Noise considerations in the determination of diffusion tensor anisotropy

Area of Science:

• Neuroimaging
• Diffusion MRI
• Biomedical Engineering

Background:

• Diffusion Magnetic Resonance Imaging (dMRI) is crucial for studying white matter microstructure.
• Anisotropy indices quantify the directionality of water diffusion, reflecting tissue organization.
• Noise sensitivity in these indices can limit the accuracy and reliability of dMRI analysis.

Purpose of the Study:

• To investigate the noise sensitivity of various anisotropy indices in dMRI.
• To compare the performance of indices based on eigenvalues versus those using eigenvalues and eigenvectors.
• To propose methods for reducing noise and partial volume effects in dMRI.

Main Methods:

• Monte-Carlo computer simulations were employed to model noise effects.
• Magnetic Resonance Imaging (MRI) measurements were conducted on a phantom and 5 healthy volunteers.
• Comparison of anisotropy indices derived from eigenvalues only versus those incorporating eigenvectors.

Main Results:

• Anisotropy indices utilizing both eigenvalues and eigenvectors demonstrated superior noise robustness compared to eigenvalue-only methods.
• Spatial averaging with neighboring pixels significantly reduced the standard deviation of anisotropy indices.
• A novel time-domain averaging method, leveraging eigenvector orientation coherence, was proposed to mitigate partial volume effects.

Conclusions:

• Indices incorporating both eigenvalues and eigenvectors are recommended for improved noise performance in dMRI.
• Spatial and time-domain averaging techniques offer effective strategies for noise and partial volume artifact reduction.
• These findings contribute to more accurate and reliable characterization of white matter microstructure using dMRI.