, the experimental group and the standard group is 8 plants and 5.4 respectively. And referring to Fig. 5, when the angle error is considered to be within 2 〇. When the brain position, the control group and the experimental group were A 92. 〇 6% and 99 〇 7% of the unit volume, respectively, while the average angle error was smaller or the angle error was 2 G. The larger the unit volume is, the closer the direction distribution function of the group to the standard group is, the better the sampling quality is, which means that the sampling efficiency is high, and this means that the actual number of samples T required is reduced. The sampling time is further shortened. In addition, referring to FIG. 6, if the experimental group and the control group do not perform the spherical window function on the DWIs first, then the probability density function of the experimental group's probability density function due to the overlay is located at the corner of the observation area. Therefore, the recognition of the pattern is not affected, and the control group affects the recognition of the pattern because the repeated patterns are not all located in the corner. If the experimental group first filters it using the spherical window function, the repeating pattern of the corners can be completely removed. However, even if the control group uses the spherical window function, the repeating pattern cannot be removed. In summary, the embodiment of the present invention directly obtains the location of the best sampled data in the Q space, and then obtains a series of desired DWIs based on these locations. And the embodiment is to obtain the best sampling data based on the sampling method of the regular non-rectangular sampling lattice in the Q space, so that the sampling quality is better under the same sampling number, in other words, the invention can be used. The large sampling interval, and because the sampling interval can be larger, the number of samples required is less than the conventional one, and the sampling time is shortened. However, the above is only the preferred embodiment of the present invention, and the scope of the present invention is not limited thereto, that is, the simple equivalent change between the scope of the invention and the description of the invention is Modifications are still within the scope of the invention. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic view showing a rectangular sampling lattice, a body-centered stacked sampling lattice, and a face-centered stacked sampling lattice and respective generation matrices; FIG. 2 is a first comparison of the diffusion spectrum contrast method of the present invention. Figure 3 is a flow chart illustrating the steps involved in the steps of the first preferred embodiment; Figure 4 is a schematic diagram showing the control group and standard under different diffusion anisotropy Average angular error of the group, experimental group and standard group; Figure 5 is a schematic diagram showing the angular error at 20 for different diffusion anisotropies. The unit volume of the control group and the experimental group; and Fig. 6 is a schematic rate density function. Explain the machine of the drill group, the experimental group and the control group

15 1345461 [Description of main component symbols] 1~3 steps 11~15 steps

Claims (1)

1345461 X. Patent Application Scope 1_ A diffusion frequency spectroscopy method consists of the following steps: (A) In the Q space, 'the position of the multiple sampling data in the Q space is obtained by the regular-non-rectangular sampling lattice' and is obtained accordingly a series of diffusion-weighted magnetic resonance images; and (B) arranging the series of diffusion-weighted magnetic resonance images of step (A) in a Q space according to a regular _ non-rectangular sampling lattice, and processing to obtain a probability density function; Step (A) includes the following sub-steps: (A-1) receiving a required number of samples set by the user in the q space and a range of the sampling area; (A-2) under the regular-non-rectangular sampling lattice, Calculating the sampling interval when an actual sampling number converges to the required number of samples in an iterative manner: and (A-3) obtaining the sampling data in the Q space based on the sampling spacing and the regular _ non-rectangular sampling lattice The location. 2. A method of diffusion spectroscopy according to the scope of the patent application, wherein the regular-non-rectangular sampling lattice comprises a body-centered stacked sampling lattice. 3. According to the diffusion spectrum contrast method described in claim 1, the rule-non-rectangular sampling lattice comprises a face-centered stacked sampling lattice. 4. The diffusion spectrum contrast method according to claim 1 of the patent application, further comprising the step (C) wherein the step (c) is to process the probability density function obtained in the step (B) to obtain a direction distribution function. 5. According to the diffusion spectrum contrast method described in claim 1 of the patent application, 17 6. and in the step (B), the series of diffusion-weighted magnetic vibrations of step (4) are inverse-Fourier transforms to obtain Probability density function. The diffusion spectrum contrast method described in claim 1 of the patent application, in the step (B), the spherical window function filtering is performed on the series of diffusion-weighted magnetic vibrating shirt images in the step (4), and then the inverse Fourier is performed. Convert 'to get the probability density function. According to Shen. In the diffusion spectrum contrast method described in the patent scope of the present invention, the sampling area is a spherical sampling area, and in the step (Α-1), the radius of the spherical sampling area set by the user is further received. According to the diffusion spectrum contrast method of claim 7, wherein the step (Α-3) is based on the sampling interval of the step (Α_2) and the rule_non-rectangular sampling lattice, and each sampling data is calculated in q. Coordinates in space, and each coordinate is multiplied by a ratio to obtain a more accurate coordinate, and the ratio is the outermost sampling data in the spherical sampling area under the sampling interval of step (Α_2) The distance between the center points of the spherical sampling area is divided by the radius of the spherical sampling area. 9. The method according to claim 7, wherein the step (Α-2) and (Α-3) further comprise a step of gradually increasing the step (Α-2). The value of the sampling interval until the distance between the outermost sampling data in the spherical sampling region and the center point of the spherical sampling region is the same as the radius of the spherical sampling region, and the sampling interval at this time is used as the updated Sampling spacing according to claim 9. According to the diffusion spectrum imaging method described in claim 9, wherein in step (Α-3), according to the updated sampling spacing and the gauge 18 1345461 1 1 non-rectangular The crystal lattice is sampled and the coordinates of each sampled data in the Q space are calculated. 11. The method according to claim 1, wherein the step (A-2) repeats the iteration according to the following formula: ά(η-\Λ) = d{n)xrj,j /, V/3 N insideO1, iy desired J where ' #/1), +" respectively is the sampling interval when the nth, n+1th iterations are 'the actual number of samples in the nth iteration, And the basic unit sampling pitch of the regular-non-rectangular sampling lattice is approximately inversely proportional to the cubic root of the number of samples. 12. A Q-space sampling method suitable for use in a diffusion spectrum imaging method. The sampling method comprises the following steps: (a) receiving a required number of samples set by the user in the Q space and a range of the sampling area; (b) Under a regular-non-rectangular sampling lattice, the sampling spacing when an actual number of samples converges to the required number of samples is found in an iterative manner; and (c) based on the sampling spacing and the regular-non-rectangular sampling lattice, The location of the sampled data in the Q space is obtained. 13. The q-space sampling method according to claim 12, wherein the sampling area is a spherical sampling area, and in step (a), the radius of the spherical sampling area set by a user is further received. 14. The q-space sampling method according to item 3 of the claim patent, wherein 'step (4) is based on the sampling interval of step (8) and the regular non-moment 19 1345461 shaped sampling lattice, and each sampled data is calculated in Q. Coordinates in space, and multiplying each-coordinate by -ratio' to obtain a more accurate coordinate, which is the outermost sampling data in the spherical sampling area and the spherical sampling area at the sampling interval of step (8) The distance between the center points is divided by the radius of the spherical sampling area. 15_ According to the Q-space sampling method described in claim 13 wherein, between steps (8) and (4), a step-by-step: adding the value of the sampling interval obtained in step (b) to the ground until the spherical shape is located When the distance between the outermost sampling data in the sampling area and the center point of the spherical sampling area is the same as the radius of the spherical sampling area, the sampling interval at this time is used as the updated sampling interval. 16. The Q-space sampling method according to claim 15, wherein in step (c), each sampled data is calculated based on the updated sampling interval and the rule_non-rectangular sampling lattice. The coordinate in the Q space 17 according to the q-space sampling method described in claim 12, wherein in step (b), the iteration is repeated according to the following formula: 1/3 d(n + l) = d( n)xu) Λ/ V desired J where ' + i) is the sampling interval of the nth, n+1th iterations respectively, and M",,. heart (n) is the actual sampling at the nth iteration The number, and the basic unit sampling spacing of the rule • non-rectangular sampling lattice is approximately inversely proportional to the cubic root of the number of samples. 18. The q-space sampling method of claim 12, wherein the regular-non-rectangular sampling lattice comprises a body-centered stacked sampling lattice in 20 1345461. 19. The Q-space sampling method of claim 12, wherein the regular-non-rectangular sampling lattice comprises a centroid stacked sampling lattice. 21 1345461 VII. Designated representative map: (1) The representative representative of the case is: (3). (2) A brief description of the symbol of the representative figure: 11-15 Step 8. If there is a chemical formula in this case, please disclose the chemical formula that best shows the characteristics of the invention: