Premium
TU‐D‐332‐05: The Benefits of Non‐Uniform Gradient Direction Specification in DTI: Simulations and Phantom Data
Author(s) -
Yanasak N,
Allison J,
Zhao Q,
Hu T,
Dhandapani K
Publication year - 2008
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.2962596
Subject(s) - imaging phantom , eigenvalues and eigenvectors , diffusion mri , tensor (intrinsic definition) , anisotropy , range (aeronautics) , physics , structure tensor , mathematical analysis , principal axis theorem , geometry , mathematics , computational physics , optics , computer science , image (mathematics) , materials science , artificial intelligence , medicine , quantum mechanics , magnetic resonance imaging , composite material , radiology
Purpose: The goal of this project was to optimize angular precision in determining the principle diffusion eigenvector of prolate tensors in a diffusion tensor imaging (DTI) series, using prior knowledge of principle eigenvector direction and non‐uniform specification of gradient directions. Additionally, the effect of non‐uniform gradient distributions on fractional anisotropy (FA) was characterized. Method and Materials: Simulations were conducted, representing diffusive behavior in tissue as manifest in a DTI image series. Diffusion‐encoding gradient directions were constrained in elevation, for a prolate tensor oriented along the z‐axis. Noise was added to generate multiple measurements as per a DTI ROI analysis. Tensors were calculated as well as FA. Angular precision was quantified as the dispersion in the angle between the principle eigenvector of the prolate and the z‐axis. To confirm simulations, a phantom containing glass capillary arrays (FA=0.68) was imaged with DTI using a 3.0T GE HDx scanner. Gradient directions were specified to match simulation results. Eigenvalues, eigenvectors, and FA were calculated within an ROI. Results: Simulations identify the range of elevation angles θ=30°–40° as being most sensitive to the determination of principle eigenvector direction. This result is fairly consistent within the range of FA=0.2–0.8. Prescription of gradients within a band of elevation angles having a width of Δθ ∼40°–60° and centered at θ ∼30° can improve angular precision by 30–40%, given an uncertainty in prior eigenvector direction of <30°. FA precision using this scheme is similar to uniform gradient prescriptions, and accuracy is improved at low SNR values. Data from phantom measurements generally agree with simulation results. Conclusion: This work suggests that prior knowledge of principle eigenvector direction can improve its final determination, using gradients prescribed non‐uniformly. This technique may be useful for increasing sensitivity in these conditions (e.g., spinal lesions, determining tumor infiltration of white matter tracts).