Error analysis of helmholtz-based MR-electrical properties tomography.

Area of Science:

• Medical Imaging
• Biophysics
• Electrical Properties Measurement

Background:

• MR electrical properties tomography (MR-EPT) measures tissue electrical properties by analyzing B1+ data derivatives.
• This process is sensitive to noise and Gibbs ringing, impacting accuracy.
• Finite difference (FD) kernels are used for spatial derivative computation in MR-EPT.

Purpose of the Study:

• To investigate the error introduced by finite difference (FD) kernels in MR-EPT.
• To evaluate the effectiveness of mitigation strategies like Gibbs ringing correction and Gaussian apodization on conductivity reconstructions.
• To assess the impact of image resolution, FD kernel size, k-space windowing, and SNR on MR-EPT accuracy.

Main Methods:

• Conductivity reconstructions were performed using simulations and MR measurements at 3T on phantoms and a human brain model.
• Accuracy was assessed based on varying image resolution, FD kernel size, k-space windowing, and signal-to-noise ratio.
• The influence of Gibbs ringing correction and Gaussian apodization was specifically investigated.

Main Results:

• Small FD kernels are sensitive to fluctuations, while large FD kernels are more noise-robust but cause boundary error propagation.
• Mitigation strategies provided only marginal improvements in conductivity reconstruction accuracy.
• MR-EPT conductivity reconstructions demonstrated low accuracy, with less than 37% of voxels achieving a relative error below 30%.

Conclusions:

• Numerical error from FD kernel computations is a primary limitation in Helmholtz-based MR-EPT.
• The accuracy of MR-EPT reconstructions is significantly hampered by FD error, especially in complex structures like the human brain.
• Further research is needed to address FD error for improved MR-EPT diagnostic capabilities.