WO2025007834A1 - 一种优化分子交换磁共振测量采样轨迹的方法、分子交换磁共振测量方法及装置 - Google Patents

一种优化分子交换磁共振测量采样轨迹的方法、分子交换磁共振测量方法及装置 Download PDF

Info

Publication number
WO2025007834A1
WO2025007834A1 PCT/CN2024/102846 CN2024102846W WO2025007834A1 WO 2025007834 A1 WO2025007834 A1 WO 2025007834A1 CN 2024102846 W CN2024102846 W CN 2024102846W WO 2025007834 A1 WO2025007834 A1 WO 2025007834A1
Authority
WO
WIPO (PCT)
Prior art keywords
magnetic resonance
sampling
network
sampling trajectory
physiological
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2024/102846
Other languages
English (en)
French (fr)
Inventor
白瑞良
程昭玮
金心宇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University
Original Assignee
Zhejiang University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhejiang University filed Critical Zhejiang University
Priority to JP2026500279A priority Critical patent/JP7875650B1/ja
Publication of WO2025007834A1 publication Critical patent/WO2025007834A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N24/00—Investigating or analyzing materials by the use of nuclear magnetic resonance, electron paramagnetic resonance or other spin effects
    • G01N24/08—Investigating or analyzing materials by the use of nuclear magnetic resonance, electron paramagnetic resonance or other spin effects by using nuclear magnetic resonance
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R33/00—Arrangements or instruments for measuring magnetic variables
    • G01R33/20—Arrangements or instruments for measuring magnetic variables involving magnetic resonance
    • G01R33/44—Arrangements or instruments for measuring magnetic variables involving magnetic resonance using nuclear magnetic resonance [NMR]
    • G01R33/48—NMR imaging systems
    • G01R33/50—NMR imaging systems based on the determination of relaxation times, e.g. T1 measurement by IR sequences; T2 measurement by multiple-echo sequences
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/30—Assessment of water resources

Definitions

  • the present invention belongs to the technical field of magnetic resonance quantitative measurement, and in particular relates to a method for optimizing a molecular exchange magnetic resonance measurement sampling trajectory, a molecular exchange magnetic resonance measurement method and a device.
  • Multidimensional nuclear magnetic resonance (NMR) spectroscopy can provide a comprehensive and detailed analysis of complex molecular systems and is widely used in biological research as a non-invasive technique [Bai R, Benjamini D, Cheng J, et al. Fast, accurate 2D-MR relaxation exchange spectroscopy (REXSY): Beyond compressed sensing [J]. The Journal of chemical physics, 2016, 145(15): 154-202].
  • the long acquisition time of multidimensional spectroscopy techniques limits their application in analyzing fast dynamics under specific physical conditions and studying human/animal bodies.
  • the diffusion exchange spectroscopy (DEXSY) method [Callaghan P T, Furo I. Diffusion-diffusion correlation and exchange as a signature for local order and dynamics [J].
  • the object of the present invention is to provide a method for optimizing the sampling trajectory of molecular exchange magnetic resonance measurement, which improves the efficiency and accuracy of physiological/physical parameter estimation and can be applied to different molecular exchange magnetic resonance measurement methods.
  • a method for optimizing a molecular exchange magnetic resonance measurement sampling trajectory comprising:
  • an end-to-end unified network model including a sampling trajectory optimization network, a noise signal generation network, and a parameter estimation network:
  • the sampling trajectory optimization network simulates the process of signal generation by the molecular exchange magnetic resonance measurement method, and generates magnetic resonance signals according to the corresponding physiological/physical parameters, sampling trajectory, and acquisition parameters;
  • the noise signal generation network simulates the noise distribution in the real magnetic resonance measurement scene, and generates a noisy magnetic resonance signal;
  • the parameter estimation network implements the magnetic resonance signal fitting function, and outputs physiological/physical parameters according to the noisy magnetic resonance signal;
  • step (3) Based on the end-to-end unified network model in step (3), the simulated physiological/physical parameters in step (1) are used as input for training under specified acquisition parameters until the network converges and the training is completed, thereby obtaining an optimized sampling trajectory and a corresponding parameter estimation network.
  • the method for optimizing the molecular exchange magnetic resonance measurement sampling trajectory mainly includes two parts: a sampling trajectory optimization network that simulates the process of generating signals by the molecular exchange magnetic resonance measurement method; and a parameter estimation network that realizes the magnetic resonance signal fitting process.
  • the two network parts are connected by a noise signal generation network, thereby combining the sampling trajectory optimization and parameter estimation into an end-to-end unified network model, realizing the network training with the direct goal of improving the parameter estimation accuracy, and obtaining an optimized sampling trajectory and parameter estimation network.
  • step (1) in order to ensure the diversity of the data and the generalization of the method, all parameters are obtained by average random sampling independently within the corresponding range.
  • physiological/physical parameters refer to relevant parameters describing the molecular exchange system.
  • the molecular exchange system is a dual-chamber exchange system or a multi-chamber exchange system;
  • the magnetic resonance measurement method is a diffusion exchange spectrum DEXSY sequence or a relaxation exchange spectrum REXSY sequence;
  • the physiological parameters/molecular exchange related parameters are: fast component diffusion rate D f , slow component diffusion rate D s , fast component to slow component exchange rate k fs , fast component proportion coefficient in equilibrium state and the signal S 0 without diffusion coding gradient.
  • the relaxation exchange spectrum REXSY sequence is very similar to the diffusion exchange spectrum DEXSY sequence, and the constructed unified network model is also similar, the diffusion exchange spectrum DEXSY sequence will be mainly used as an example for explanation in the present invention.
  • step (3) the sampling trajectory optimization network simulates the process of generating signals by the molecular exchange magnetic resonance measurement method DEXSY sequence, and generates magnetic resonance signals according to the corresponding physiological/physical parameters, sampling trajectory and acquisition parameters, and the expression is as follows:
  • O D1 and O D2 represent the signal evolution process in the first and second diffusion coding modules of the DEXSY sequence, and the expressions are as follows:
  • b1 and b2 represent the b-value in the first and second diffusion coding modules respectively.
  • the number of sampling points b1 and b2 is set according to the number of sampling trajectory points under a single exchange time in step (2).
  • the numerical values of the sampling points b1 and b2 are the parameters related to the sampling trajectory to be optimized and are set as trainable parameters.
  • tm is the exchange time
  • switching matrix is the longitudinal relaxation matrix
  • T 1,s(f) is the longitudinal relaxation time
  • step (3) in the sampling trajectory optimization network, b1 and b2 satisfy the following conditions:
  • the value range of the variable v is [- ⁇ ,+ ⁇ ], ⁇ represents the sigmoid activation function, b max and b min represent the distribution constraints of the sampling trajectory, which are the maximum and minimum values of the value range of b 1 and b 2 respectively.
  • step (3) the noise signal generation network simulates a real magnetic resonance scanning scene, adds independent and identically distributed Gaussian noise to the real and imaginary parts of the signal, and obtains a magnetic resonance signal with Rice noise, the formula is as follows:
  • n re and n im represent independent and identically distributed Gaussian noise added to the real and imaginary parts, respectively
  • SNR is the signal-to-noise ratio
  • S DEXSY represents the noise-free magnetic resonance signal output by the sampling mode optimization network in step (3)
  • step (3) the parameter estimation network is constructed using a fully connected layer, including an input layer, five hidden layers and an output layer; the number of neurons in the input layer is equal to the number of magnetic resonance signals, and the number of neurons in the output layer is equal to the number of physiological/physical parameters.
  • the physiological/physical parameters are output:
  • f DNN represents the parameter estimation network
  • ⁇ represents the network parameter to be optimized of f DNN .
  • the output layer neuron value is not directly a physiological/physical parameter, but a scale value of the physiological/physical parameter: in step (3), the output physiological/physical parameter to be estimated is obtained by scaling the maximum-minimum value, as shown below:
  • Output P represents the neuron value corresponding to parameter P in the output layer.
  • step (4) the optimization goal is to improve the accuracy of parameter estimation.
  • the training loss function is based on the traditional mean square error, which consists of the losses of all physiological/physical parameters:
  • P represents the true value of the physiological parameter
  • N represents the number of physiological parameters P.
  • step (4) after the unified network model is trained, the obtained optimized sampling trajectory is discretized into the b1 - b2 sampling space, and b1 and b2 in the sampling trajectory optimization network are set according to the discretized sampling trajectory and fixed as non-trainable parameters, and then the parameter estimation network is trained until the network converges and the training is completed, thereby obtaining the optimal parameter estimation network corresponding to the discretized sampling trajectory.
  • the method further comprises: in step (2), adjusting the signal-to-noise ratio of the signal acquired by a single magnetic resonance in the acquisition parameters, setting the Gaussian noise standard deviation in the noise signal generating network described in step (3), and performing step (4) to train a unified network model to obtain optimized sampling trajectories and corresponding parameter estimation networks under different signal-to-noise ratios.
  • the method further comprises: in step (2), adjusting the number of sampling trajectory points under a single exchange time in the acquisition parameters, setting the number of sampling points in the b1 - b2 space in the sampling trajectory optimization network described in step (3), and performing step (4) to train a unified network model to obtain optimized sampling trajectories and corresponding parameter estimation networks under different sampling ratios.
  • the method for optimizing the molecular exchange magnetic resonance measurement sampling trajectory provided by the present invention is based on a data-driven approach, trains a unified network model under specified experimental settings, obtains an optimized sampling trajectory and a corresponding parameter estimation network, and analyzes the distribution characteristics of the optimal sampling trajectory under different numbers of sampling trajectory points and signal-to-noise ratios.
  • the technology has certain versatility and can be used to optimize other molecular exchange magnetic resonance measurement sampling trajectories.
  • the present invention also provides a molecular exchange magnetic resonance measurement method, the method comprising:
  • the sampling trajectory optimized by the above method for optimizing the sampling trajectory of molecular exchange magnetic resonance measurement is set as the sampling trajectory in the molecular exchange magnetic resonance measurement, and the magnetic resonance signal is collected;
  • the collected signal is used as input, and the physiological/physical parameter to be measured is output by the network based on the corresponding parameter estimation in the above-mentioned method for optimizing the molecular exchange magnetic resonance measurement sampling trajectory.
  • the present invention also provides a molecular exchange magnetic resonance measurement device, which includes a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, they are used to implement the above-mentioned molecular exchange magnetic resonance measurement method.
  • the processor must be equipped with Intel(R) Xeon(R) CPU@3.20GHz, 256GB memory, and an Nvidia Tesla P4 (8GB) GPU.
  • the present invention has the following technical effects:
  • FIG1 is a flow chart showing the method for optimizing the molecular exchange magnetic resonance measurement sampling trajectory in an embodiment and applying it to the molecular exchange magnetic resonance measurement, which mainly includes the following steps:
  • Independent uniform random sampling was performed according to the specified range of each physiological parameter to generate a simulation data set, from which 1,000,000 groups were selected as training sets, 10,000 groups as validation sets, and 10,000 groups as test sets.
  • the DEXSY sequence is used as the molecular exchange magnetic resonance measurement method
  • the DEXSY sequence is used as the molecular exchange magnetic resonance measurement method, and the magnetic resonance signal generation expression is as follows:
  • b1 and b2 represent the b-value in the first and second diffusion coding modules respectively.
  • the number of sampling trajectory points in the b1 - b2 space is fixed to 12, indicating that the numerical values of b1 and b2 of the sampling trajectory are parameters to be optimized.
  • this embodiment defines b1 and b2 in the sampling trajectory optimization network to meet the following conditions:
  • the relaxation times of the fast and slow components are set to 1.8 s and 0.9 s respectively.
  • n re and n im represent independent and identically distributed Gaussian noise added to the real and imaginary parts, respectively.
  • SNR is the signal-to-noise ratio (SNR), which is fixed to 40 according to the experimental setting in step (2).
  • S DEXSY represents the noise-free magnetic resonance signal output by the sampling trajectory optimization network in step (3). Represents a magnetic resonance signal with Rice noise.
  • f DNN represents the parameter estimation network
  • ⁇ represents the network parameter to be optimized of f DNN .
  • this embodiment only uses a fully connected layer to construct a parameter estimation network, which includes an input layer, five hidden layers and an output layer.
  • the number of neurons in the input layer is equal to the number of collected signals, which is equal to the product of the number of exchange times and the number of sampling trajectory points under a single exchange time in this embodiment.
  • the number of exchange times is 4, and the number of sampling trajectory points under a single exchange time is 12, then the number of neurons in the input layer is set to 48.
  • the number of neurons in the five hidden layers is set to 128, 256, 512, 256 and 128, respectively, and each layer is followed by an exponential linear unit (ELU) activation function.
  • the number of neurons in the output layer is equal to the number of molecular exchange-related parameters/physiological parameters to be estimated.
  • it is set to 5 in this embodiment, corresponding to the estimated physiological/physical parameters and
  • P represents the true value of the physiological parameter
  • P max and P min represent the upper and lower boundaries of the physiological parameter range, respectively
  • N represents the number of physiological parameters P.
  • the sampling trajectory optimized in step (6) is discretized into the b1 - b2 sampling space to meet the sampling interval constraint, and b1 and b2 in the sampling trajectory optimization network are set according to the discretized sampling trajectory and fixed as non-trainable parameters.
  • the unified network model is trained based on the simulation data set in step (1) until the network converges, and the optimal parameter estimation network corresponding to the discretized sampling trajectory is obtained.
  • the four b1 - b2 space sampling trajectories proposed so far are compared: FEXSY, DEWI, ANTI and UNIFORM (as shown in (a) in Figure 4).
  • b1 and b2 in the sampling trajectory optimization network are set according to different sampling trajectories and fixed as non-trainable parameters.
  • the corresponding parameter estimation network is trained under the same experimental parameters.
  • the performance of the sampling trajectory is evaluated from the perspectives of accuracy and repeatability on the simulation data set. Besides, the repeatability of the sampled trajectories is evaluated on real data.
  • FIG7 shows the repeatability experimental results on simulated data and real data. It can be seen from FIG7 that the present invention exhibits the best parameter repeatability in both data sets, showing its stability, and also proving that the sampling trajectory optimized based on the present invention has a certain universality.
  • the present invention sets the SNR and the number of sampling trajectory points to fixed values in turn, and changes another parameter to analyze the relationship between the optimal sampling trajectory and the two.
  • the specific steps are as follows:
  • step (3) The number of sampling trajectory points in the b 1 -b 2 sampling space in the sampling trajectory optimization network in step (3) is fixed to 12, the SNR of the noise signal generation network in step (4) is set to 10, 40 and 400 respectively, and step (6) is repeated to train the network 100 times to obtain the optimized b 1 -b 2 space sampling trajectory under different signal-to-noise ratios.
  • Step (4) The SNR of the noise signal generation network in step (4) is fixed to 40, and the number of sampling trajectory points in the b1 - b2 sampling space in the sampling trajectory optimization network in step (3) is set to 3, 6, 12 and 18 respectively.
  • Step (6) is repeated to train the network 100 times to obtain the optimized b1 - b2 space sampling trajectories under different sampling ratios.
  • the network is repeatedly trained 100 times based on step (6). Because the training of deep neural networks is a random process, different network weights may be obtained for the same data. In order to obtain statistically significant results, the training is repeated 100 times for the same experimental parameter settings.
  • the distribution of the optimal sampling trajectory under different SNRs obtained in step (8) is highly regular.
  • the sampling trajectory tends to improve the signal-to-noise ratio by repeated sampling in important areas.
  • the sampling trajectory tends to obtain more information by uniform sampling in the b1 - b2 space.
  • step (9) The distribution of the optimal sampling trajectory under different sampling ratios obtained in step (9) shows that the importance of each sampling point in the b 1 -b 2 space is different.
  • the sampling trajectory mainly shows a distribution similar to ANTI.
  • the sampling trajectory gradually expands along the diagonal of the b1 - b2 sampling space and the optimization results under low sampling rate. From the above two results, it can be seen that it is necessary to design a corresponding optimal sampling trajectory for a specific experimental environment.
  • the device for optimizing the molecular exchange magnetic resonance measurement sampling trajectory includes one or more processors, and the memory stores executable code.
  • the processor executes the executable code, it is used to implement the method for optimizing the molecular exchange magnetic resonance measurement sampling trajectory in the above embodiment.
  • Taking software implementation as an example as a device in a logical sense, it is formed by the processor of any device with data processing capability in which it is located to read the corresponding computer program instructions in the non-volatile memory into the memory for execution.
  • any device with data processing capability in which the device in the embodiment is located can also include other hardware according to the actual function of the device with data processing capability, which will not be described in detail.

Landscapes

  • Physics & Mathematics (AREA)
  • High Energy & Nuclear Physics (AREA)
  • General Physics & Mathematics (AREA)
  • Chemical & Material Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Condensed Matter Physics & Semiconductors (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)

Abstract

本发明公开了一种优化分子交换磁共振测量采样轨迹的方法,所述方法包括:(1)确定分子交换系统的生理/物理参数数量和对应范围,生成仿真生理/物理参数数据集;(2)确定分子交换磁共振测量方法并设置采集参数;(3)构建端到端的统一网络模型,包括采样轨迹优化网络、生成噪声信号网络和参数估计网络;(4)将仿真生理/物理参数作为输入进行训练,直到网络收敛完成训练,得到优化后的采样轨迹和对应的参数估计网络。本发明还公开了一种分子交换磁共振测量方法及装置。本发明提供的优化分子交换磁共振测量采样轨迹的方法提高了生理/物理参数估计的效率和准确度,可应用于不同分子交换磁共振测量方法中。

Description

一种优化分子交换磁共振测量采样轨迹的方法、分子交换磁共振测量方法及装置 技术领域
本发明属于磁共振定量测量的技术领域,特别涉及一种优化分子交换磁共振测量采样轨迹的方法、分子交换磁共振测量方法及装置。
背景技术
多维核磁共振(nuclear magnetic resonance,NMR)波谱技术可以对复杂分子系统进行全面详细的分析,并且作为一种非侵入性技术,在生物研究领域广为青睐[Bai R,Benjamini D,Cheng J,et al.Fast,accurate 2D-MR relaxation exchange spectroscopy(REXSY):Beyond compressed sensing[J].The Journal of chemical physics,2016,145(15):154202]。然而,多维波谱技术的采集时间较长,这限制了它们在特定物理条件下分析快速动力学和研究人体/动物体内的应用。例如,扩散交换光谱(diffusion exchange spectroscopy,DEXSY)方法[Callaghan P T,Furo I.Diffusion-diffusion correlation and exchange as a signature for local order and dynamics[J].The Journal of chemical physics,2004,120(8):4032-4038],它根据不同微环境或室内分子表观扩散速率不同来测量分子交换过程,序列图如图2中的(a)所示,它由两个独立的扩散编码模块组成,并且被时间混合模块分隔开,是一个包含第一扩散模块-混合模块-第二扩散模块的三维采集序列。每个模块的对应的采集变量为扩散加权b1,交换/混合时间tm,以及扩散加权b2。对于每一个信号点,都需要在特定的第一和第二扩散编码以及混合时间下单独收集,因此DEXSY技术非常耗时。
实时测量分子交换在化学的动态过程研究和生物的活体研究中具有重要意义,目前有很多研究提出加速分子交换测量的方法。超快核磁共振(ultrafast NMR,UF NMR)光谱技术运用磁共振成像(magnetic resonance imaging,MRI)的基本原理,将增量演化时间以空间方式编码到样品各层中,从而实现在单次扫描内获得多维光谱[Mankinen O,Zhivonitko V V,Selent A,et al.Ultrafast diffusion exchange nuclear magnetic resonance[J].Nature Communications,2020,11(1):3251]。然而,UF NMR技术单次扫描获得的结果的灵敏度有限,无法同时实现扫描时间短和分辨率高[Giraudeau P,Akoka S.Sources of sensitivity losses in ultrafast 2D NMR[J].Journal of Magnetic Resonance,2008,192(1):151-158],因此,该技术在探索复杂系统中的应用受到阻碍。另一部分研究提出通过在b1-b2空间降采样来缩短扫描时间,例如10%的采样比例对应10倍加速。S.Ramadan提出扩散交换加权成像技术(Diffusion-Exchange Weighted Imaging,DEWI)[Ramadan S.Diffusion-exchange weighted imaging[J].Magnetic Resonance Insights,2009,3:MRI.S3504],该技术要求第一个扩散编模块和第二个扩散编码模块的b值(b-value)相等,例如b1=b2。Nilsson等人提出过滤交换光谱技术(filter-exchange spectroscopy,FEXSY)[Nilsson M,J,van Westen D,et al.Noninvasive mapping of water diffusional exchange in the human brain using filter-exchange imaging[J].Magnetic resonance in medicine,2013,69(6):1572-1580],该技术将第一个扩散编码模块的b-value设置为固定值,通过减少待设置扫描参数的个数来缩短扫描时间。Cai等人提出反对角线采样模式(antidiagonal strategy,ANTI)[Williamson N H,Ravin R,Cai T X,et al.Real-time measurement of diffusion exchange rate in biological tissue[J].Journal of Magnetic Resonance,2020,317:106782],该采样模式的特点是要求第一个扩散编模块和第二个扩散编码模块的b-value的和为定植。Ordinola等人提出均匀采样模式(uniform undersampling strategy,UNIFORM)[A.Ordinola,S.Cai,R.Bai,and E.the 15th International Bologna Conference on Magnetic Resonance in Porous Media,15,Hangzhou,China.(2022)],该采样模式提出在b1-b2空间的上三角区域进行均匀分布的采样。然而这些基于启发式方法获得的采样模式的最优性无法得到证明。
最近,深度学习的出现为采样模式优化提供了新的基于数据驱动的优化技术。在特定的实验条件下,已经有研究基于该技术成功实现了MRI中频域空间采样模式的优化[Xue S,Cheng Z,Han G,et al.2D probabilistic undersampling pattern optimization for MR image reconstruction[J].Medical Image Analysis,2022,77:102346],基于该技术重建的磁共振(magnetic resonance,MR)图像质量优于基于其他采样模式获得的图像。
因此,如何进一步提高参数估计的效率和准确度是目前本领域的研究热点。
发明内容
本发明的目的在于提供一种优化分子交换磁共振测量采样轨迹的方法,该方法提高了生理/物理参数估计的效率和准确度,可应用于不同分子交换磁共振测量方法中。
本发明提供如下技术方案:
一种优化分子交换磁共振测量采样轨迹的方法,所述方法包括:
(1)确定分子交换系统的生理/物理参数数量和对应范围,作为仿真参数,在指定范围内进行相互独立的均匀随机采样,生成仿真生理/物理参数数据集;
(2)确定分子交换磁共振测量方法并设置采集参数;
(3)构建端到端的统一网络模型,包括采样轨迹优化网络、生成噪声信号网络和参数估计网络:采样轨迹优化网络模拟分子交换磁共振测量方法生成信号的过程,根据对应的生理/物理参数、采样轨迹和采集参数生成磁共振信号;生成噪声信号网络模拟真实磁共振测量场景中的噪声分布,生成带噪声的磁共振信号;参数估计网络实现磁共振信号拟合功能,根据带噪声的磁共振信号输出生理/物理参数;
(4)基于步骤(3)中的端到端的统一网络模型,在指定采集参数下,将步骤(1)中的仿真生理/物理参数作为输入进行训练,直到网络收敛完成训练,得到优化后的采样轨迹和对应的参数估计网络。
本发明提供的优化分子交换磁共振测量采样轨迹的方法主要包括两个部分:模拟分子交换磁共振测量方法生成信号的过程的采样轨迹优化网络;实现磁共振信号拟合过程的参数估计网络;该两部分网络由生成噪声信号网络连接,从而将采样轨迹优化和参数估计结合到端到端的统一网络模型中,实现了以提高参数估计准确度为直接目标训练网络,得到优化的采样轨迹和参数估计网络。
在步骤(1)中,为了保证数据的多样性,以及方法的泛化性,所有的参数都是相互独立地在对应的范围内进行平均随机采样得到的。
在本发明中,生理/物理参数是指描述分子交换系统的相关参数。
所述分子交换系统为双室交换系统或多室交换系统;所述磁共振测量方法为扩散交换光谱DEXSY序列或弛豫交换光谱REXSY序列;在双室交换系统中,所述生理参数/分子交换相关参数为:快组分扩散速率Df,慢组分扩散速率Ds,快组分到慢组分的交换速率kfs,平衡状态下快组分占比系数和无扩散编码梯度下的信号S0。
由于弛豫交换光谱REXSY序列与扩散交换光谱DEXSY序列非常相似,构建的统一网络模型也类似,因此在本发明中将主要以扩散交换光谱DEXSY序列为例进行说明。
在步骤(2)中,所述采集参数包括交换时间数值及个数、单次交换时间下的采样轨迹点个数及分布约束、单次磁共振采集的信号信噪比。
在步骤(3)中,所述采样轨迹优化网络模拟了分子交换磁共振测量方法DEXSY序列生成信号的过程,根据对应的生理/物理参数、采样轨迹和采集参数生成磁共振信号,表达式如下:
SDEXSY=IOD2OEOD1S0
其中I=(1,1),
运算符OD1和OD2代表DEXSY序列的第一个和第二个扩散编码模块中信号演变过程,表达式如下:
其中是扩散矩阵,b1和b2分别代表第一个和第二个扩散编码模块中的b-value,根据步骤(2)中单次交换时间下的采样轨迹点个数设置采样点b1和b2的个数,采样点b1和b2的数值大小为待优化采样轨迹相关参数,设置为可训练参数。
运算符OE描述了DEXSY序列的混合模块中信号演变过程,表达式如下:
其中tm是交换时间,是交换矩阵, 是纵向弛豫矩阵,R1,s(f)=1/T1,s(f),T1,s(f)是纵向弛豫时间。
为了保证b1和b2在采样轨迹分布约束的范围内同时能够被充分优化,本发明在采样模式优化网络中进行下述定义:在步骤(3)中,在采样轨迹优化网络中,b1和b2满足提下条件:
b1(2)=(bmax-bmin)σ(v)+bmin
其中变量v的取值范围是[-∞,+∞],σ代表sigmoid激活函数,bmax和bmin表示采样轨迹的分布约束,分别为b1和b2的取值范围的最大值和最小值。
在步骤(3)中,所述生成噪声信号网络模拟真实磁共振扫描场景,在信号的实部和虚部分别添加独立同分布的高斯噪声,获得具有莱斯噪声的磁共振信号,公式如下:
其中nre和nim分别代表添加到实部和虚部的独立同分布的高斯噪声,该高斯噪声的标准差σ=S0/SNR,SNR是信噪比,SDEXSY代表步骤(3)中采样模式优化网络输出的无噪声磁共振信号,代表具有莱斯噪声的磁共振信号。
在步骤(3)中,所述参数估计网络采用全连接层构建,包括一个输入层、五个隐藏层和一个输出层;输入层神经元个数等于磁共振信号的个数,输出层的神经元个数等于生理/物理参数的个数,根据输入的具有莱斯噪声的磁共振信号,输出生理/物理参数:
其中和代表估计的生理参数,fDNN代表参数估计网络,θ表示fDNN的待优化网络参数。
优选地,为了加速网络收收敛,输出层神经元数值并不直接是生理/物理参数,而是作为生理/物理参数的尺度值:在步骤(3)中,所述输出的待估计的生理/物理参数是经过最大-最小值的缩放得到的,如下所示:
P=(Pmax-Pmin)*OutputP+Pmin
其中Pmax和Pmin分别代表生理/物理参数范围的上下边界,OutputP代表输出层中对应参数P的神经元数值。
在步骤(4)中,以提高参数估计的准确度为优化目标,训练的损失函数基于传统的均方误差,由所有生理/物理参数的损失组成:
其中单个生理参数的损失定义如下:
其中P代表生理参数的真实值,代表网络估计的生理参数,N代表生理参数P的个数。
优选地,在步骤(4)中,统一网络模型完成训练后,将得到的优化后的采样轨迹离散化到b1-b2采样空间中,并根据离散后的采样轨迹设置采样轨迹优化网络中的b1和b2,并固定为不可训练参数,然后训练参数估计网络,直到网络收敛完成训练,得到与离散后的采样轨迹对应的最优参数估计网络。
优选地,所述方法还包括:在步骤(2)中,调整采集参数中的单次磁共振采集的信号的信噪比,设置步骤(3)中所述的生成噪声信号网络中的高斯噪声标准差,进行步骤(4)训练统一网络模型,得到不同信噪比下优化后的采样轨迹和对应的参数估计网络。
优选地,所述方法还包括:在步骤(2)中,调整采集参数中的单次交换时间下的采样轨迹点个数,设置步骤(3)中所述的采样轨迹优化网络中b1-b2空间的采样点个数,进行步骤(4)训练统一网络模型,得到不同采样比例下优化后的采样轨迹和对应的参数估计网络。
本发明提供的优化分子交换磁共振测量采样轨迹的方法,基于数据驱动的方式,在指定实验设置下训练统一网络模型,获得优化的采样轨迹和对应的参数估计网络,并分析了不同的采样轨迹点个数和信噪比下最优采样轨迹的分布特性,该技术具有一定的通用性,可用于优化其他的分子交换磁共振测量采样轨迹。
本发明还提供了一种分子交换磁共振测量方法,所述方法为:
针对于待测量的分子交换系统,将上述优化分子交换磁共振测量采样轨迹的方法优化后的采样轨迹设置为分子交换磁共振测量中的采样轨迹,采集磁共振信号;
将采集到的信号作为输入,基于上述优化分子交换磁共振测量采样轨迹的方法中对应的参数估计网络输出待测量的生理/物理参数。
本发明还提供了一种分子交换磁共振测量装置,所述装置包括存储器和一个或多个处理器,所述存储器中存储有可执行代码,所述一个或多个处理器执行所述可执行代码时,用于实现上述的分子交换磁共振测量方法。
在上述装置中,所述处理器需配备Intel(R)Xeon(R)CPU@3.20GHz,256GB内存,一个Nvidia Tesla P4(8GB)GPU。
与现有技术相比,本发明具有以下技术效果:
在本发明提出的优化分子交换磁共振测量采样轨迹的方法中:模拟分子交换磁共振测量方法生成信号过程的采样轨迹优化网络通过将采样轨迹相关参数设置为可训练参数,克服了传统的优化方法中存在的待优化参数自由度高、模型复杂等问题;本发明提出的实现磁共振信号拟合过程的参数估计网络利用深度学习在求解逆问题上的优势,相比于传统的参数估计方法,提高了参数估计的效率和准确度;本发明提出的统一的网络模型实现了在特定的分子交换磁共振测量方法中,以参数估计准确度为直接目标训练网络,得到优化的采样轨迹和参数估计网络。
对比现有的采样轨迹和参数估计方法,在同等的采样比例下,基于本发明得到的采样轨迹和参数估计方法能够提供最可靠的参数。因此,本发明将其应用到分子交换磁共振测量方法中,可为实现不同微环境或室间分子交换的实时监测和定量测量提供了一种高效、准确的方法,并且该方法具有一定的通用性,可适应多种分子交换磁共振测量方法中。
附图说明
图1为实施例中优化分子交换磁共振测量采样轨迹的方法应用到分子交换磁共振测量的流程图。
图2为本发明提供的一种统一网络模型的结构图:(a)实施例优化的DEXSY扫描序列;(b)本发明提出的网络结构图:输入层,DEXSY序列的采样模式优化网络,生成噪声信号网络,参数估计网络,输出层。
图3为基于细胞模型的双室交换系统。
图4为对比的b1-b2空间采样模式以及在不同的实验条件下基于本发明优化得到的b1-b2空间采样模式:(a)对比采样模式:FEXSY、DEWI、ANTI和UNIFORM;(b)b1-b2空间采样点数固定为12,SNR分别为10、40和400时,重复100次实验得到的优化结果;(c)SNR固定为40,b1-b2空间采样点数分别为3、6、12和18时,重复100次实验得到的优化结果。
图5为SNR为40,b1-b2空间采样点数为12时优化得到的一组采样模式。
图6为在不同的SNR和参数估计方法下,基于不同的采样模式估计的参数误差结果。
图7为不同的采样模式估计参数的可重复性性能:(a)基于仿真数据的参数估计可重复性性能;(b)基于真实数据的参数估计可重复性性能。
具体实施方式
为使本发明的目的、技术方案及优点更加清楚明白,以下结合附图及实施例对本发明进行进一步的详细说明。应当理解,此处所描述的具体实施方式仅仅用以解释本发明,并不限定本发明的保护范围。
本发明提出的模型基于Pytorch框架实现。
图1为实施例中优化分子交换磁共振测量采样轨迹的方法应用到分子交换磁共振测量的流程图,主要包括以下步骤:
(1)针对于分子交换磁共振测量的采样轨迹优化,确定待测量分子交换系统的生理/物理参数和对应范围,作为仿真参数,在指定范围内进行相互独立的均匀随机采样,生成一定数量的仿真生理/物理参数,并随机划分为训练集、验证集和测试集三个部分。
本实施例以双室交换系统为实验模型,如图3所示,包含快慢(fast,f;slow,s)两个组分,该系统中待测量的生理参数/分子交换相关参数以及其合理的范围分别是:快组分扩散速率Df∈[1.5,2.0]×10-3mm2/s,慢组分扩散速率Ds∈[0.03,0.06]×10-3mm2/s,快组分到慢组分的交换速率kfs∈[0,10]s-1平衡状态下快组分占比系数无扩散编码梯度下的信号S0∈[500,1500]。
根据各个生理参数的指定范围进行相互独立的均匀随机采样生成仿真数据集,从其中选取1000000组作为训练集,10000组作为验证集,10000组作为测试集。
(2)针对分子交换磁共振测量的采样轨迹优化,确定分子交换磁共振测量实验中相关实验设置(或表述为,确定分子交换磁共振测量方法并设置采集参数),包括交换时间数值及个数、单次交换时间下的采样轨迹点个数及分布约束、单次磁共振采集的信号信噪比。
本实施例以DEXSY序列为分子交换磁共振测量方法,采集参数包括:交换时间tm固定为[10,100,200,350]ms,单次交换时间下的采样轨迹点个数为12,采样空间为b1-b2空间,满足bmin≤b1,b2≤bmax,其中bmin=100s/mm2,bmax=1200s/mm2,离散b1-b2空间的采样间隔为100s/mm2,单次磁共振采集的信号信噪比为40。
(3)构建采样轨迹优化网络,模拟分子交换磁共振测量方法生成信号的过程,根据对应的生理/物理参数、采样轨迹和实验条件(采集参数)生成磁共振信号。
本实施例以DEXSY序列为分子交换磁共振测量方法,磁共振信号生成表达式如下:
SDEXSY=IOD2OEOD1S0
其中I=(1,1),
运算符OD1和OD2描述了DEXSY序列的第一个和第二个扩散编码模块中信号演变过程,表达式如下:
其中是扩散矩阵,b1和b2分别代表第一个和第二个扩散编码模块中的b-value,根据步骤(2)中的实验设置,b1-b2空间的采样轨迹点个数固定为12,表示采样轨迹的b1和b2的数值大小为待优化参数,为了保证b1和b2在采样轨迹分布约束的范围内能够被充分优化,本实施例在采样轨迹优化网络中定义b1和b2满足以下条件:
b1(2)=(bmax-bmin)σ(v)+bmin
其中变量v的取值范围是[-∞,+∞],σ代表sigmoid激活函数,根据步骤(2)中的实验设置,bmax=1200s/mm2,bmin=100s/mm2。
运算符OE描述了混合模块中信号演变过程,表达式如下:
其中tm是交换时间,根据步骤(2)中的实验设置固定为[10,100,200,350]ms,是交换矩阵,是纵向弛豫矩阵,R1,s(f)=1/T1,s(f),T1,s(f)是纵向弛豫时间,在本实施例中快慢组分的弛豫时间分别设置为1.8s和0.9s。
(4)构建生成噪声信号网络,模拟真实磁共振实验场景中的噪声分布,将步骤(3)中采样轨迹优化网络输出的磁共振信号作为输入,在信号的实部和虚部分别添加独立同分布的高斯噪声,输出具有莱斯噪声的信号:
其中nre和nim分别代表添加到实部和虚部的独立同分布的高斯噪声,该高斯噪声的标准差σ=S0/SNR,SNR是信噪比(Signal-to-Noise Ratio,SNR),根据步骤(2)中的实验设置固定为40,SDEXSY代表步骤(3)中采样轨迹优化网络输出的无噪声磁共振信号,代表具有莱斯噪声的磁共振信号。
(5)构建参数估计网络,实现磁共振信号拟合功能,将步骤(4)中生成的具有莱斯噪声的信号作为该网络的输入,输出待估计的生理/物理参数值:
其中和代表估计的生理/物理参数,fDNN代表参数估计网络,θ表示fDNN的待优化网络参数。
具体地,本实施例仅采用全连接层构建参数估计网络,该网络包括一个输入层、五个隐藏层和一个输出层。输入层神经元个数等于采集信号个数,在本实施例中等于交换时间个数和单次交换时间下的采样轨迹点个数的乘积,根据步骤(2)中的实验设置,交换时间个数为4,单次交换时间下的采样轨迹点个数为12,则输入层神经元个数设置为48。五个隐藏层的神经元个数分别设置为128、256、512、256和128,每一层后面接着指数线性单元(Exponential Linear Unit,ELU)激活函数。输出层的神经元个数等于待估计的分子交换相关参数/生理参数的个数,根据步骤(1)中的实验模型描述在本实施例中设置为5,分别对应估计的生理/物理参数和
(6)将步骤(3)-(5)中的三部分网络整合成端到端的统一网络,在指定采集参数下,基于步骤(1)获得的仿真数据集对网络进行训练,直到网络收敛,得到优化的采样轨迹。
编写计算机程序,实现步骤(3)、(4)、(5)中的三个网络,构建端到端的统一网络模型,如图2中的(b)所示,实现在特定的分子交换磁共振测量方法中,以参数估计准确度为直接目标训练网络,得到优化的b1-b2空间采样轨迹和参数估计网络。训练的损失函数基于参数估计的均方误差,由所有生理参数的损失组成:
其中单个生理参数的损失定义如下:
其中P代表生理参数的真实值,代表网络估计的生理参数,Pmax和Pmin分别代表生理参数范围的上下边界,N代表生理参数P的个数。
(7)根据步骤(2)中采样轨迹的分布约束,将步骤(6)优化的采样轨迹离散化到b1-b2采样空间中以符合采样间隔约束,并根据离散化后的采样轨迹设置采样轨迹优化网络中的b1和b2,并固定为不可训练参数,基于步骤(1)中的仿真数据集对统一网络模型进行训练,直到网络收敛,得到与离散后的采样轨迹对应的最优参数估计网络。
为了证明基于本发明提出的数据驱动方式优化得到的采样轨迹的优越性能,对比了目前提出的四种b1-b2空间采样轨迹:FEXSY、DEWI、ANTI和UNIFORM(如图4中的(a)所示),为了保证对比的公平性,根据不同的采样轨迹设置对应的采样轨迹优化网络中的b1和b2并固定为不可训练参数,在同样的实验参数下训练得到相应的参数估计网络。并在仿真数据集上从准确度和可重复性两个角度评估采样轨迹的性能, 除此之外,在真实数据上评估了采样轨迹的可重复性。
如图6所示,展示了在不同信噪比和不同的参数估计方法下,仿真数据集上的参数估计准确度测试结果,从结果上看,本发明提出的采样轨迹能够显著提高参数估计的准确性,无论是基于参数估计网络还是传统的非线性最小二乘法进行模型拟合。
图7展示了在仿真数据和真实数据上的可重复性实验结果,从图7中可以看出来在两个数据集中本发明都展现出最优的参数可重复性,展现了其稳定性,也证明了基于本发明优化的采样轨迹具有一定的普适性。
本发明为了定性分析最优采样轨迹与单次磁共振采集的信号信噪比和单次交换时间下的采样轨迹点个数的关系,依次将SNR和采样轨迹点个数设置为定值,变化另一个参数分析最优采样轨迹与两者之间的关系,具体步骤为:
(8)固定步骤(3)中采样轨迹优化网络中b1-b2采样空间中的采样轨迹点个数为12,分别将步骤(4)中生成噪声信号网络的SNR设置为10、40和400,重复步骤(6)训练网络100次,得到不同信噪比下优化的b1-b2空间采样轨迹。
(9)固定步骤(4)中生成噪声信号网络的SNR为40,分别将步骤(3)中采样轨迹优化网络中b1-b2采样空间中的采样轨迹点个数设置为3、6、12和18,重复步骤(6)训练网络100次,得到不同采样比例下优化的b1-b2空间采样轨迹。
(9)定性分析不同的采样比例和信噪比对应的最优采样轨迹的规律及特点。
对于特定的SNR和采样比例,基于步骤(6)对网络重复训练100次,因为深度神经网络的训练是一个随机过程,对于同样的数据可能会获得不同的网络权重,为了获得具有统计意义的结果,对于同样的实验参数设置,重复训练100次。
步骤(8)中获得的不同SNR下最优采样轨迹的分布情况具有高度的规律性,当SNR较低时,采样轨迹倾向于在重要区域通过重复采样提高信噪比,当SNR比较高时,采样轨迹倾向于在b1-b2空间中进行均匀采样来获得更多的信息。
步骤(9)中获得的不同采样比例下最优采样轨迹的分布情况显示了b1-b2空间每个采样点的重要性不同。
结果分别如图4中的(b)和(c)所示。从结果图4中的(b)中可以看出来,采样轨迹点个数固定时,最优采样轨迹受到SNR的影响,当SNR比较低时,采样轨迹趋向于在某些重要区域重复采样,以此减弱噪声对信号的影响,当SNR比较高时,最优采样轨迹趋向于均匀分布,以此增加信号的多样性,但是在高b-value区域采样密度依然比较大,因为高b-value区域受噪声影响相比于低b-value区域较明显。从结果图4中的(c)中可以看出来,SNR固定时,最优采样轨迹受到采样率的影响,当采样率比较低时,采样轨迹主要展现类似于ANTI的分布,当采样率逐渐升高时,采样轨迹逐渐沿着b1-b2采样空间的对角线以及低采样率下的优化结果扩张。从以上两个结果可以看出,对于具体的实验环境有必要设计一个对应的最优采样轨迹。
本实施例提供的优化分子交换磁共振测量采样轨迹的装置,包括一个或多个处理器,存储器中存储有可执行代码,处理器执行可执行代码时,用于实现上述实施例中的优化分子交换磁共振测量采样轨迹的方法。以软件实现为例,作为一个逻辑意义上的装置,是通过其所在任意具备数据处理能力的设备的处理器将非易失性存储器中对应的计算机程序指令读取到内存中运行形成的从硬件层面而言,除处理器、内存、网络接口、以及非易失性存储器之外,实施例中装置所在的任意具备数据处理能力的设备通常根据该任意具备数据处理能力的设备的实际功能,还可以包括其他硬件,对此不再赘述。
以上,仅为本申请较佳的具体实施方式,但本申请的保护范围并不局限于此,任何熟悉本技术领域的技术人员在本申请揭露的技术范围内,可轻易想到的变化或替换,都应涵盖在本申请的保护范围之内。因此,本申请的保护范围应该以权利要求的保护范围为准。

Claims (13)

  1. 一种优化分子交换磁共振测量采样轨迹的方法,其特征在于,所述方法包括:
    (1)确定分子交换系统的生理/物理参数数量和对应范围,作为仿真参数,在指定范围内进行相互独立的均匀随机采样,生成仿真生理/物理参数数据集;
    (2)确定分子交换磁共振测量方法并设置采集参数;
    (3)构建端到端的统一网络模型,包括采样轨迹优化网络、生成噪声信号网络和参数估计网络:采样轨迹优化网络模拟分子交换磁共振测量方法生成信号的过程,根据对应的生理/物理参数、采样轨迹和采集参数生成磁共振信号;生成噪声信号网络模拟真实磁共振测量场景中的噪声分布,生成带噪声的磁共振信号;参数估计网络实现磁共振信号拟合功能,根据带噪声的磁共振信号输出生理/物理参数;
    (4)基于步骤(3)中的端到端的统一网络模型,在指定采集参数下,将步骤(1)中的仿真生理/物理参数作为输入进行训练,直到网络收敛完成训练,得到优化后的采样轨迹和对应的参数估计网络。
  2. 根据权利要求1所述的优化分子交换磁共振测量采样轨迹的方法,其特征在于,所述分子交换系统为双室交换系统或多室交换系统;所述磁共振测量方法为扩散交换光谱DEXSY序列或弛豫交换光谱REXSY序列;在双室交换系统中,所述生理/物理参数为:快组分扩散速率Df,慢组分扩散速率Ds,快组分到慢组分的交换速率kfs,平衡状态下快组分占比系数和无扩散编码梯度下的信号S0。
  3. 根据权利要求1所述的优化分子交换磁共振测量采样轨迹的方法,其特征在于,在步骤(2)中,所述采集参数包括交换时间数值及个数、单次交换时间下的采样轨迹点个数及分布约束、单次磁共振采集的信号信噪比。
  4. 根据权利要求3所述的优化分子交换磁共振测量采样轨迹的方法,其特征在于,在步骤(3)中,所述采样轨迹优化网络模拟分子交换磁共振测量方法DEXSY序列生成信号的过程, 根据对应的生理/物理参数、采样轨迹和采集参数生成磁共振信号,表达式如下:
    SDEXSY=IOD2OEOD1S0
    其中I=(1,1),
    运算符OD1和OD2代表DEXSY序列的第一个和第二个扩散编码模块中信号演变过程,表达式如下:
    其中是扩散矩阵,b1和b2分别代表第一个和第二个扩散编码模块中的b-value,根据步骤(2)中单次交换时间下的采样轨迹点个数设置采样点b1和b2的个数,采样点b1和b2的数值大小为待优化采样轨迹相关参数,因此设置为可训练参数并满足以下条件:
    b1(2)=(bmax-bmin)σ(v)+bmin
    其中变量v的取值范围是[-∞,+∞],σ代表sigmoid激活函数,bmax和bmin表示采样轨迹的分布约束,分别为b1和b2的取值范围的最大值和最小值;
    运算符OE描述了DEXSY序列的交换模块中信号演变过程,表达式如下:
    其中tm是交换时间,是交换矩阵,是纵向弛豫矩阵,R1,s(f)=1/T1,s(f),T1,s(f)是纵向弛豫时间。
  5. 根据权利要求4所述的优化分子交换磁共振测量采样轨迹的方法,其特征在于,在步骤(3)中,所述生成噪声信号网络模拟真实磁共振扫描场景,在磁共振信号的实部和虚部分别添加独立同分布的高斯噪声,获得具有莱斯噪声的磁共振信号,公式如下:
    其中nre和nim分别代表添加到实部和虚部的独立同分布的高斯噪声,该高斯噪声的标准差σ=S0/SNR,SNR是信噪比,SDEXSY代表步骤(3)中采样轨迹优化网络输出的无噪声磁共振信号,代表具有莱斯噪声的磁共振信号。
  6. 根据权利要求5所述的优化分子交换磁共振测量采样轨迹的方法,其特征在于,在步骤(3)中,所述参数估计网络采用全连接层构建,包括一个输入层、五个隐藏层和一个输出层;输入层神经元个数等于磁共振信号的个数,输出层的神经元个数等于生理/物理参数的个数,根据输入的具有莱斯噪声的磁共振信号,输出生理/物理参数:
    其中和代表估计的生理参数,fDNN代表参数估计网络,θ表示fDNN的待优化网络参数。
  7. 根据权利要求1所述的优化分子交换磁共振测量采样轨迹的方法中的参数估计网络,其特征在于,在步骤(3)中,所述输出的生理/物理参数是经过最大-最小值的缩放得到的,如下所示:
    P=(Pmax-Pmin)*OutputP+Pmin
    其中Pmax和Pmin分别代表生理/物理参数范围的上下边界,OutputP代表输出层中对应参数P的神经元数值。
  8. 根据权利要求1所述的优化分子交换磁共振测量采样轨迹的方法,其特征在于,在步骤(4)中,以提高参数估计的准确度为优化目标,训练的损失函数基于传统的均方误差,由所有生理/物理参数的损失组成:
    其中单个生理参数的损失定义如下:
    其中P代表生理参数的真实值,代表网络估计的生理参数,N代表生理参数P的个数。
  9. 根据权利要求1所述的优化分子交换磁共振测量采样轨迹的方法,其特征在于,在步骤(4)中,统一网络模型完成训练后,将得到的优化后的采样轨迹离散化到b1-b2采样空间中,并根据离散后的采样轨迹设置采样轨迹优化网络中的b1和b2,并固定为不可训练参数,然后训练参数估计网络,直到网络收敛完成训练,得到与离散后的采样轨迹对应的参数估计网络。
  10. 根据权利要求1所述的优化分子交换磁共振测量采样轨迹的方法,其特征在于,所述方法还包括:在步骤(2)中,调整采集参数中的单次磁共振采集的信号信噪比,设置步骤(3)中生成噪声信号网络中的高斯噪声标准差,进行步骤(4)训练统一网络模型,得到不同信噪比下优化后的采样轨迹和对应的参数估计网络。
  11. 根据权利要求1所述的优化分子交换磁共振测量采样轨迹的方法,其特征在于,所述方法还包括:在步骤(2)中,调整采集参数中的单次交换时间下的采样轨迹点个数,设置步骤(3)中所述的采样轨迹优化网络中b1-b2空间的采样点个数,进行步骤(4)训练统一网络模型,得到不同采样比例下优化后的采样轨迹和对应的参数估计网络。
  12. 一种分子交换磁共振测量方法,其特征在于,所述方法为:
    针对于待测量的分子交换系统,将权利要求1-11任一所述的优化后的采样轨迹设置为分子交换磁共振测量中的采样轨迹,采集磁共振信号;
    将采集到的信号作为输入,基于权利要求1-11任一所述的对应的参数估计网络输出待测量的生理/物理参数。
  13. 一种分子交换磁共振测量装置,其特征在于,所述装置包括存储器和一个或多个处理器,所述存储器中存储有可执行代码,所述一个或多个处理器执行所述可执行代码时,用于实现权利要求12所述的分子交换磁共振测量方法。
PCT/CN2024/102846 2023-07-06 2024-07-01 一种优化分子交换磁共振测量采样轨迹的方法、分子交换磁共振测量方法及装置 Ceased WO2025007834A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2026500279A JP7875650B1 (ja) 2023-07-06 2024-07-01 分子交換磁気共鳴測定のサンプリング軌跡を最適化する方法、分子交換磁気共鳴測定方法及び装置

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202310824584.3 2023-07-06
CN202310824584.3A CN116930241B (zh) 2023-07-06 2023-07-06 一种优化分子交换磁共振测量采样轨迹的方法、分子交换磁共振测量方法及装置

Publications (1)

Publication Number Publication Date
WO2025007834A1 true WO2025007834A1 (zh) 2025-01-09

Family

ID=88393344

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2024/102846 Ceased WO2025007834A1 (zh) 2023-07-06 2024-07-01 一种优化分子交换磁共振测量采样轨迹的方法、分子交换磁共振测量方法及装置

Country Status (3)

Country Link
JP (1) JP7875650B1 (zh)
CN (1) CN116930241B (zh)
WO (1) WO2025007834A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116930241B (zh) * 2023-07-06 2025-05-16 浙江大学 一种优化分子交换磁共振测量采样轨迹的方法、分子交换磁共振测量方法及装置
CN120779309B (zh) * 2025-09-04 2025-11-11 复旦大学 时间依赖性扩散磁共振成像的参数优化方法、系统及设备

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102914753A (zh) * 2011-12-08 2013-02-06 中国科学院深圳先进技术研究院 磁共振成像采样轨迹优化方法
CN108132274A (zh) * 2017-12-21 2018-06-08 厦门大学 不均匀磁场下回波平面成像无参考扫描图像畸变矫正方法
CN111062994A (zh) * 2019-11-26 2020-04-24 厦门大学 一种化学交换饱和转移对比图像的重建方法及系统
US20200134887A1 (en) * 2018-10-25 2020-04-30 The Board Of Trustees Of The Leland Stanford Junior University Reinforcement Learning for Online Sampling Trajectory Optimization for Magnetic Resonance Imaging
CN114544689A (zh) * 2022-01-13 2022-05-27 浙江大学 测量水分子跨细胞膜流出速率的方法及应用、胶质瘤磁共振影像学标志物的测量方法及系统
CN114820849A (zh) * 2022-04-19 2022-07-29 浙江大学 基于深度学习的磁共振cest图像重建方法、装置及设备
CN115063499A (zh) * 2022-05-26 2022-09-16 天津大学 面向磁共振成像放射状采样轨迹的切片自适应确定方法
CN116930241A (zh) * 2023-07-06 2023-10-24 浙江大学 一种优化分子交换磁共振测量采样轨迹的方法、分子交换磁共振测量方法及装置

Family Cites Families (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
SE0950363A1 (sv) * 2009-05-22 2010-06-29 Cr Dev Ab Metod och system för magnetisk resonanstomografi, samt användning därav.
WO2015179049A1 (en) * 2014-05-19 2015-11-26 The United States Of America, As Represented By The Secretary, Department Of Health And Human Services Magnetic resonance 2d relaxometry reconstruction using partial data
US11415652B2 (en) * 2016-08-11 2022-08-16 The United States Of America, As Represented By The Secretary, Department Of Health And Human Services Multi-dimensional spectroscopic NMR and MRI using marginal distributions
GB201913481D0 (en) * 2019-09-18 2019-10-30 Institute Of Cancer Res Royal Cancer Hospital Diffusion-weighted magnetic resonance imaging
US12044762B2 (en) * 2020-02-03 2024-07-23 The General Hospital Corporation Estimating diffusion metrics from diffusion- weighted magnetic resonance images using optimized k-q space sampling and deep learning
CN111751770A (zh) * 2020-06-12 2020-10-09 浙江大学 一种非磁共振造影剂依赖的定量测量血管内外水分子交换速度的方法
CN112461881B (zh) * 2020-10-16 2022-02-18 北京大学 一种检测rna中弱稳定性碱基对的固体核磁共振方法
CN113379867B (zh) * 2021-07-05 2023-09-12 北京大学深圳研究生院 一种基于联合优化采样矩阵的核磁共振图像重建方法
CN113743596B (zh) * 2021-11-03 2022-02-18 中国科学院精密测量科学与技术创新研究院 一种基于无监督神经网络的nmr弛豫时间反演方法
US20250044246A1 (en) * 2021-11-10 2025-02-06 The United States Of America, As Represented By The Secretary, Department Of Health And Human Nuclear magnetic resonance methods of determining homeostatic perturbations
CN114155225B (zh) * 2021-12-07 2023-06-06 浙江大学 一种定量测量脑白质髓鞘内外水分子交换速率的方法
CN114581550B (zh) * 2021-12-31 2023-04-07 浙江大学 基于跨域网络的磁共振成像降采样和重建方法
CN115329802B (zh) * 2022-07-13 2025-10-31 厦门大学 一种基于深度学习的核磁共振信号弛豫时间分布计算方法

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102914753A (zh) * 2011-12-08 2013-02-06 中国科学院深圳先进技术研究院 磁共振成像采样轨迹优化方法
CN108132274A (zh) * 2017-12-21 2018-06-08 厦门大学 不均匀磁场下回波平面成像无参考扫描图像畸变矫正方法
US20200134887A1 (en) * 2018-10-25 2020-04-30 The Board Of Trustees Of The Leland Stanford Junior University Reinforcement Learning for Online Sampling Trajectory Optimization for Magnetic Resonance Imaging
CN111062994A (zh) * 2019-11-26 2020-04-24 厦门大学 一种化学交换饱和转移对比图像的重建方法及系统
CN114544689A (zh) * 2022-01-13 2022-05-27 浙江大学 测量水分子跨细胞膜流出速率的方法及应用、胶质瘤磁共振影像学标志物的测量方法及系统
CN114820849A (zh) * 2022-04-19 2022-07-29 浙江大学 基于深度学习的磁共振cest图像重建方法、装置及设备
CN115063499A (zh) * 2022-05-26 2022-09-16 天津大学 面向磁共振成像放射状采样轨迹的切片自适应确定方法
CN116930241A (zh) * 2023-07-06 2023-10-24 浙江大学 一种优化分子交换磁共振测量采样轨迹的方法、分子交换磁共振测量方法及装置

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
XIAO, TAOHUI ET AL.: "Fast Magnetic Resonance Imaging with Deep Learning and Design of Undersampling Trajectory", JOURNAL OF IMAGE AND GRAPHICS, vol. 23, no. 2, 16 February 2018 (2018-02-16), pages 194 - 208, XP055822195, ISSN: 1006-8961, DOI: 10.11834/jig.170274 *

Also Published As

Publication number Publication date
JP7875650B1 (ja) 2026-06-18
JP2026521969A (ja) 2026-07-02
CN116930241B (zh) 2025-05-16
CN116930241A (zh) 2023-10-24

Similar Documents

Publication Publication Date Title
KR102622283B1 (ko) 맥스웰 병렬 이미징
US12007455B2 (en) Tensor field mapping with magnetostatic constraint
Koay et al. A signal transformational framework for breaking the noise floor and its applications in MRI
US20210134028A1 (en) Method and system for simultaneous quantitative multiparametric magnetic resonance imaging (mri)
CN111870245B (zh) 一种跨对比度引导的超快速核磁共振成像深度学习方法
JP6884344B2 (ja) 脳内ネットワークの活動推定システム、脳内ネットワークの活動推定方法、脳内ネットワークの活動推定プログラム、および、学習済み脳活動推定モデル
JP7595379B2 (ja) 外挿磁化移動信号に基づくcestデータフィッティング方法、装置及び媒体
CN116930241B (zh) 一种优化分子交换磁共振测量采样轨迹的方法、分子交换磁共振测量方法及装置
CN105147288B (zh) 脑磁源强度定位方法
CN113743596B (zh) 一种基于无监督神经网络的nmr弛豫时间反演方法
CN111062994B (zh) 一种化学交换饱和转移对比图像的重建方法及系统
CN115137347B (zh) 三维超短回波时间磁共振指纹成像的髓鞘定量成像方法
CN110664378B (zh) 磁共振成像方法、装置、系统及存储介质
CN103325119A (zh) 一种基于模态融合的默认态脑网络中心节点检测方法
CN111898734A (zh) 一种基于mlp的nmr弛豫时间反演方法
Tu et al. K-space and image domain collaborative energy-based model for parallel MRI reconstruction
CN111445546A (zh) 图像重建方法、装置、电子设备、存储介质
Xiao et al. Deep learning for dense Z-spectra reconstruction from CEST images at sparse frequency offsets
CN114742116A (zh) 一种模拟采集脑电信号的生成方法和系统
CN118230911A (zh) 一种基于深度学习的脑磁图磁源定位方法、电子设备及存储介质
US20260140092A1 (en) Method and apparatus for measuring ultrasonic viscoelasticity, device and medium
CN119762608B (zh) 人体5.0t多核31p磁共振波谱成像的自监督重建方法
CN109118526A (zh) 一种基于虚拟现实的老年痴呆图像分析系统及分析方法
CN120405538A (zh) 用于多尺度脑组织微环境测量的弥散磁共振成像方法及系统
CN118299057A (zh) 一种基于物理信息神经网络的房室模型参数估计方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 24835315

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2026500279

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 24835315

Country of ref document: EP

Kind code of ref document: A1