Parallelization methods for implementation of a magnetic induction tomography forward model in symmetric multiprocessor systems

Maimaitijiang, Yasheng; Roula, M.A.; Watson, S.J.; Patz, R.; Williams, Robert J.; Griffiths, Huw

doi:10.1016/j.parco.2008.03.008

Cited by 11 publications

(5 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Furthermore, for a 2D problem, the inverse solver can usually be handled by a central processing unit (CPU), but for a 3D problem (or any problem with a large number of voxels), this could be inefficient for the solving process. High performance computing techniques using graphics processing units (GPUs) have been proposed to address this, where the parallelisation scheme was implemented on both the forward and inverse solver [78,79,80], significantly reducing computational time.…”

Section: Discussionmentioning

confidence: 99%

Magnetic induction tomography methods and applications: a review

Soleimani

2017

Meas. Sci. Technol.

104

View full text Add to dashboard Cite

Magnetic induction tomography (MIT) is a tomographic technique capable of imaging the passive electromagnetic properties of an object. It has the advantages of being contact-less and non-invasive, as the process involves interrogating the electromagnetic field of the imaging subject. As such, the potential applications of MIT are broad, with various domains of operation including biomedicine, industrial process tomography and nondestructive evaluation. Consequently, there is a rich-yet underexploredresearch landscape for the practical applications of MIT. The aim of this review is to provide a non-exhaustive overview of this landscape. The fundamental principles of MIT are discussed, alongside the instrumentation and techniques necessary to obtain and interpret MIT measurements.

show abstract

Section: Discussionmentioning

confidence: 99%

Magnetic induction tomography methods and applications: a review

Soleimani

2017

Meas. Sci. Technol.

104

View full text Add to dashboard Cite

show abstract

“…Incidentally, the image reconstruction was based on the calculation of the sensitivity matrix, which required a large number of forward problem solutions and several hours of computing time. To reduce this computational burden, the finite-difference algorithm, which was used to calculate the magnetic and scalar electric potentials in the imaging region, could for instance be parallelized using message passing interface library, as outlined in [20]. In MIT system design, resonant receiver circuits can lead to phase instabilities at the frequency of operation.…”

Section: Discussionmentioning

confidence: 99%

Rotational magnetic induction tomography

Trakic¹,

Eskandarnia²,

Li³

et al. 2012

Meas. Sci. Technol.

View full text Add to dashboard Cite

In magnetic induction tomography (MIT), an array of excitation coils is typically used to apply time-varying magnetic fields to induce eddy currents in the material to be studied. The magnetic fields from the eddy currents are then detected by an array of sensing coils to form an image of passive electromagnetic properties (i.e. conductivity, permittivity and permeability). Increasing the number of transmitters and receivers can provide a better image quality at the expense of a larger and more expensive MIT system. Instead of increasing the number of coils, this study investigates the possibility of rotating a single transmit-receive coil to image the electrical properties of the sample, by emulating an array of 200 transmit-receive coils by time-division multiplexing. Engineering details on the electromechanical design and development of a rotating MIT system are presented. The experimental results indicate that representative images of conductive samples can be obtained at 5 MHz by rotating a single transmit-receive coil.

show abstract

“…The implementation of GPU techniques has been investigated in the area of both hardfield [16] and soft-field tomography [14] to achieve computational gain, including electrical impedance tomography (EIT) [17] and electrical capacitance tomography (ECT) [18]. Compared to the use of GPU in previous MIT research [19,20], where the parallelisation of the finite difference (FD) algorithm was achieved, in this work, the time reduction using GPU is realised differently. An edge finite element method (FEM) is used to solve the forward model, which involves generating the meshes, solving the magnetic vector potential and fields in every single tetrahedral element, as well as solving the system linear equation which is a large scale sparse matrices calculation.…”

Section: Introductionmentioning

confidence: 99%

Magnetic Induction Tomography With High Performance Gpu Implementation

Ma¹,

Banasiak

Soleimani

2016

PIER B

View full text Add to dashboard Cite

Magnetic induction tomography (MIT) is a non-invasive medical imaging technique with promising applications such as brain imaging and cryosurgery monitoring. Despite its potential, the realisation of medical MIT application is challenging. The computational complexity of both the forward and inverse problems, and specific MIT hardware design are the major limitations for the development of MIT research in medical imaging. The MIT forward modeling and linear system equations for large scale matrices are computationally expensive. This paper presents the implementation of GPU (graphics processing unit) for both forward and inverse problems in MIT research. For a given MIT mesh geometry composed of 167,488 tetrahedral elements, the GPU accelerated Biot-Savart Law for solving the free space magnetic field and magnetic vector potential is proved to be over 200 times faster compared to the time consumption of a CPU (central processing unit). The linear system equation arising from the forward and inverse problem, can also be accelerated using GPU. Both simulations and experimental results are presented based on a new GPU implementation. Laboratory experimental results are shown for a phantom study representing potential cryosurgery monitoring using an MIT system.

show abstract

Parallelization methods for implementation of a magnetic induction tomography forward model in symmetric multiprocessor systems

Cited by 11 publications

References 11 publications

Magnetic induction tomography methods and applications: a review

Magnetic induction tomography methods and applications: a review

Rotational magnetic induction tomography

Magnetic Induction Tomography With High Performance Gpu Implementation

Contact Info

Product

Resources

About