Decoding and quantizing method for sequential arterial spin marker image

Abstract

The invention discloses a method for decoding and quantifying a sequential artery spin labeling image, and belongs to the technical field of magnetic resonance imaging. The method comprises the following steps: firstly, acquiring to-be-decoded Hadamard time sequence pCASL data consisting of coded signal images; then decoding the acquired N coded signal images by using the decoding matrix to obtain a multi-time-point pCASL signal image formed by the decoded signal images; wherein the decoding matrix is obtained by performing matrix multiplication operation on an upper triangular matrix and a Hadamard matrix and then removing a first row. According to the invention, the signal strength and the signal-to-noise ratio of the decoded image can be improved, the artery transmission time estimation under different scanning protocols is realized, and the application range is wide.

Description

Technical Field

[0001] The invention belongs to the technical field of magnetic resonance imaging, and in particular relates to a decoding and quantification method for time-series arterial spin labeling images. Background Art

[0002] Arterial spin labeling (ASL) is a noninvasive magnetic resonance imaging technique widely used to quantify brain perfusion, providing rich information on cerebral hemodynamics. (Williams DS, Detre JA, Leigh JS, Koretsky AP. Magnetic resonance imaging of perfusion using spin inversion of arterial water. Proc Natl Acad Sci US A. 1992;89:4220–4220.) Carotid arterial blood is labeled with a radiofrequency pulse inversion technique. A flow-labeled image is obtained after a post-labeling delay (PLD). A control image is obtained using the same sequence of non-phase-modulated radiofrequency pulses. Pairwise subtraction of the control and flow-labeled images yields an image proportional to cerebral blood flow (CBF), known as a perfusion-weighted image.

[0003] The time required for labeled blood to travel from the labeled plane to the imaging plane is defined as arterial transit time (ATT), which provides valuable information for clinical diagnosis. ATT varies depending on the location of the brain, and a single PLD scan may have accuracy issues when used to calculate CBF. When the PLD is shorter than the ATT, CBF will be underestimated. (Woods JG, Chappell MA, Okell T W. A general framework for optimizing arterial spin labeling MRI experiments [J]. Magnetic resonance in medicine, 2019, 81 (4): 2474-2488.) Such risks can be avoided by using a longer PLD. However, since the labeled signal decays exponentially at the T1 rate (T1 of blood at 3T is 1.6 seconds), the labeled signal obtained after a long PLD time is very weak (usually less than 1% of the original signal), and the signal-to-noise ratio (SNR) is relatively low.

[0004] The International Society of Magnetic Resonance (ISMR) consensus guidelines recommend the use of multi-timepoint ASL scanning to simultaneously estimate ATT and CBF. (Woods JG, Achten E, Asllani I, et al. Recommendations for quantitative cerebral perfusion MRI using multi-timepoint arterial spin labeling: Acquisition, quantification, and clinical applications [J]. Magnetic resonance in medicine, 2024.) Multi-PLD ASL primarily involves sequential acquisition and Hadamard-coded acquisition. Sequential acquisition is typically achieved by assigning PLDs within an empirical time range that reflect the expected ATT, using a fixed label duration (LD). Sequential acquisition yields longer LDs than Hadamard-coded acquisition, resulting in stronger signals and higher SNRs, but suffers from the drawback of longer acquisition times. For Hadamard-coded acquisition, the ASL LD is divided into N-1 hour time blocks. These blocks are temporally encoded using an N×(N-1) encoding matrix formed by removing the first column from an N×N Hadamard matrix. Positive coefficients in the encoding matrix correspond to the label, while negative coefficients correspond to the control. Each image acquired with Hadamard coding is composed of the marker and control signals from all small time blocks. The ASL signal under each effective PLD can be separated by Hadamard diagonal decoding. If ASL images of N-1 effective PLDs are required, the time required for Hadamard coding acquisition is about half that of sequential acquisition (N / (2(N-1))). More importantly, the standard deviation (SD) of the noise after Hadamard diagonal decoding is times. However, due to the short LD of each small time block, the obtained ASL signal is also small and the SNR is low, which limits clinical accessibility. (Dai W, Shankaranarayanan A, Alsop DC. Volumetric measurement of perfusion and arterial transit delay using hadamard encoded continuous arterial spinlabeling[J]. Magnetic resonance in medicine, 2013, 69(4): 1014-1022.) Therefore, it is crucial to design an acquisition and decoding method that can effectively combine the advantages of both.

[0005] In terms of quantitative calculation of blood flow, Dai et al. proposed a signal weighted delay method (Dai W, Robson PM, Shankaranarayanan A, et al. Reduced resolution transit delay prescan for quantitative continuous arterial spinlabeling perfusion imaging [J]. Magnetic resonance in medicine, 2012, 67 (5): 1252-1265.) to address the problem of poor stability of time-series ASL quantitative calculation. This method obtains ATT by performing PLD weighting on a series of signals and matching them with a theoretical signal curve dictionary. However, this method is only applicable to cases where LD is fixed and has a limited scope of application.

[0006] In summary, although the traditional Hadamard encoding-decoding method can shorten the scanning time of multi-time point pCASL, it cannot improve the low signal-to-noise ratio problem caused by short LD. Secondly, the current weighted delay method is limited to the quantitative arterial transit time of time-series pCASL with a fixed LD. Therefore, it is particularly important to design a universal weighted delay calculation method. Summary of the Invention

[0007] The purpose of the present invention is to solve the above problems in the prior art and to provide a method for decoding and quantifying time-series arterial spin labeling images.

[0008] The specific technical solutions adopted in the present invention are as follows:

[0009] In a first aspect, the present invention provides a method for decoding time-series arterial spin labeling images, comprising:

[0010] S1. Obtain Hadamard time-series pCASL data consisting of N coded signal images to be decoded; the Hadamard time-series pCASL data is obtained by removing the first column from an N×N Hadamard matrix to form an N×(N-1) coding matrix, mapping each row of the coding matrix to a time-series coding pCASL sequence in a mapping manner in which a matrix element +1 corresponds to a labeling period and a matrix element -1 corresponds to a control period, and executing the time-series coding pCASL sequence one by one by a magnetic resonance imaging device and acquiring coded signal images;

[0011] S2. Use an (N-1)×N decoding matrix to decode the N collected coded signal images to obtain a multi-time point pCASL signal image composed of N-1 decoded signal images; the (N-1)×N decoding matrix is obtained by performing a matrix multiplication operation on an N×N upper triangular matrix and an N×N Hadamard matrix and then removing the first row, and the elements below the main diagonal of the N×N upper triangular matrix are all 0, and the elements on and above the main diagonal are all +1.

[0012] As a preferred embodiment of the above-mentioned first aspect, when using the (N-1)×N decoding matrix to decode the N collected coded signal images, the N coded signal images are sequentially organized into column vectors with the image as the smallest unit, and the (N-1)×N decoding matrix is subjected to matrix multiplication with the column vector to obtain N-1 decoded signal images.

[0013] As a preference for the above first aspect, N=8.

[0014] As a preferred embodiment of the first aspect, the N×(N-1) encoding matrix is:

[0015]

[0016] As a preferred embodiment of the first aspect, the (N-1)×N decoding matrix is:

[0017]

[0018] In a second aspect, the present invention provides a method for quantifying arterial transit time based on time-series arterial spin labeling images, comprising:

[0019] According to the decoding method of the time-series arterial spin labeling image of any one of the solutions of the first aspect, after obtaining the multi-time point pCASL signal image, the universal weighted delay of each voxel is calculated according to the following formula: Among them LD i is the effective marking time of the i-th decoded signal image in the multi-time point pCASL signal image, PLD i is the effective post-labeling delay time of the i-th decoded signal image in the multi-time point pCASL signal image, ΔM i is the i-th decoded signal image among the pCASL signal images at multiple time points;

[0020] According to the pre-constructed GWD-ATT comparison table, the general weighted delay GWD of each voxel is mapped to the arterial transit time ATT of the voxel, and finally the arterial transit time quantization map corresponding to all voxels is obtained.

[0021] As a preferred embodiment of the second aspect, the GWD-ATT comparison table is obtained by fitting Monte Carlo simulation data.

[0022] In a third aspect, the present invention provides a computer program product comprising a computer program / instruction, characterized in that when the computer program / instruction is executed by a processor, it can implement the method for decoding time-series arterial spin labeling images as described in any of the schemes of the first aspect above, or implement the method for quantifying arterial transmission time based on time-series arterial spin labeling images as described in any of the schemes of the second aspect above.

[0023] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the computer program can implement the method for decoding time-series arterial spin labeling images as described in any one of the schemes of the first aspect above, or implement the method for quantifying arterial transmission time based on time-series arterial spin labeling images as described in any one of the schemes of the second aspect above.

[0024] In a fifth aspect, the present invention provides a computer electronic device comprising a memory and a processor;

[0025] The memory is used to store computer programs;

[0026] The processor is configured to implement, when executing the computer program, the method for decoding time-series arterial spin labeling images as described in any one of the schemes of the first aspect, or the method for quantifying arterial transit time based on time-series arterial spin labeling images as described in any one of the schemes of the second aspect.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] The Hadamard upper triangle decoding method of the present invention is compared with the current Hadamard diagonal decoding method, and is verified on Monte Carlo simulation data and real collected data respectively. The experimental results show that:

[0029] 1. The present invention provides a time-series arterial spin labeling image decoding method based on a Hadamard upper triangular decoding matrix. This method utilizes Hadamard time-series pCASL data and achieves effective LD length variation at different pCASL time points through decoding, significantly improving image signal strength and SNR. The designed decoding method can use existing Hadamard time-series encoded pCASL data, eliminating the need for new acquisition and demonstrating excellent versatility.

[0030] 2. This invention provides a method for quantifying arterial transit time based on time-series arterial spin labeling images. This method constructs a universal weighted delay (GWD)-ATT estimator, which can be used for ATT estimation under various scanning protocols and has high universality. Using GWD to estimate the ATT of Hadamard upper triangle decoded data improves accuracy within a certain range, demonstrating broad applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Schematic diagram of the steps of a method for decoding time-series arterial spin labeling images;

[0032] Figure 2 is a flow chart in an embodiment of the present invention;

[0033] Figure 3 Figure 2 is the encoding-decoding signal diagram; (a) Hadamard temporal encoding pCASL; (b) Hadamard upper triangle decoding pCASL. "L" represents marker, "C" represents control, and "S" represents control-marker.

[0034] Figure 4 The design process of the Hadamard upper triangular decoding matrix is as follows: the upper triangular matrix is multiplied by the Hadamard diagonal decoding matrix to obtain the Hadamard upper triangular decoding matrix. Specifically, the first row of the lower decoding matrix needs to be removed; the elements of each row of the decoding matrix represent the multiplication operation performed on the signal of each row of the encoding matrix;

[0035] Figure 5 Comparison of signal strength and SNR after two types of Monte Carlo data decoding; (a) Signal strength comparison, solid spheres are Hadamard diagonal decoding signals, hollow spheres are Hadamard upper triangular decoding signals, and the lines are corresponding signal fittings; (b) SNR comparison, the change of SNR of the two decoded signals with observation time and ATT. "Had_diag" represents Hadamard diagonal decoding, and "Had_tri" represents Hadamard upper triangular decoding of the present invention;

[0036] Figure 6 This is a comparison of the signal strength of the real data after two decoding methods; (a) Hadamard diagonal decoding results, PLD decreases from left to right; (b) Hadamard upper triangle decoding results, LD decreases from left to right. Note that the color bar ranges of (a) and (b) are different;

[0037] Figure 7 Schematic diagram of the GWD method for estimating ATT; (a) GWD-ATT comparison table; (b) GWD estimated ATT error; (c) GWD estimated ATT standard deviation;

[0038] Figure 8 Computer electronic device including memory and processor. DETAILED DESCRIPTION

[0039] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways than those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below. The technical features in the various embodiments of the present invention can be combined accordingly without conflicting with each other.

[0040] The present invention provides a decoding method for time-series arterial spin labeling images. This method leverages the advantages of sequential acquisition and designs a Hadamard upper triangle decoding method that can directly utilize Hadamard time-series pCASL data. This decoding method reduces acquisition time while enhancing the signal by extending the effective LD. The specific implementation of this decoding method is described in detail below.

[0041] like Figure 1 As shown, in a preferred embodiment of the present invention, a method for decoding time-series arterial spin labeling images is provided, which includes the following steps S1 to S2. Each step is explained in detail below.

[0042] S1. Obtain Hadamard time-series pCASL data consisting of N coded signal images to be decoded; the Hadamard time-series pCASL data is obtained by removing the first column from an N×N Hadamard matrix to form an N×(N-1) coding matrix, and mapping each row of the coding matrix to a time-series coded pCASL sequence according to a mapping method in which a matrix element +1 corresponds to a labeling period and a matrix element -1 corresponds to a control period. The time-series coded pCASL sequence is executed one by one by a magnetic resonance imaging device and the coded signal images are acquired.

[0043] It should be noted that the above-mentioned acquisition of Hadamard time-series pCASL data should be understood in a generalized manner. It can be understood as offline reading from a database that has stored Hadamard time-series pCASL data, or it can be understood as online acquisition and reading of Hadamard time-series pCASL data from a magnetic resonance imaging device.

[0044] It should also be noted that the specific value of N can be adjusted according to actual conditions. When the value of N is determined, the N×N Hadamard matrix is also determined accordingly. In the embodiment of the present invention, N=8 is preferred. Therefore, the corresponding 8×8 Hadamard matrix is:

[0045]

[0046] Therefore, corresponding to the above 8×8 Hadamard matrix, after removing the first column, the 8×7 encoding matrix is formed as follows:

[0047]

[0048] The 8×7 encoding matrix needs to be mapped row by row into a temporal encoding pCASL sequence according to the specified mapping rules. Matrix element +1 is mapped to the labeled period in the sequence, and matrix element -1 is mapped to the control period in the sequence. This results in a total of eight temporal encoding pCASL sequences. The magnetic resonance imaging device can capture eight encoded signal images by executing each temporal encoding pCASL sequence. The eight encoded signal images obtained in this process are denoted as image1 through image8.

[0049] S2. Use an (N-1)×N decoding matrix to decode the N collected coded signal images to obtain a multi-time point pCASL signal image composed of N-1 decoded signal images; the (N-1)×N decoding matrix is obtained by performing a matrix multiplication operation on an N×N upper triangular matrix and an N×N Hadamard matrix and then removing the first row, and the elements below the main diagonal of the N×N upper triangular matrix are all 0, and the elements on and above the main diagonal are all +1.

[0050] It should be noted that when the present invention uses the (N-1)×N decoding matrix to decode the N collected coded signal images, the N coded signal images are sequentially organized into column vectors with the image as the smallest unit, and the (N-1)×N decoding matrix is subjected to matrix multiplication with the column vector to obtain N-1 decoded signal images.

[0051] Taking the above-mentioned case of N=8 as an example, the N collected coded signal images image1~image8 constitute a column vector with a dimension of 8. Therefore, when using a 7×8 decoding matrix to decode them, the 8 coded signal images are actually calculated using the element values of each row of the decoding matrix, and each row of the decoding matrix will obtain a corresponding decoded signal image.

[0052] In an embodiment of the present invention, the 7×8 decoding matrix is obtained by performing a matrix multiplication operation on an 8×8 upper triangular matrix and an 8×8 Hadamard matrix, and removing the first row. The 8×8 upper triangular matrix is expressed as:

[0053]

[0054] The 8×8 Hadamard matrix is as described above, and the calculated 7×8 decoding matrix is as follows:

[0055]

[0056] When decoding based on the 7×8 decoding matrix, if the eight matrix elements in a row of the matrix are a1, a2, a3, a4, a5, a6, a7, and a8 respectively, then the decoded signal image obtained by decoding this row is a1*image1+a2*image2+a3*image3+a4*image4+a5*image5+a6*image6+a7*image7+a8*image8.

[0057] Thus, the N-1 decoded signal images obtained through decoding can be output as multi-time point pCASL signal images, that is, decoding results.

[0058] The above-mentioned multi-time point pCASL signal images can be further used to quantify the arterial transit time ATT. Therefore, in an embodiment of the present invention, a method for quantifying the arterial transit time based on time-series arterial spin labeling images is also provided, which includes:

[0059] According to the decoding method of the time-series arterial spin labeling image described in S1 and S2 above, after obtaining the multi-time point pCASL signal image, the universal weighted delay of each voxel is calculated according to the following formula Among them LD i is the effective marking time of the i-th decoded signal image in the multi-time point pCASL signal image, PLD i is the effective post-labeling delay time of the i-th decoded signal image in the multi-time point pCASL signal image, ΔM i is the i-th decoded signal image among the pCASL signal images at multiple time points;

[0060] According to the pre-constructed GWD-ATT comparison table, the general weighted delay GWD of each voxel is mapped to the arterial transit time ATT of the voxel, and finally the arterial transit time quantization map corresponding to all voxels is obtained.

[0061] It should be noted that the GWD-ATT comparison table is a chart that records the mapping relationship between different GWD values and different ATT values. It can be in the form of a curve or a table, or other form that can record the association relationship, without limitation. The GWD-ATT comparison table can be obtained by fitting a large amount of sample data. In the embodiment of the present invention, this sample data can be obtained by Monte Carlo simulation. Therefore, the GWD-ATT comparison table can be obtained by fitting the data using Monte Carlo simulation in advance.

[0062] The present invention will further illustrate the decoding method of the time-series arterial spin labeling image shown in steps S1 to S2 and the arterial transmission time quantification method based on the time-series arterial spin labeling image through a specific embodiment below, and the detailed implementation process and technical effects based on specific data to facilitate understanding of the essence of the present invention.

[0063] Example

[0064] In this embodiment, the specific process is as follows Figure 2 As shown, a Hadamard time-series pCASL data generation method was first designed. Monte Carlo simulations of extended ATT and variable SNR were used to design a Hadamard upper triangular decoder. A universal weighted delay (GWD)-ATT estimator was then designed to effectively estimate the ATT under different PLD / LD protocols. Finally, the designed decoding scheme was used to decode simulated and real data, and the signal strength and SNR of the decoded pCASL were tested. Specifically, this embodiment designed an experiment to test the Hadamard upper triangular decoding method of the present invention. Similar simulation and scanning parameters were used in the experiment. The simulated data was generated using a Monte Carlo Hadamard time-series pCASL data generator, while the real data was Hadamard pCASL data from a single subject acquired at a GE 3T MRI. In the experiment, signal strength and SNR were used to evaluate the decoding results of the Monte Carlo simulated and real data. SNR reflects the clarity or quality of the signal relative to the background noise and depends on the signal strength and noise level. A higher SNR indicates a clearer image with richer details. The designed GWD-ATT estimator is used to estimate the ATT of Monte Carlo simulation decoded data. The bias and standard deviation are used to evaluate the estimated ATT performance. The smaller the bias, the higher the accuracy; the smaller the standard deviation, the higher the estimation precision.

[0065] The entire process is described in detail below.

[0066] 1. Design of Monte Carlo Hadamard Time Series pCASL Data Generator

[0067] This embodiment first designs a Hadamard time series pCASL data generation method, uses Monte Carlo simulation to extend the ATT and variable SNR, and uses it to design a Hadamard upper triangle decoding scheme.

[0068] During the Monte Carlo data generation process, 36 CBF values were uniformly sampled from [20-90] ml / 100 g / min, and 93 ATT values were uniformly sampled from [700-3000] ms, which were used as the true reference values. The pCASL signal is given by Buxton's general kinetic model:

[0069]

[0070] where ΔM is the perfusion signal, M0 is the equilibrium magnetization of brain tissue (1), α is the labeling efficiency (0.8), λ is the tissue-water partition coefficient (0.9 ml / g), and T 1t is the T1 value of the tissue (1500ms), T 1b is the T1 value of blood (1650ms), PLD is the post-labeling delay time, and LD is the labeling time. Figure 3 (a) The pCASL preparation phase is divided into seven equally spaced time blocks, where "L" represents the marker and "C" represents the control. After a certain period of time, an encoded image is acquired. By subtracting this matrix from a full "C" encoding matrix of the same size, the encoding block corresponding to "L" becomes "CL," the pCASL signal; the encoding block corresponding to "C" becomes "CC," or zero. The signals generated by different encoding blocks in one encoding cycle under their corresponding LD and PLD conditions are superimposed to produce a Hadamard temporally encoded pCASL signal. This is repeated eight times to obtain the complete encoded signal. Zero-mean Gaussian noise is added to each encoding acquisition, and the standard deviation of the noise is controlled by varying the relative SNR (reference signal parameters: LD / PLD / ATT = 1000 / 2000 / 0ms). (Guo J, Holdsworth SJ, Fan AP, et al. Comparing accuracy and reproducibility of sequential and Hadamard-encoded multidelaypseudocontinuous arterial spin labeling for measuring cerebral blood flow and arterial transit time in healthy subjects: a simulation and in vivo study [J]. Journal of Magnetic Resonance Imaging, 2018, 47(4): 1119-1132.)

[0071] It should be noted that the Monte Carlo Hadamard time-series pCASL data generator described above is primarily intended to generate Hadamard time-series pCASL data under different experimental conditions to facilitate subsequent fitting and effect verification. However, in actual applications of the present invention, the Hadamard time-series pCASL data to be decoded, consisting of N coded signal images, is not generated by the Monte Carlo Hadamard time-series pCASL data generator. Instead, it is generated by mapping each row of the coding matrix into a time-series coded pCASL sequence, which is then executed line by line by the magnetic resonance imaging device, which then acquires the coded signal images.

[0072] 2. Hadamard upper triangle decoder design

[0073] The Hadamard time series pCASL data is generated by Monte Carlo simulation and decoded. The decoding matrix needs to ensure that the decoded "L" and "C" exist in pairs. It can effectively suppress the signal of static tissue, which is crucial in real data. The existing decoding method is to directly use the Hadamard matrix (such as Figure 4 The Hadamard diagonal decoding matrix (shown in the middle) decodes the Hadamard time-series pCASL data to generate a diagonal matrix, i.e., multi-PLD pCASL data, where LD is the length of each small time block and PLD is the interval between that time block and image acquisition. However, this method decodes each signal contributed by only a single small time block, resulting in a short LD and a small pCASL signal. This results in a low SNR, limiting clinical applicability. The present invention designs a Hadamard upper triangular decoder to enhance the decoded signal strength and improve the SNR.

[0074] Hadamard time-series pCASL data is decoded using the original Hadamard matrix to generate a multi-PLD diagonal matrix (Had_diag). To obtain an upper triangular matrix, simply multiply this diagonal matrix by an upper triangular matrix to generate a multi-LD upper triangular matrix (Hadamard upper triangular matrix, Had_tri), which also meets the definition of multi-time point pCASL. The decoding matrix then becomes an upper triangular matrix multiplied by a Hadamard matrix. Correspondingly, to decode real data, the decoding matrix should be multiplied by a negative sign.

[0075] Specifically, in the decoding process of this embodiment, the Figure 4 The Hadamard upper triangular decoding matrix shown in FIG (the first row of the matrix needs to be removed, the dimension is 7×8) decodes the 8 collected coded signal images to obtain a multi-time point pCASL signal image composed of 7 decoded signal images. Figure 4 As shown in , this 7×8 decoding matrix is obtained by performing matrix multiplication of an 8×8 upper triangular matrix and an 8×8 Hadamard matrix and removing the first row. The schematic diagram of the decoded pCASL signal is shown in Figure 3 (b), where "S" represents 4 times "CL". The Hadamard upper triangular decoding matrix satisfies the requirement that "L" and "C" exist in pairs after decoding, and the extension of LD greatly enhances the signal, compensates for the increase in noise, and ultimately improves the SNR.

[0076] 3. SNR comparison module design

[0077] To highlight the advantages of Hadamard upper triangle decoding, an SNR comparison module is designed in this embodiment. Based on Monte Carlo simulation, SNR can be calculated in two ways: 1. Apply Gaussian noise with zero mean and standard deviation = σ0 (where σ0 is determined by the relative SNR) to each encoded data, and calculate the average signal strength / signal standard deviation of n repeated acquisitions at different time points after decoding; 2. Apply Gaussian noise with zero mean and standard deviation = σ to the decoded data, and calculate the average signal / signal standard deviation of n repeated acquisitions at different time points. The standard deviation σ is calculated as follows:

[0078]

[0079] Where N+1 is the dimension of the Hadamard matrix, then the noise standard deviation of the first pCASL signal after upper triangular decoding can be expressed as SNR is also related to the number of k-space excitations (NEX), which is usually The SNR calculation based on real data is to obtain the average value of the region of interest (ROI) of the decoded pCASL image between different repeated acquisitions, subtract the average image signal from the image signal of each repeated acquisition, and then calculate the noise standard deviation, so as to obtain the SNR of the decoded image.

[0080] It should be noted that the SNR comparison module is only designed in this embodiment to verify the advantages of the Hadamard upper triangle decoding solution, but in actual applications, it is not necessary to compare SNRs.

[0081] 4. General Weighted Delay GWD-ATT Estimator

[0082] In principle, ATT can be obtained by fitting multiple time points, but nonlinear least squares fitting lacks robustness when there are few data points and may produce completely unrealistic ATT values. Given the given parameters, if GWD is a monotonic function of ATT and is unaffected by CBF, ATT can be found using the GWD-ATT comparison table:

[0083]

[0084] Among them, ΔM i (ATT,LD i ,PLD i ) is the i-th decoded signal image, the signal value in the image is related to ATT,LD i ,PLD i The three parameters are related; N-1 is the number of time points, that is, the total number of decoded signal images, LD i is the effective marking time of the i-th decoded signal image, PLD iis the effective post-marking delay of the i-th decoded signal image. By plotting a monotonic curve of GWD with respect to ATT, we can infer ATT from GWD. When LD is fixed and PLD varies, the GWD-ATT estimator degenerates to the previous signal weighted delay.

[0085] Based on the universal weighted delay GWD-ATT estimator, after obtaining the multi-time point pCASL signal image according to the decoding method of the time-series arterial spin labeling image of the present invention, the universal weighted delay GWD of each voxel can be first calculated using the above formula; then, based on the aforementioned GWD-ATT comparison table, the universal weighted delay GWD of each voxel is mapped to the arterial transmission time ATT of the voxel, and finally a quantized arterial transmission time map corresponding to all voxels is obtained.

[0086] 5. Experimental verification

[0087] Based on the Monte Carlo Hadamard timing pCASL data generator, Hadamard upper triangle decoder, SNR comparison module and general weighted delay GWD-ATT estimator designed above, specific experiments are conducted to verify the effects of the decoding scheme and quantization scheme of the present invention.

[0088] 5.1. Monte Carlo Hadamard time series pCASL data generation and real data acquisition: The parameters of Monte Carlo simulation are as follows: total encoding time 4200ms, that is, each small time block LD = 600ms, PLD = 700ms, SNR = 10, NEX = 7, number of repetitions n = 10000, CBF uniformly samples 36 values in [20-90] ml / 100g / min, and ATT uniformly samples 93 values in [700-3000] ms. Each Hadamard time series pCASL signal generated by Monte Carlo is composed of the superposition of multiple small time blocks LD and the corresponding effective PLD generated signals, that is, multiple "CL". The real data scanning protocol is similar to the simulation: total encoding time 4200ms, PLD = 700ms, NEX = 1, and number of repeated acquisitions Repetition = 7. Each real acquired signal is based on Figure 3 (a) The coding matrix arrangement of a string of markers “L” and controls “C” that contribute to the results.

[0089] 5.2. Hadamard upper triangle decoding results: The results of Hadamard upper triangle decoding of Monte Carlo simulation and real collected data are shown in Figure 3(b) Each pCASL signal has an effective PLD of 700ms and an effective LD of 4200 / 3600 / 3000 / 2400 / 1800 / 1200 / 600ms. The proposed method is compared with the conventional Hadamard diagonal decoding method. After Hadamard diagonal decoding, each signal has an effective LD of 600ms and an effective PLD of 4300 / 3700 / 3100 / 2500 / 1900 / 1300 / 700ms. Both decoding methods yield multi-time-point pCASL signals: the former is a multi-LD pCASL signal, and the latter is a multi-PLD pCASL signal.

[0090] 5.3. Comparison of signal intensity and SNR: Since CBF plays a role as a coefficient in the pCASL model, the CBF value was set to 50 ml / 100 g / min in the signal simulation and SNR comparison. Figure 5 (a) Simulation of the noise-free signals at seven different observation points of the Hadamard time series pCASL generated by Monte Carlo simulation when ATT = 1800ms after two decoding methods. The Hadamard upper triangular decoding signal Had_tri increases with the increase of LD; the Hadamard diagonal decoding signal Had_diag increases first and then decreases with the increase of PLD. The signal strength of Had_tri is twice that of Had_diag or even higher. Figure 5 (b) shows the SNR variation with observation time (PLD+LD) and ATT, simulated using the above simulation parameters and the first type of SNR calculation method. Generally, a higher ATT is used to simulate pathological changes such as atherosclerosis or vascular stenosis. The SNR of Had_tri decreases as the ATT increases, but for longer observation times, the larger signal compensates well for the SNR, and the SNR is generally higher than Had_diag. In real acquisition, the decoded pCASL diagram is shown in Figure 6 To more clearly illustrate the Had_diag signal, the two decoded results are displayed using different color bars. Clearly, the Had_tri signal is stronger than the Had_diag signal, consistent with the Monte Carlo simulation results. Comparing the SNRs, as the PLD decreases, the SNRs for Had_diag decrease to 6.47 / 6.78 / 6.75 / 7.30 / 8.04 / 7.23 / 6.65, respectively; while as the LD decreases, the SNRs for Had_tri decrease to 9.22 / 9.56 / 9.20 / 8.89 / 8.07 / 7.33 / 6.65, respectively, which also meets expectations.

[0091] 5.4. General Weighted Delay GWD ATT Estimation Results: Since the true ATT value of the collected data is unknown, ATT estimation is performed only on the pCASL data generated by Monte Carlo and decoded by Hadamard diagonal / upper triangle. After Had_diag and Had_tri decoding, the effective PLD and LD are obtained respectively. In the range of prior ATT = [700-3000]ms, the GWD curve with respect to ATT is shown in Figure 7 (a), Had_tri's GWD-ATT curve has a small slope and is more robust to noise. Using GWD to estimate ATT, the results are as follows Figure 7 (b, c) The estimated ATT errors for the two decoded pCASL signals using the GWD method are similar, both being nearly unbiased. The only significant fluctuation is when ATT = PLD / LD + PLD, which is due to inherent limitations of the pCASL model. When ATT is less than 2500 ms, the Had_tri estimate has even less uncertainty, meeting the ATT range for most people.

[0092] It should be noted that each step of the above-mentioned method for decoding time-series arterial spin labeling images and the method for quantifying arterial transit time based on time-series arterial spin labeling images can essentially be implemented in the form of a computer program.

[0093] Therefore, based on the same inventive concept, Figure 8 As shown, the present invention also provides a computer electronic device provided in the above embodiment, which includes a memory and a processor;

[0094] The memory is used to store computer programs;

[0095] The processor is configured to implement the aforementioned method for decoding time-series arterial spin labeling images, or the aforementioned method for quantifying arterial transit time based on time-series arterial spin labeling images, when executing the computer program.

[0096] Furthermore, the logic instructions in the aforementioned memory can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention.

[0097] Therefore, based on the same inventive concept, the present invention provides a computer-readable storage medium corresponding to a method for decoding time-series arterial spin-labeling images, wherein the storage medium stores a computer program. When the computer program is executed by a processor, it can implement the aforementioned method for decoding time-series arterial spin-labeling images, or implement the aforementioned method for quantifying arterial transmission time based on time-series arterial spin-labeling images.

[0098] Therefore, based on the same inventive concept, the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, can implement the above-mentioned method for decoding time-series arterial spin labeling images, or implement the above-mentioned method for quantifying arterial transit time based on time-series arterial spin labeling images.

[0099] It is understood that the storage medium may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Furthermore, the storage medium may be any medium capable of storing program code, such as a USB flash drive, a mobile hard drive, a magnetic disk, or an optical disk.

[0100] It is understandable that the above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.

[0101] It should also be noted that those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working process of the system described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the various embodiments provided in this application, the division of steps or modules in the system and method is only a logical function division. In actual implementation, there may be other division methods, for example, multiple modules or steps can be combined or integrated together, and a module or step can also be split.

[0102] The embodiments described above are merely some preferred implementations of the present invention and are not intended to limit the present invention. Persons skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, any technical solution obtained by equivalent substitution or equivalent transformation falls within the scope of protection of the present invention.

Claims

1. A decoding method for time-series arterial spin labeling images, characterized in that: include: S1. Obtain Hadamard time series pCASL data consisting of N coded signal images to be decoded; The Hadamard time-series pCASL data is obtained by removing the first column from an N×N Hadamard matrix to form an N×(N-1) encoding matrix, and mapping each row of the encoding matrix to a time-series encoding pCASL sequence in a manner such that a matrix element +1 corresponds to a labeling period and a matrix element -1 corresponds to a control period. The time-series encoding pCASL sequence is executed one by one by a magnetic resonance imaging device and an encoded signal image is acquired; S2. Use an (N-1)×N decoding matrix to decode the N collected coded signal images to obtain a multi-time point pCASL signal image composed of N-1 decoded signal images; the (N-1)×N decoding matrix is obtained by performing a matrix multiplication operation on an N×N upper triangular matrix and an N×N Hadamard matrix and then removing the first row, and the elements below the main diagonal of the N×N upper triangular matrix are all 0, and the elements on and above the main diagonal are all +1.

2. The decoding method of time-series arterial spin labeling images according to claim 1, wherein: When using the (N-1)×N decoding matrix to decode the N collected coded signal images, the N coded signal images are sequentially organized into column vectors with the image as the smallest unit, and the (N-1)×N decoding matrix is multiplied by the column vector to obtain N-1 decoded signal images.

3. The decoding method of time-series arterial spin labeling images according to claim 1, wherein: Said N=8.

4. The decoding method of time-series arterial spin labeling images according to claim 3, characterized in that: The N×(N-1) encoding matrix is:

5. The decoding method of time-series arterial spin labeling images according to claim 3, characterized in that: The (N-1)×N decoding matrix is:

6. A method for quantifying arterial transit time based on time-series arterial spin labeling images, characterized in that: include: According to the decoding method of time-series arterial spin labeling images according to any one of claims 1 to 5, after obtaining the multi-time point pCASL signal images, the universal weighted delay of each voxel is calculated according to the following formula: Among them LD i is the effective marking time of the i-th decoded signal image in the multi-time point pCASL signal image, PLD i is the effective post-labeling delay time of the i-th decoded signal image in the multi-time point pCASL signal image, ΔM i is the i-th decoded signal image among the pCASL signal images at multiple time points; According to the pre-constructed GWD-ATT comparison table, the general weighted delay GWD of each voxel is mapped to the arterial transit time ATT of the voxel, and finally the arterial transit time quantization map corresponding to all voxels is obtained.

7. The method for quantifying arterial transit time based on time-series arterial spin labeling images according to claim 6, wherein: The GWD-ATT comparison table is obtained by fitting the Monte Carlo simulation data.

8. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, it can implement the decoding method of time-series arterial spin labeling images as described in any one of claims 1 to 5, or implement the arterial transit time quantification method based on time-series arterial spin labeling images as described in claim 6 or 7.

9. A computer-readable storage medium, characterized in that The storage medium stores a computer program. When the computer program is executed by the processor, the method for decoding the time-series arterial spin-labeling image according to any one of claims 1 to 5 is implemented, or the method for quantifying arterial transit time based on the time-series arterial spin-labeling image according to claim 6 or 7 is implemented.

10. A computer electronic device, characterized in that: including memory and processor; The memory is used to store computer programs; The processor is configured to implement the method for decoding time-series arterial spin labeling images according to any one of claims 1 to 5, or the method for quantifying arterial transit time based on time-series arterial spin labeling images according to claim 6 or 7, when executing the computer program.

Images