Method and device for processing ultra-short echo sequence magnetic resonance image and medium

Abstract

The invention discloses an ultra-short echo sequence magnetic resonance image processing method and device and a medium. The processing method comprises the following steps: performing magnetic resonance scanning on a to-be-imaged part of a patient by adopting a UTE sequence, acquiring N (N is greater than or equal to 4) magnetic resonance images of echo time, and storing the magnetic resonance images as image data in a DICOM format; converting the magnetic resonance image data from a DICOM format to an NIFTI format for storage; and performing analytic calculation on the long and short values by using a four-point analytic model based on piecewise exponential decay hypothesis, and finally obtaining a three-dimensional parameter distribution diagram of the rapid relaxation time, the slow relaxation time and the proportion (f) of rapid relaxation time protons in the ultra-short echo signal. According to the method, the processing time of the three-dimensional UTE image data is remarkably shortened to about 10 seconds from 10 hours of a traditional method by adopting an analytical algorithm, and meanwhile, the proportion of the number of slow relaxation protons and the number of fast relaxation protons can be effectively distinguished.

Description

Technical Field

[0001] This invention relates to the field of magnetic resonance image processing technology, and more specifically, to a rapid quantitative processing method, apparatus, and medium for ultra-short echo sequence (UTE) magnetic resonance images. Background Technology

[0002] Magnetic resonance imaging (MRI) uses radiofrequency pulses to excite spin nuclei (mainly hydrogen nuclei) in a static magnetic field, causing nuclear magnetic resonance. After the radiofrequency pulse stops, the hydrogen nuclei release energy and relax, and the resulting signal is acquired by an induction coil. Data processing then produces an image. MRI is widely used in the examination of various systems in the human body. With the increasingly serious aging population, the incidence of diseases affecting the musculoskeletal system is also increasing year by year. Bone has an extremely short T2 value, resulting in very rapid signal attenuation. Conventional MRI sequences used clinically cannot capture the magnetic resonance signal of bone. Some tissues in the human body have very short T2 values, generally less than 10 ms, and some even less than 1 ms. Conventional MRI sequences cannot encode these tissues, leading to the loss of anatomical and physiological information, which poses a challenge to disease diagnosis. Ultrashort echo time (UTE) sequences can effectively image tissues with extremely short T2 values. UTE sequences can not only clearly show the morphology and structure of the cortical bone and articular cartilage, but also provide good quantitative analysis of the components of the cortical bone and articular cartilage.

[0003] The intervertebral disc, consisting of the central nucleus pulposus, the annulus fibrosus surrounding the nucleus pulposus, and the cartilaginous endplate (CEP) located beneath the bony margins of the vertebral body, forms the intervertebral disc and is the largest avascular structure in the body. MRI-UTE sequence imaging can reveal the cartilaginous endplates and their layering in the lumbar intervertebral disc, which are not visible with conventional MRI sequences, thus allowing for the observation of damage and providing a new method for non-invasive clinical assessment of lumbar spine structures. Water content can serve as a biomarker for cartilage degeneration, allowing for the detection of subtle changes in the biochemical composition of articular cartilage earlier than changes in the morphology and structure of bones in patients with osteoarthritis.

[0004] Chinese patent document (application number: 202510291187.3, application date: 2025.03.12) discloses a method, apparatus, device, and medium for imaging cartilage calcification layers, comprising: acquiring a knee joint specimen to be imaged, and performing magnetic resonance scanning processing on the knee joint specimen to be imaged using a preset ultra-high field magnetic resonance scanner according to preset scanning parameters to obtain a number of corresponding images to be processed; taking the image to be processed located at a first echo time as the first image to be processed, and taking the image to be processed located at a second echo time as the second image to be processed; wherein, the first echo time and the second echo time are respectively the time signal for capturing calcified cartilage and the time signal for attenuating the image background in each magnetic resonance scanning imaging; performing double echo subtraction processing on the first image to be processed and the second image to be processed to obtain a soft cartilage layer containing high signal bands. The target image of the bone calcification layer meets the preset high-resolution conditions, and the location and morphological information of the cartilage calcification layer in the target image are determined; a single exponential fitting process is performed on all images to be processed to obtain a target quantitative map, and the target quantitative map is processed based on the location and morphological information to obtain quantitative information of the cartilage calcification layer, and a target high-resolution image of the knee joint specimen to be imaged is obtained based on the quantitative information and the target image; this scheme combines short echo and long echo signals and uses a single exponential model to fit the average relaxation time distribution, but it cannot distinguish the relaxation time of the short echo signal and the long echo signal and the proportion of the short echo signal component. A double exponential model fitting is required. For the three-dimensional image acquired by magnetic resonance, the double exponential model fitting needs to be performed point by point, which takes a very long time, generally more than one hour.

[0005] Current quantitative analyses of UTE mostly focus on images with short echo components or use long and short echo signals for subtraction. A few studies employ the traditional voxel-by-voxel double exponential nonlinear least squares (NLLS) fitting to obtain parameter distribution maps. However, for high-resolution three-dimensional (3D) volumetric imaging data, NLLS fitting requires iterative calculations for each voxel, resulting in long processing times.

[0006] Therefore, 1. how to distinguish the relaxation time of short echo signals and long echo signals and the proportion of short echo signal components, and 2. how to reduce processing time while ensuring accuracy have become urgent technical problems to be solved. Summary of the Invention

[0007] The objective of the present invention is to provide an ultra-short echo train (UTE) magnetic resonance imaging analysis model based on the segmented exponential decay hypothesis, and to propose a data processing method for efficiently obtaining the fast relaxation time, slow relaxation time, and the spatial distribution of the signal proportion fractions of two components, so as to overcome the defects of the traditional fitting method, such as long time consumption and inconvenient use. The present invention intends to propose an equivalent algorithm to achieve the fast calculation of the long and values and the short proportion distribution, and apply it to clinical practice. This method solves the problem of long time consumption in the traditional voxel-level double exponential non-linear fitting, and realizes the fast and efficient acquisition of the tissue biophysical parameter distribution map.

[0008] In the first aspect of the present application, a method for processing ultra-short echo train (UTE) magnetic resonance images is provided, including:

[0009] Performing magnetic resonance scanning on the part of the patient to be imaged by using an ultra-short echo train, collecting magnetic resonance images at N echo times t, and saving them as DICOM format magnetic resonance image data. The magnetic resonance image data includes: the T2 diffusion weighted image of the part to be imaged, the T2 diffusion weighted image of the part to be imaged after fat suppression, the sagittal image of the part to be measured scanned under the ultra-short echo train parameters, and the proton density weighted image. Here, N is an integer greater than or equal to 4; the echo times corresponding to the magnetic resonance images include at least 2 long echo times and at least 2 short echo times. Among them, the long echo time is greater than the echo time boundary value, and the short echo time is less than the echo time boundary value. Select 2 long echo times t = t1 and t = t2, where t2 > t1 > the echo time boundary value, and select 2 short echo times t = t3 and t = t4, where t3 < t4 < the echo time boundary value; obtain the signal intensity S at the echo time, including S1, S2, S3, S4, where S1 is the signal intensity at the echo time t1, S2 is the signal intensity at the echo time t2, S3 is the signal intensity at the echo time t3, and S4 is the signal intensity at the echo time t4;

[0010] Converting the magnetic resonance image data from the DICOM format to the NIFTI format for storage;

[0011] Using a four-point analysis model based on the segmented exponential decay hypothesis to perform analytical calculation on the long and values, and performing numerical calculation on the fast relaxation time slow relaxation time and the proportion f of protons with the fast relaxation time to obtain a three-dimensional parameter distribution map, including the following steps:

[0012] Obtaining the three-dimensional distribution of the slow relaxation time through Equation 1,

[0013]

[0014] The fast relaxation time is obtained from Equation 2. The three-dimensional distribution

[0015]

[0016] The proportion f of protons with fast relaxation time in the ultrashort echo signal is obtained from equations 3, 4, and 5.

[0017]

[0018]

[0019] Among them, A S A is the proportional coefficient corresponding to the rapid relaxation time. L This is the proportionality coefficient corresponding to the slow relaxation time; To achieve rapid relaxation time, This is the slow relaxation time.

[0020] Optionally, the minimum short echo time is less than 0.1 ms.

[0021] Optionally, the threshold value can be greater than 1 ms. For this purpose, images with more than four echo times can be recorded, such as N=6, and four groups can be selected from the recorded images for processing based on the new threshold value.

[0022] Optionally, the four-point analytical model based on the piecewise exponential decay assumption is obtained from a double exponential model, which is:

[0023]

[0024] Optionally, the slow relaxation time The three-dimensional distribution satisfies the relationship described by equations 7-8:

[0025] At t=t1,

[0026] At t = t2,

[0027] Optionally, the fast relaxation time The three-dimensional distribution satisfies the relationship described by equations 9-12;

[0028] At t=t3,

[0029] At t=t4,

[0030] Based on Equations 7 and 9, we obtain Equation 11:

[0031]

[0032] Based on Equations 8 and 10, Equation 12 is obtained:

[0033]

[0034] According to Equations 11-12, the fast relaxation time described in Equation 2 is obtained.

[0035] A second aspect of this application provides a processing apparatus for ultrashort echo sequence magnetic resonance images, used to implement the above-described processing method for ultrashort echo sequence magnetic resonance images; the processing apparatus includes:

[0036] The scanning module uses an ultra-short echo sequence to perform magnetic resonance scanning on the patient's imaging area, acquires magnetic resonance images for N echo times, and stores the acquired magnetic resonance images in DICOM format as image files, the image files including magnetic resonance image data;

[0037] The data preprocessing module, coupled to the scanning module, converts the magnetic resonance image data in the image file stored in the scanning module into NIFTI format and obtains the signal intensity at N echo times;

[0038] The numerical processing module, coupled to the data preprocessing module, calculates the proportion of protons in the fast relaxation time, slow relaxation time, and fast relaxation time according to Equations 1 to 5 based on the obtained echo time and the signal strength at the corresponding echo time, thereby obtaining a parameter distribution map.

[0039] The slow relaxation time is obtained using Equation 1. The three-dimensional distribution

[0040]

[0041] The fast relaxation time is obtained from Equation 2. The three-dimensional distribution

[0042]

[0043] The proportion f of protons during the fast relaxation time in the ultrashort echo signal is obtained from equations 3, 4, and 5.

[0044]

[0045] Among them, A S A is the proportional coefficient corresponding to the rapid relaxation time. L This is the proportionality coefficient corresponding to the slow relaxation time; To achieve rapid relaxation time, This is the slow relaxation time.

[0046] A third aspect of this application provides a computer-readable storage medium storing a computer program or computer instructions that can be loaded by a processor and executed as described above for processing ultrashort echo sequence magnetic resonance images.

[0047] Compared with the prior art, the processing method, apparatus and medium for ultrashort echo sequence magnetic resonance images provided in this application achieve at least the following beneficial effects:

[0048] First, this application considers using an equivalent echo time based on a single-exponential model during the long echo period, and then calculating a three-dimensional distribution map of the short echo time based on a double-exponential waveform during the short echo period, which can obtain the ratio of long echo protons to short echo protons.

[0049] Secondly, this method efficiently acquires UTE data and quickly processes it to obtain the proportion of protons with fast relaxation time, slow relaxation time, and fast relaxation time. For the three-dimensional lumbar spine data in this embodiment, it can process the three-dimensional short echo time distribution, long echo time distribution, and the proportion of short echo protons within a few seconds.

[0050] Of course, any product implementing this invention does not necessarily need to achieve all of the technical effects described above at the same time.

[0051] Other features and advantages of the invention will become clear from the following detailed description of exemplary embodiments of the invention with reference to the accompanying drawings. Attached Figure Description

[0052] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments of the invention and, together with their description, serve to explain the principles of the invention.

[0053] Figure 1 This is the overall flowchart of the present invention.

[0054] Figure 2 In the figure, Figure 201 is a T2 diffusion-weighted image of the lumbar spine in the sagittal position, Figure 202 is a T2 diffusion-weighted image of the lumbar spine in the sagittal position after fat suppression, and Figure 203 is a lumbar spine sagittal image scanned under the corresponding UTE sequence parameter conditions, where TE = 0.1ms.

[0055] Figure 3In the figure, Figures 301-304 show the images acquired at four echo times of the UTE sequence. Figure 301 corresponds to TE = 0.1ms, Figure 302 corresponds to TE = 0.2ms, Figure 303 corresponds to TE = 0.8ms, and Figure 304 corresponds to TE = 1.2ms. Figure 305 shows the relationship between the voxel signal intensity and the TE value, as indicated by the white arrows. The dots represent the normalized signal intensity, and the black solid line is the curve obtained by double exponential fitting.

[0056] Figure 4 Figure 401 shows the spatial distribution of rapid relaxation time, slow relaxation time, and the proportion of protons in the rapid relaxation time of the lumbar spine obtained by fitting using the traditional double exponential model NLLS. Specifically, Figure 402 shows the spatial distribution of rapid relaxation time in the intermediate layer of the lumbar spine in the sagittal plane; Figure 403 shows the spatial distribution of the proportion of protons in the rapid relaxation time in the intermediate layer of the lumbar spine in the sagittal plane; and Figure 404 shows the spatial distribution of the background signal of the fitted signal in the intermediate layer of the lumbar spine in the sagittal plane.

[0057] Figure 5 The image shown is a sagittal UTE image of the lumbar spine acquired in Embodiment 1 of the present invention, with a corresponding echo time TE = 0.1 ms. The relationship between the signal intensity of the central voxel marked by the box in Figure 501 and the echo time is shown in Figure 502.

[0058] Figure 6 Figure 601 shows the spatial distribution of the rapid relaxation time, slow relaxation time, and proton proportion of the rapid relaxation time in the lumbar spine obtained by the rapid analysis method proposed in this invention. Figure 602 shows the spatial distribution of the rapid relaxation time in the intermediate layer of the lumbar spine in the sagittal plane, Figure 603 shows the spatial distribution of the proportion of protons in the rapid relaxation time in the intermediate layer of the lumbar spine in the sagittal plane, and Figure 604 shows the spatial distribution of the background of the fitted signal in the intermediate layer of the lumbar spine in the sagittal plane.

[0059] Figure 7 This is a schematic diagram of the structure of the ultrashort echo sequence magnetic resonance image processing device proposed in this invention. Detailed Implementation

[0060] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention.

[0061] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use.

[0062] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0063] In all the examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0064] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0065] The core of this invention lies in utilizing long Relaxor components at long echo times (TE) and short echo times (TE) The physical assumption that the relaxation component signal decays to zero transforms the complex double-exponential nonlinear fitting problem into a solvable analytical expression, enabling efficient acquisition of the fast relaxation time component, the slow relaxation time component, and the spatial distribution of both components.

[0066] In MRI-UTE sequence imaging, if the tissue of interest is predominantly composed of short T2 components, a single T2 exponential decay model can be used. However, if ultra-short T2 tissue constitutes a minority, the single exponential model is not suitable, necessitating the development of a multi-component model. It should be noted that the proportion of ultra-short T2 tissue varies depending on the tissue, generally f < 0.4. When the tissue is diseased, the proportion f value will also change. This application does not impose specific limitations; the choice should be made based on the actual situation.

[0067] This application provides a method for processing ultrashort echo sequence magnetic resonance images, such as... Figure 1 As shown, it includes:

[0068] Step 101: Use an ultrashort echo sequence to perform magnetic resonance scanning on the patient's imaging site, collect magnetic resonance images at N echo times (TE), and save them as DICOM - format magnetic resonance image data. The magnetic resonance image data includes: the T2 diffusion - weighted image of the imaging site, the T2 diffusion - weighted image of the imaging site after fat suppression, the sagittal image of the measured site scanned under the ultrashort echo sequence parameters, and the proton density - weighted image. Here, N is an integer greater than or equal to 4. The echo times corresponding to the magnetic resonance images include at least 2 long echo times and at least 2 short echo times. Among them, the long echo time is greater than the echo - time boundary value, and the short echo time is less than the echo - time boundary value. In the present invention, the specific echo - time boundary value is not limited. In this embodiment, the echo - time boundary value can be selected as 1 ms. Select 2 long echo times t = t1 and t = t2, where t2>t1>1 ms, and select 2 short echo times t = t3 and t = t4, where t3<t4<1 ms. Obtain the signal intensities S at the echo times, including S1, S2, S3, S4. S1 is the signal intensity at the echo time t1, S2 is the signal intensity at the echo time t2, S3 is the signal intensity at the echo time t3, and S4 is the signal intensity at the echo time t4.

[0069] Step 102: Convert the magnetic resonance image data from DICOM format to NIFTI format for storage.

[0070] Step 103: Use a four - point analytical model based on the piece - wise exponential decay assumption to perform analytical calculations on the long and short values to obtain a three - dimensional parameter distribution map of the fast relaxation time slow relaxation time and the proportion f of protons with fast relaxation time, including the following steps:

[0071] Step 1031: Obtain the three - dimensional distribution of the slow relaxation time through Equation 1

[0072]

[0073] Step 1032: Obtain the three - dimensional distribution of the fast relaxation time through Equation 2

[0074]

[0075] Step 1033: Obtain the proportion f of protons with fast relaxation time in the ultrashort echo signal according to Equations 3, 4, and 5.

[0076]

[0077]

[0078] Among them, A S A is the proportional coefficient corresponding to the rapid relaxation time. L This is the proportionality coefficient corresponding to the slow relaxation time; To achieve rapid relaxation time, This is the slow relaxation time.

[0079] Specifically, the method for acquiring MRI images based on UTE sequences and processing them to obtain the fast relaxation time, slow relaxation time, and the proportion of protons corresponding to the fast relaxation time includes the following processing steps:

[0080] 1. For UTE sequences, set at least four echo times, with two greater than the echo time threshold and two less than the echo time threshold. The minimum echo time is recommended to be less than 0.1ms or set to a smaller echo time according to the characteristics of the tissue to be imaged.

[0081] 2. Use magnetic resonance imaging equipment to scan the human body to obtain MRI images of the target area and save them in DICOM format.

[0082] 3. Use software tools such as 3D Slicer and ITK-SNAP to convert UTE-MRI images with multiple TE values ​​(minimum value less than 1ms) from DICOM format and save them as NIFTI format.

[0083] 4. Based on the double exponential model, the distribution map of UTE sequence correlation parameters is obtained by fitting. More generally, for four determined echo times, the distribution map of each component can be calculated by the formula derived in this invention, and the spatial distribution map of each component is saved on the computer in NIFTI format.

[0084] Specifically, the processing method of this application includes data acquisition, which involves acquiring magnetic resonance images at multiple echo times using a magnetic resonance imaging (MRI) device and saving them as DICOM format image files. Data acquisition requires selecting at least four echo times, two of which are greater than the echo time boundary value and two of which are less than the echo time boundary value.

[0085] Specifically, the processing method in this application requires preprocessing the acquired DICOM format image files. Software tools such as 3D Slicer and ITK-SNAP are used to convert UTE-MRI images with multiple TE values ​​(minimum value less than the echo time threshold) into NIFTI format. The software tool is opened, the stored patient image data in DICOM format is retrieved from the software, and then the image data is saved to the computer using the .nii format selected from the open-source software.

[0086] In some optional embodiments provided by this invention, the minimum short echo time is less than 0.1 ms. Since the short relaxation time of cartilage, ligaments, and lumbar intervertebral discs is generally on the order of 0.1 ms, the minimum echo value is recommended to be less than 0.1 ms, or a smaller echo time should be set according to the characteristics of the tissue to be imaged.

[0087] It should be noted that for biological tissues where the boundary values ​​are not very clear, images with more than four echo times can be acquired, such as N=6, and then the echo time boundary values ​​can be flexibly selected according to the lesion to be examined. In particular, when the range of tissue relaxation time is not clear, image data corresponding to more than four echo values ​​can be acquired. Specifically, when acquiring MRI images with multiple echo times using an MRI machine, at least four TE values ​​should be selected, and six TE values ​​are recommended. This allows for more flexible selection of boundary values ​​for tissues with complex composition or diseased tissues where the range of short relaxation time is uncertain. When N=6, images corresponding to two larger TE values ​​are selected to calculate the three-dimensional distribution of slow relaxation time, and images corresponding to two smaller TE values ​​are selected to calculate the three-dimensional distribution of fast relaxation time; the middle two of the six TE values ​​can be used to calculate different sizes of fast echo sequences depending on the tissue composition being examined.

[0088] In some optional embodiments provided by the present invention, the four-point analytical model based on the piecewise exponential decay assumption is obtained from the double exponential model, which is as follows:

[0089]

[0090] In some optional embodiments provided by the present invention, the slow relaxation time The three-dimensional distribution satisfies the relationship described by equations 7 and 8:

[0091] At t=t1,

[0092] At t = t2,

[0093] In some optional embodiments provided by the present invention, the fast relaxation time The three-dimensional distribution satisfies the relationship described by equations 9-12;

[0094] At t=t3,

[0095] At t=t4,

[0096] Based on equations 7 and 9, we obtain equation 11:

[0097]

[0098] According to Equation 8 and Equation 10, Equation 12 is obtained:

[0099]

[0100] According to Equation 11 - Equation 12, the fast relaxation time described by Equation 2 can be obtained

[0101] It should be noted that in this application, the double exponential model is

[0102]

[0103] where A S [[ID=No.18]]and A L are the proportionality coefficients corresponding to the two components of the fast relaxation time and the slow relaxation time respectively, is the short relaxation time, is the long relaxation time. However, in actual clinical practices such as lumbar spine and bone joints, a relatively large resolution often needs to be set to clearly image the bone surface, which will increase the voxel number of the image. Based on the double exponential model, the fitting time-consuming will be greatly extended, and the processing time required by a general desktop computer ranges from several hours to dozens of days, etc. For this reason, a method for quickly processing UTE data is needed.

[0104] Generally, for the long relaxation time, such as t > 1ms, the fast term tends to zero. Therefore, Equation 6 is simplified to:

[0105]

[0106] By collecting two echo times, t = t1 and t = t2, where t2 > t1 > 1ms, the corresponding magnetic resonance images S1 and S2 are respectively given by the following equations:

[0107]

[0108] Dividing Equation 7 by Equation 8, we get the three-dimensional distribution of

[0109]

[0110] For the UTE sequence, two sets of echo times t3 < t4 < 1ms need to be collected, and the corresponding images S3 and S4 are given by Equation 6,

[0111]

[0112] From Equation 7 and Equation 9, we can obtain

[0113]

[0114] Similarly, from equations 8 and 10, we can obtain...

[0115]

[0116] Dividing equations 11 and 12 and combining them with equation 1, we can obtain:

[0117]

[0118] That is, based on the above analysis, the fast relaxation time corresponding to the ultrashort echo sequence can be calculated using Equations 1 and 2 from the magnetic resonance images corresponding to the four sets of echo times. and slow relaxation time The three-dimensional distribution.

[0119] The UTE signal includes short-echo signal protons and long-echo signal protons. Considering the proportion f of short-echo signal protons (fast relaxation time protons), we can obtain the following from Equation 6:

[0120]

[0121] Combining equations 7 and 11, we get:

[0122]

[0123] The proportion of protons during the fast relaxation time can be obtained from equations 3, 4 and 5.

[0124] Therefore, by performing data processing and fitting according to Equations 1-5, we can quickly obtain parameter distribution maps such as the proportion of protons in fast relaxation time, slow relaxation time, and fast relaxation time, thus achieving the goal of this application.

[0125] The embodiments of this application, using the lumbar spine as an example, illustrate the specific process of data acquisition, preprocessing, data processing, and storage. The described process and methods are also applicable to UTE imaging of other body parts.

[0126] Comparative Example 1

[0127] This comparative study uses the UTE sequence parameter distribution based on a double exponential model for magnetic resonance image processing.

[0128] Here, we take the ultra-short echo sequence images of the lumbar spine as an example.

[0129] Step 1: Perform MRI scans of the patient's lumbar spine using an ultra-short echo UTE sequence. The scan parameters are set as follows: repetition time (TR) is 14.3 ms, the four echo times are TE1 = 0.096 ms, TE2 = 0.2 ms, TE3 = 0.8 ms and TE4 = 1.2 ms, the flip angle is 10°, the bandwidth is 488 Hz, the field of view (FOV) is 380 x 380 mm, the matrix is ​​512 x 512, and the slice thickness is 2 mm.

[0130] With the above scanning parameter settings, magnetic resonance imaging (MRI) images were acquired at four echo times at different echo times and saved as DICOM format image files. The image files included MRI image data, which included: T2 diffusion-weighted images of the area to be imaged, T2 diffusion-weighted images of the area to be imaged after fat suppression, sagittal images of the lumbar spine taken with ultra-short echo sequences, and proton number density-weighted images. The echo times included two long echo times and two short echo times, where the long echo time was greater than 1 ms and the short echo time was less than 1 ms. The two long echo times t = t1 and t = t2 were selected, specifically, t2 = 1.2 ms and t1 = 0.8 ms. The two short echo times t = t3 and t = t4 were selected, specifically, t3 = 0.1 ms and t4 = 0.2 ms. Obtain the signal strength S at the echo time, including S1, S2, S3, and S4. S1 is the signal strength at echo time t1, S2 is the signal strength at echo time t2, S3 is the signal strength at echo time t3, and S4 is the signal strength at echo time t4.

[0131] Step 2: Convert the MRI image data from DICOM format to NIFTI format for storage. Specifically, use the 3DSlicer open-source software to convert UTE-MRI images with multiple TE values ​​(minimum value less than 1ms) to NIFTI format. The specific steps are: open the open-source software, retrieve the stored DICOM format data of the patient's lumbar spine from the software, and then select the .nii format in the open-source software to save the image data to the computer.

[0132] Figure 201 shows a T2-weighted diffusion image of the lumbar spine in the sagittal plane. Figure 202 shows a T2-weighted diffusion image of the lumbar spine in the sagittal plane after fat suppression. Figure 203 shows a lumbar spine sagittal image scanned under the corresponding UTE sequence parameters, where TE = 0.1 ms.

[0133] Figure 3In the figure, Figures 301-304 show the images acquired at four echo times (TE) of the UTE sequence. In this embodiment, the echo time boundary value can be selected as 0.6 ms. Figure 301 corresponds to TE = 0.1 ms, Figure 302 corresponds to TE = 0.2 ms, Figure 303 corresponds to TE = 0.8 ms, and Figure 304 corresponds to TE = 1.2 ms. Figure 305 shows the fitting result of a single voxel, specifically the relationship between the voxel signal intensity indicated by the white arrow and the TE value. The dots represent the normalized signal intensity, and the black solid line is the curve obtained by double exponential fitting. From the fitting result, the shortest T2* is 0.45 ms, and the component proportion corresponding to the shortest T2* is 69.6%.

[0134] Step 3: Use a bi-exponential model to fit and obtain the distribution map of the proportion of protons with short relaxation time, long relaxation time and short relaxation time in the lumbar spine. Specifically, the bi-exponential model is as follows:

[0135]

[0136] The acquired multi-echo UTE 3D images were substituted into a bi-exponential model for fitting. After about 10 hours of fitting calculations, the distribution map of the proportion of protons in short relaxation time, long relaxation time and short relaxation time in the lumbar spine was obtained.

[0137] Figure 4 Figure 401 shows the spatial distribution of rapid relaxation time, slow relaxation time, and the proportion of protons in the rapid relaxation time of the lumbar spine obtained by fitting the traditional double exponential model NLLS. Figure 402 shows the spatial distribution of rapid relaxation time in the intermediate layer of the lumbar spine in the sagittal plane, Figure 403 shows the spatial distribution of the proportion of protons in the rapid relaxation time of the intermediate layer of the lumbar spine in the sagittal plane, and Figure 404 shows the spatial distribution of the background signal of the fitted signal in the intermediate layer of the lumbar spine in the sagittal plane.

[0138] Example 1

[0139] This embodiment performs magnetic resonance image processing based on the UTE sequence parameter distribution of the fast processing model proposed in this application. Steps 1 and 2 are the same as in the comparative example.

[0140] Step 3: A four-point analytical model based on the piecewise exponential decay assumption is used to analyze the length and shortness of the model. The value is analyzed and calculated to obtain the fast relaxation time. Slow relaxation time The method includes the following steps: substituting the obtained echo time and corresponding signal intensity data into the following fitting formula for fitting calculation, where,

[0141] The fast relaxation time is obtained from Equation 2. The three-dimensional distribution of (short T2 value)

[0142]

[0143] The slow relaxation time is obtained using Equation 1. The three-dimensional distribution of (long T2 value),

[0144]

[0145] Based on equations 3, 4, and 5, the proportion of short echo signal protons in the ultrashort echo signal is f.

[0146]

[0147] Among them, A S A is the proportional coefficient corresponding to the rapid relaxation time. L This is the proportionality coefficient corresponding to the slow relaxation time; To achieve rapid relaxation time, This is the slow relaxation time.

[0148] The data processing method proposed in this invention performs calculations on each voxel in the tissue to be tested, and the calculations for all voxels can be quickly achieved through matrix calculations. Figure 5 Figure 501 shows the lumbar spine MRI image corresponding to an echo time of 0.1 ms. The relationship between the MRI signal intensity and echo time at the location marked by the red box is as follows: Figure 5 Figure 502 shows the method proposed in this invention. Here, the method is described for this specific voxel. The two long echo times are selected as t = t1 = 5.3 ms and t = t2 = 7.9 ms, and the corresponding signal intensities are as follows: Signal intensity S1 at echo time t1 = 501 arbitrary intensity units (corresponding to...). Figure 5 The signal strength S2 at echo time t2 is 453 arbitrary intensity units (the third point from the left in Figure 502). Figure 5 (The fourth point from left to right in Figure 502). The slow relaxation time at this voxel is obtained using Equation 1. The value of (long T2 value) is,

[0149]

[0150] Combining the signal strength S3 = 4060 arbitrary intensity units at two short echo times t3 = 0.1 ms (corresponding to...) Figure 5 The first point from left to right in Figure 502, and the signal strength S4 = 1132 arbitrary intensity units corresponding to t4 = 2.7ms (corresponding to...). Figure 5 (The second point from the left in Figure 502). The fast relaxation time is then obtained using Equation 2. The value of (short T2 value) is,

[0151]

[0152] The proportional coefficient A corresponding to the slow relaxation time L for

[0153]

[0154] The proportional coefficient A corresponding to the fast relaxation time S for,

[0155]

[0156] Therefore, according to Equation 3, the proportion of protons in the short echo signal is f.

[0157]

[0158] Taking the Lenovo Thinkpad T14 laptop as an example, with an Intel(R) Core™ i7-10510U CPU and 16GB of memory, the calculation time for a single voxel data described by Equations 1-5 on the Matlab R2025a platform is approximately 2.5ms, while the time required for fitting using the double exponential model described by Equation 6 in Comparative Example 1 is 5.83s. The double exponential fitting model in Comparative Example 1 requires point-by-point calculations on the three-dimensional volumetric image, and even with parallel computing algorithms, it would take several weeks to process a set of UTE images for a group of cases. The method of this invention can also greatly improve the data processing speed through array matrix operations, enabling the processing of three-dimensional UTE volumetric data to be completed within tens of seconds or even seconds, providing rapid short-echo signal distribution maps for clinical diagnosis.

[0159] The distribution maps of multiple parameters, such as long T2, short T2, and the proportion of short T2 components, obtained after fitting for approximately 10 seconds are shown below. Figure 6 As shown, Figure 6 Figure 601 shows the spatial distribution of rapid relaxation time, slow relaxation time, and the proportion of protons in the rapid relaxation time of the lumbar spine obtained by the rapid analysis method proposed in this invention. Specifically, Figure 602 shows the spatial distribution of rapid relaxation time in the intermediate layer of the lumbar spine in the sagittal plane; Figure 603 shows the spatial distribution of the proportion of protons in the rapid relaxation time of the intermediate layer of the lumbar spine in the sagittal plane; and Figure 604 shows the spatial distribution of the background of the fitted signal in the intermediate layer of the lumbar spine in the sagittal plane. All the above images are saved as NIFTI format files (file extension ".nii") for subsequent clinical diagnosis or quantitative analysis.

[0160] Because the UTE sequence is a two-component model and focuses more on the short echo component, in Figure 6 In the images, the lumbar spine, intervertebral discs, and vertebrae are clearly visible, and the image quality is comparable to that of the double exponential model fitting results. Therefore, the method proposed in this invention can be more effectively applied to clinical practice.

[0161] Example 2

[0162] This embodiment provides an ultrashort echo sequence (UTE) magnetic resonance imaging processing device, such as... Figure 6 As shown, the processing apparatus for implementing the ultrashort echo sequence magnetic resonance image processing method of Embodiment 1 includes:

[0163] The scanning module 100 uses an ultra-short echo sequence to perform magnetic resonance scanning on the patient's imaging area, acquiring magnetic resonance images at N echo times. The scanned magnetic resonance images are stored as image files in DICOM format, and the image files include magnetic resonance image data; that is, the image data used to generate the patient's image data, including the magnetic resonance scanning equipment, which can perform magnetic resonance scanning on different parts of the patient, and store the multi-echo images obtained from the scan in DICOM format on the local computer, and can also be uploaded to the hospital's internal data storage workstation;

[0164] The data preprocessing module 200, coupled to the scanning module 100, converts the magnetic resonance imaging data in the image files stored in the scanning module 100 into NIFTI format and obtains the signal intensity at N different echo times. That is, it converts the patient's image data from the machine-storage format into a format that the data processing module can easily read. Specifically, the medical images of multiple patients acquired are downloaded to the local computer on the workstation, and the patient's image data is converted from DICOM format to NIFTI format (file extension .nii) using open-source software such as 3D Slicer and ITK-SNAP.

[0165] The data processing module 300, coupled to the data preprocessing module 200, calculates the fast relaxation time, slow relaxation time, and the proportion of protons during the fast relaxation time according to Equations 1 to 5 based on the obtained echo time and the corresponding signal strength at the echo time, thereby obtaining a parameter distribution map. In other words, the original data is processed using the calculation algorithm proposed in this application to obtain the three-dimensional spatial distribution of the slow echo time, fast echo time, and the proportion of protons during the slow echo. The specific process includes:

[0166] Select two long echo acquisition times, t = t1 and t = t2, where t2 > t1 > 1 ms, and the corresponding magnetic resonance images S1 and S2 are obtained. The slow relaxation time can be obtained through the following relationship. The three-dimensional distribution

[0167]

[0168] For two sets of echo times \(t3 \lt t4 \lt 1\ ms\), the corresponding images \(S3\) and \(S4\), the fast relaxation time can be obtained through the following relational expression of the three-dimensional distribution,

[0169]

[0170] Through the magnetic resonance images corresponding to four sets of echo times, the fast relaxation time corresponding to the ultrashort echo sequence can be calculated from Equation 1 and Equation 2 and the slow relaxation time of the three-dimensional distribution.

[0171] The UTE signal includes short echo signal protons and long echo signal protons. According to Equation 3, Equation 4, and Equation 5, the proportion rate \(f\) of fast relaxation time protons in the ultrashort echo signal is obtained,

[0172]

[0173] The proportion of fast relaxation time protons can be obtained from Equation 3 - Equation 5.

[0174] Example 3

[0175] Specifically, according to the embodiments of the present application, the above processing method can be implemented as a computer software program. ​​​​​​​This invention considers using an equivalent echo time based on a single-exponential model during the long echo period, and then calculating a three-dimensional distribution map of the short echo time based on a double-exponential waveform during the short echo period, which can distinguish the ratio of long-echo protons to short-echo protons. This method efficiently acquires UTE data and quickly processes it to obtain the proportion of protons in fast relaxation time, slow relaxation time, and fast relaxation time. For three-dimensional lumbar spine data, conventional magnetic resonance imaging processing based on the UTE sequence parameter distribution of a double-exponential model requires approximately 10 hours to obtain the distribution map of the proportion of protons in short relaxation time, long relaxation time, and short relaxation time. However, the rapid processing method of this application can obtain the three-dimensional distribution of short echo time, long echo time, and the proportion of short echo protons in approximately 10 seconds, greatly improving the processing speed.

[0179] While specific embodiments of the invention have been described in detail by way of examples, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of the invention. Those skilled in the art should understand that modifications can be made to the above embodiments without departing from the scope and spirit of the invention. The scope of the invention is defined by the appended claims.

Claims

1. A method for processing ultrashort echo sequence magnetic resonance images, characterized in that, Comprising: Performing magnetic resonance scanning on the part of the patient to be imaged using an ultrashort echo sequence, collecting magnetic resonance images at N echo times, and saving them as magnetic resonance image data in DICOM format. The magnetic resonance image data includes: T2 diffusion-weighted image of the part to be imaged, T2 diffusion-weighted image of the fat-suppressed part to be imaged, sagittal image of the part to be measured scanned under ultrashort echo sequence parameters, and proton density-weighted image; where N is an integer greater than or equal to 4; the echo times corresponding to the magnetic resonance images include at least 2 long echo times and at least 2 short echo times, where the long echo time is greater than the echo time boundary value, and the short echo time is less than the echo time boundary value; selecting 2 long echo times t = t1 and t = t2, where t2 > t1 > the echo time boundary value, and selecting 2 short echo times t = t3 and t = t4, where t3 < t4 < the echo time boundary value; obtaining the signal intensity S at the echo time, including S1, S2, S3, S4, where S1 is the signal intensity at the echo time t1, S2 is the signal intensity at the echo time t2, S3 is the signal intensity at the echo time t3, and S4 is the signal intensity at the echo time t4; Converting the magnetic resonance image data from the DICOM format to the NIFTI format for storage; Using a four-point analytical model based on the piecewise exponential decay assumption to analyze the length and shortness The value is analyzed and calculated to obtain the fast relaxation time. Slow relaxation time The proportion of protons during the fast relaxation time, f, is numerically calculated to obtain a three-dimensional parameter distribution map, including the following steps: The slow relaxation time is obtained using Equation 1. The three-dimensional distribution The fast relaxation time is obtained from Equation 2. The three-dimensional distribution Obtaining the proportion f of protons with fast relaxation time in the ultrashort echo signal according to Equation 3, Equation 4, and Equation 5, Among them, A S A is the proportional coefficient corresponding to the rapid relaxation time. L This is the proportionality coefficient corresponding to the slow relaxation time; To achieve rapid relaxation time, This is the slow relaxation time.

2. The method for processing ultrashort echo sequence magnetic resonance images according to claim 1, wherein The minimum value of the short echo time is less than 0.1 ms.

3. The method for processing ultrashort echo sequence magnetic resonance images according to claim 1, wherein The four-point analytical model based on the piecewise exponential decay assumption is obtained from the double exponential model, and the double exponential model is:

4. The method for processing ultrashort echo sequence magnetic resonance images according to claim 3, wherein The slow relaxation time The three-dimensional distribution satisfies the relationship described by equations 7-8: At t=t1, At t = t2, 5. The method for processing ultrashort echo sequence magnetic resonance images according to claim 4, wherein The fast relaxation time The three-dimensional distribution satisfies the relationship described by equations 9-12; At t=t3, When t = t4, According to Equation 7 and Equation 9, obtaining Equation 11: According to Equation 8 and Equation 10, obtaining Equation 12: According to Equations 11-12, the fast relaxation time described in Equation 2 is obtained.

6. A processing apparatus for ultrashort echo sequence magnetic resonance images, characterized in that, For implementing the method for processing ultrashort echo sequence magnetic resonance images according to any one of claims 1 to 5; the processing device includes: A scanning module that performs magnetic resonance scanning on the part of the patient to be imaged using an ultrashort echo sequence, collects magnetic resonance images at N echo times, and stores the scanned magnetic resonance images as image files in DICOM format. The image files include magnetic resonance image data; A data preprocessing module coupled to the scanning module, converting the magnetic resonance image data in the image files stored by the scanning module into the NIFTI format, and obtaining the signal intensity at N echo times; A data processing module coupled to the data preprocessing module, numerically calculating the fast relaxation time, slow relaxation time, and the proportion of protons with fast relaxation time according to the obtained echo time and the signal intensity at the corresponding echo time, in accordance with Equation 1 to Equation 5, to obtain a parameter distribution map, where The slow relaxation time is obtained using Equation 1. The three-dimensional distribution The fast relaxation time is obtained from Equation 2. The three-dimensional distribution The proportion f of protons during the fast relaxation time in the ultrashort echo signal is obtained from equations 3, 4, and 5. Among them, A S A is the proportional coefficient corresponding to the rapid relaxation time. L This is the proportionality coefficient corresponding to the slow relaxation time; To achieve rapid relaxation time, This is the slow relaxation time.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program or computer instructions that can be loaded by a processor and executed as described in any one of claims 1 to 5 for processing ultrashort echo sequence magnetic resonance images.

Images