Detail publikace

Fast Calculation of T2 Relaxation Time in Magnetic Resonance Imaging

Originální název

Fast Calculation of T2 Relaxation Time in Magnetic Resonance Imaging

Anglický název

Fast Calculation of T2 Relaxation Time in Magnetic Resonance Imaging

Jazyk

en

Originální abstrakt

The main parameters displayed by means of magnetic resonance include, for example, relaxation times T1 and T2 or diffusion parameters. This paper presents the computation of relaxation time T2 measured indirectly with the Spin Echo method. The sensing coil of the tomograph provides a signal in which the important factor is the location of the peaks from individual measurements. These points must be interleaved with an exponential function. The relaxation time T2 can be directly determined from the exponential shape. The described process has to be repeated for each pixel of the sensed tissue, and this requirement makes the processing of larger images very demanding in terms of both the actual computation and the time needed for the entire operation. More concretely, if we assume the common resolution of 256x256, 20 slices, and five measurements with different times TE, it is necessary to reconstruct 1.3x106 exponential functions in total, which requires the processing of more than 6 MB of data. At present, such computation lasts approximately 3 minutes if performed by means of a regular PC. The author discusses various approaches to the parallelization of the given problem. In the described context, the time required for the processing of the applied three-dimensional image was shortened to 300 ms thanks to simple interpolation approach. The final section of the paper comprises a detailed comparison of the computation times characterizing both the sequential and the parallel solutions.

Anglický abstrakt

The main parameters displayed by means of magnetic resonance include, for example, relaxation times T1 and T2 or diffusion parameters. This paper presents the computation of relaxation time T2 measured indirectly with the Spin Echo method. The sensing coil of the tomograph provides a signal in which the important factor is the location of the peaks from individual measurements. These points must be interleaved with an exponential function. The relaxation time T2 can be directly determined from the exponential shape. The described process has to be repeated for each pixel of the sensed tissue, and this requirement makes the processing of larger images very demanding in terms of both the actual computation and the time needed for the entire operation. More concretely, if we assume the common resolution of 256x256, 20 slices, and five measurements with different times TE, it is necessary to reconstruct 1.3x106 exponential functions in total, which requires the processing of more than 6 MB of data. At present, such computation lasts approximately 3 minutes if performed by means of a regular PC. The author discusses various approaches to the parallelization of the given problem. In the described context, the time required for the processing of the applied three-dimensional image was shortened to 300 ms thanks to simple interpolation approach. The final section of the paper comprises a detailed comparison of the computation times characterizing both the sequential and the parallel solutions.

BibTex


@inproceedings{BUT109572,
  author="Jan {Mikulka} and Pavel {Dvořák}",
  title="Fast Calculation of T2 Relaxation Time in Magnetic Resonance Imaging",
  annote="The main parameters displayed by means of magnetic resonance include, for example, relaxation times T1 and T2 or diffusion parameters. This paper presents the computation of relaxation time T2 measured indirectly with the Spin Echo method. The sensing coil of the tomograph provides a signal in which the important factor is the location of the peaks from individual measurements. These points must be interleaved with an exponential function. The relaxation time T2 can be directly determined from the exponential shape. The described process has to be repeated for each pixel of the sensed tissue, and this requirement makes the processing of larger images very demanding in terms of both the actual computation and the time needed for the entire operation. More concretely, if we assume the common resolution of 256x256, 20 slices, and five measurements with different times TE, it is necessary to reconstruct 1.3x106 exponential functions in total, which requires the processing of more than 6 MB of data. At present, such computation lasts approximately 3 minutes if performed by means of a regular PC. The author discusses various approaches to the parallelization of the given problem. In the described context, the time required for the processing of the applied three-dimensional image was shortened to 300 ms thanks to simple interpolation approach. The final section of the paper comprises a detailed comparison of the computation times characterizing both the sequential and the parallel solutions.",
  booktitle="Proceedings of PIERS 2014 in Guangzhou",
  chapter="109572",
  howpublished="online",
  year="2014",
  month="september",
  pages="2331--2335",
  type="conference paper"
}