磁共振成像设备的数字孪生建模关键技术与应用.pdf

系统仿真学报系统仿真学报Journal of System Simulation第 35 卷第 10 期2023 年 10 月Vol.35 No.10Oct.2023Key Technology and Application of Digital Twin Modeling for MRIChen Shanshan1，Wang Hongzhi2*，Xia Tian3(1.College of Medical Imaging,Shanghai University of Medicine&Health Science,Shanghai 201318,China;2.Shanghai Key Laboratory of Magnetic Resonance,East China Normal University,Shanghai 200062,China;3.Shanghai Training Cloud Education Technol.Co.,Ltd.,Shanghai 200433,China)Abstract：With the accelerating digitalization in education,the construction of digital resources and application platforms has caught increasing attention.The framework of MRI equipment digital twin five-dimensional model is constructed to solve the problems in teaching and training for magnetic resonance imaging(MRI).A modeling and simulation method based on the mechanism model is proposed.The multi-dimensional physical data are obtained to perform digital human modeling,and the virtual acquisition and image reconstruction method is proposed to generate images.The digital twin data are adopted for iterative optimization to implement the whole process of the three-dimensional visual operation including preparation before inspection,coil selection,patient positioning,parameter setting,and image processing.Application verification is carried out based on the imaging sequences of spin echo,gradient echo,and echo planar.Driven by real-time data,the simulation results are consistent with the operation data of the physical equipment.This meets the needs of teaching training and auxiliary design and guarantees personalized and intelligent adaptive learning under massive teaching arrangements.Keywords：digital twin;magnetic resonance imaging;sequence;mechanism modeling磁共振成像设备的数字孪生建模关键技术与应用磁共振成像设备的数字孪生建模关键技术与应用陈珊珊1，汪红志2*，夏天3(1.上海健康医学院 医学影像学院，上海 201318；2.华东师范大学 上海市磁共振重点实验室，上海 200062；3.上海培云教育科技有限公司，上海 200433)摘要摘要：随着教育数字化的加快推进，数字化资源及其应用平台搭建日益受到重视。针对磁共振教学培训中存在的难题，基于数字孪生五维模型概念，构建磁共振成像设备数字孪生五维模型框架。提出一种基于机理模型的建模仿真方法。使用物理实体获取多维物理数据完成数字人建模，提出虚拟采集和图像重建方法生成图像，并利用融合数据进行迭代优化，实现了检查前准备、线圈选择、体位摆位、参数设置、图像处理等全流程三维可视化操作。基于自旋回波、梯度回波和平面回波成像序列开展了应用验证，在实时数据驱动下，仿真结果与实体设备运行结果相一致，满足了教学培训和辅助设计需求，为规模化教学的个性化学习和智能自适应学习提供了资源保障。关键词关键词：数字孪生；磁共振成像；序列；机理建模Received date:2023-07-02 Revised date:2023-08-21Foundation:Research Fund of Shanghai University of Medicine&Health Sciences(E4-6101-15-002)First author:Chen Shanshan(1986-),female,lecture,PhD,research area:MRI technology.E-mail:Corresponding author:Wang Hongzhi(1975-),male,associate professor,PhD,research area:MRI technology,numerical simulation technology.E-mail:第 35 卷第 10 期2023 年 10 月Vol.35 No.10Oct.2023Chen Shanshan,et al:Key Technology and Application of Digital Twin Modeling for MRIhttp:/www.china-中图分类号：TP391.9 文献标志码：A 文章编号：1004-731X(2023)10-2122-11DOI:10.16182/j.issn1004731x.joss.23-FZ0799E引用格式引用格式:陈珊珊,汪红志,夏天.磁共振成像设备的数字孪生建模关键技术与应用J.系统仿真学报,2023,35(10):2122-2132.Reference format:Chen Shanshan,Wang Hongzhi,Xia Tian.Key Technology and Application of Digital Twin Modeling for MRIJ.Journal of System Simulation,2023,35(10):2122-2132.0IntroductionThe MRI equipment has been increasingly installed year by year during its wide applications in clinical diagnosis and treatment 1.MRI equipment is relatively complex in working principles and associated technologies.More and more attention has been paid to the teaching and training quality of MRI professionals.However,the traditional teaching and training processes present some problems,such as limited learning time and space,dependence on equipment with operational risks,and closed and invisible imaging process,which is difficult to meet the requirements of the new trend of global medical development for cultivating innovative talents.Digital twin(DT)technology has been gradually applied to the aerospace industry,industrial manufacturing,biomedicine,and other industries 2-5,among which training and rehearsal is an important DT application scenario 6.The DT five-dimensional model provides a general reference model to support landing applications of DT in different fields.It includes five dimensions of physical entity,DT model,DT data,connection interaction,and functional services.Based on the functional services,the DT can also be classified into six maturity levels according to the connection interaction method and the automation degree:virtual replication,virtual reflection,virtual control,virtual prediction,virtual optimization,and virtual-real symbiosis.These levels correspond to six types of functions respectively:design verification and equivalent analysis,visualized monitoring,remote operation and maintenance control,diagnosis and prediction,intelligent decision-making and optimization,and life cycle tracking,backtracking,and management 7.The core function of the MRI DT experimental platform is to provide teachers and students with a teaching environment that can enable them to carry out practical teaching activities in the virtual world and also meet the development demands of MRI equipment.The platform construction requires two technologies to complete the virtual reflection.The first is the construction of a digital environment,virtual human and digital equipment,and the main technical index is fidelity and immersion.The other is mechanism modeling between digital environment,virtual human and digital equipment,and its technical index is computing power and real-time performance 8-9.The platform operation promotes teaching reform in practice modes and evaluation methods and satisfies students learning needs from anywhere and at any time.The numerical MRI simulation for scientific research purposes has been in progress.Many research groups such as Wilhjelm et al.mainly focused on developing a virtual MRI scanner for utilization in biomedical engineering education 10,magnetic resonance angiography of simulated fluids 11,effects of uniformity of main magnetic fields on images under arbitrary sequence generator 12,image processing evaluation and image classification13,2123第 35 卷第 10 期2023 年 10 月Vol.35 No.10Oct.2023系统仿真学报Journal of System Simulationhttp:/www.china-simulation signal calculation under real electromagnetic field characteristics,noise and specific absorption rate calculation 14,and the design simulation of gradient coil,RF coil,and RF pulse 15-16.These simulations are primarily adopted for simulation and verification comparison in methods or equipment development.They do not pay attention to the equipment operation and the imaging process.Hacklanders group 17 developed an MRI simulation software that can simulate the basic operation of clinical MRI.However,the software does not consider the influence of factors such as human body information,equipment and environment,and does not visualize the original data acquisition.Based on previous studies 18,this paper proposed a mechanism modeling and simulation method for the construction of an MRI DT system and designed an MRI system framework.Some specific experimental cases have been employed to explain the data capture,virtual acquisition,iterative optimization,and application service.1Platform frameworkThe MRI DT experiment platform consists of physical devices,virtual devices,twin data,application services,and connections among various components.The platform framework is shown in Fig.1.Fig.1 Framework of DT experiment platform for MRI 2124第 35 卷第 10 期2023 年 10 月Vol.35 No.10Oct.2023Chen Shanshan,et al:Key Technology and Application of Digital Twin Modeling for MRIhttp:/www.china-1.1 Physical MRI equipmentThe physical MRI equipment includes magnet unit,RF unit,gradient unit,spectrometer unit,and computer unit,which are respectively responsible for generating static magnetic field,transmitting RF field,receiving signals,generating gradient magnetic field,and realizing human-computer interaction.The magnetic resonance phenomenon occurs when the hydrogen proton in the static magnetic field is excited by the RF field,and the spatial positioning of the signal is realized by the gradient magnetic field.The magnetic resonance signal during the relaxation is received by the RF receiving coil,and finally the image is reconstructed.The parameters reflected in the images are the hydrogen proton density(PD),spin-lattice relaxation time(T1),spin-spin relaxation time(T2),chemical shift(CS),and diffusion coefficient(D)of human tissues.1.2 Virtual MRI EquipmentThe DT virtual platform includes models of digital environment,virtual humans,and digital equipment.WebGL and Unity tools were adopted to complete the development of the DT virtual entity,in which the digital environment was divided into a virtual scene and a virtual process.The virtual scene was realized by 3D modeling according to the hospital environment,and the virtual process was formulated according to the clinical examination specifications.The shape of the virtual human model is constructed according to gender,age,and body type,and the motion information is constructed according to the number,type,and degree of freedom of joints to realize the arbitrary position of the whole body with a high degree of freedom to meet the experimental teaching needs.1.3 Feedback interactionThere is a two-way interaction between the physical MRI equipment and the virtual MRI equipment.Firstly,a physical digital human model is built using high-resolution isotropic quantitative sensing data from the physical MRI equipment,such as PD,T1,T2,CS,and D.These data are utilized to drive the synchronous operation of the MRI DT model.The MRI DT provides a complete set of imaging and control logic that is consistent with the physical MRI equipment and can accept parameter input instructions and output the same results as the physical MRI equipment.It is not limited by time,space,and environment,but also can master the running results of the physical equipment without touching it to a certain extent(virtual reflection).Secondly,the MRI DT can also control the operation of physical MRI equipment indirectly,and the DT calculation results can be fed back to the physical MRI equipment(virtual control).For example,the examination procedure of Patient B is set on the MRI DT without interrupting the examination of Patient A to connect the physical equipment seamlessly to perform the examination of Patient B after completing the examination of Patient A.Thirdly,the MRI DT can predict process and status of the physical equipment in the future,and employ two-way closed-loop interaction to transform unknowns into predictions(virtual prediction).For instance,failure information gathered by sensors,such as voluntary movements of humans,hardware and software errors,and environment anomalies,is transmitted to the DT for simultaneous computing of artefacts,obtaining artefacts solutions,and performing proactive maintenance in advance.Fourthly,the MRI DT can realize intelligent control by optimizing the physical 2125第 35 卷第 10 期2023 年 10 月Vol.35 No.10Oct.2023系统仿真学报Journal of System Simulationhttp:/www.china-equipment(virtual optimization).The MRI DT would continuously monitor and forecast the equipment to form the optimal maintenance scheme set for different fault situations during batches of patient examinations.After running for a period of time,the optimal solution for automatic disposal can be selected based on the actual fault situation.Ultimately,it is possible to achieve autonomous twinning through long-term synchronous operation and dynamic reconstruction,and thus the dynamic consistency of the DT with the physical equipment is maintained.Additionally,the visualization,prediction,decision,and optimization services can be maintained,and a sustainable,low-cost,and high-quality MRI DT can be achieved(virtual-real symbiosis).1.4 Mechanism modeling simulationIn employing MATLAB to realize the natural law simulation,the key point is to simulate the internal mechanism relationship between the digital environment,virtual human,and digital equipment,including physical data acquisition,virtual signal acquisition,and virtual image reconstruction.Finally,virtual images of MRI such as T1-weighted image(T1WI),T2-weighted image(T2WI),PD-weighted image(PDWI),short TI inversion recovery(STIR)images,and fluid attenuated inversion recovery(FLAIR)images are obtained.1.5 Application servicesFirstly,the digital environment,virtual human,and digital equipment model are built to represent the operation information of the physical MRI equipment in the virtual space and construct the experimental platform.Secondly,the DT simulation model is built in the virtual space to realize the interactive feedback between the virtual and physical equipment,and then the application service is designed.The scanning sequences supported in the platform are shown in Fig.1,including the spin echo(SE)sequence,inversion recovery(IR)sequence,gradient echo(GRE)sequence,and echo planar image(EPI)sequence.2Mechanism modelingMulti-dimensional physical data are obtained from the physical MRI equipment to implement the MRI DT mechanism modeling.Secondly,virtual human modeling is performed,and then virtual acquisition and image reconstruction are completed.Virtual MRI images are generated and DT fusion data are adopted for iterative optimization.Finally,the overall improvement of computing power,real-time performance,fidelity,immersion,interactivity,and user autonomy is achieved.2.1 Digital human modelingAs the digital human is the object to scan in the virtual equipment,its construction accuracy determines the fidelity of the reconstructed virtual images.The digital human construction requires multidimensional information,including the three-dimensional spatial information of human tissue(X,Y,Z spatial resolution of 1 mm 1 mm 1 mm),and the attribute information of the human tissues reflected by MR signals,such as PD,T1,T2,CS and D.The images generated by the physical MRI equipment were preprocessed with nonuniformity correction and rigid registration.The region growth method was employed for preliminary structure segmentation,and the k-nearest neighbor was to achieve fine boundary segmentation.Finally,the tissue parameters PD,T1,T2,CS and D were assigned.2126第 35 卷第 10 期2023 年 10 月Vol.35 No.10Oct.2023Chen Shanshan,et al:Key Technology and Application of Digital Twin Modeling for MRIhttp:/www.china-2.2 Virtual acquisition and image reconstructionAfter constructing the digital human,the virtual acquisition of physical data of the digital human is carried out according to the working process of the physical MRI equipment,and slice selection,phase encoding,frequency encoding,signal acquisition,and image processing are executed in sequence.k-space of a single slice can then be expressed assnmkl=AM0n=0N-1m=0M-1k=0K-1l=0L-1(k l)e j(2/N)kne j(2/M)lme-n/T*2(1)In Formula(1),sn,m,k,l is a four-dimensional complex sequence of double space and double time;A represents the signal amplification factor item;M0 is the original signal amplitude item;(k l)represents the level of proton density distribution item;e j