Resolution enhancement of diffusion imaging biomarkers in magnetic resonance imaging
View Patent ↗Resolution is enhanced for diffusion MR imaging. The tensors modeling the underlying water diffusion in brain tissues are used to interpolate other diffusion tensors, providing higher resolution diffusion biomarker images. Each diffusion tensor is represented by a pair of elements, the one in an ‘orientation space’ and another in a ‘shape space.’ The tensors are iteratively interpolated by averaging the aforementioned elements in separate mathematical spaces. The weighted average of the shape components of the diffusion tensors is computed in closed form, which decreases the runtime.
1. A method for resolution enhancement for diffusion MR imaging, the method comprising:
generating first diffusion tensors from the measurements of a diffusion magnetic resonance imaging system, the diffusion tensors being at a scan resolution of the magnetic resonance imaging system;
interpolating, by a diffusion tensor processor of the magnetic resonance imaging system, second diffusion tensors from the first diffusion tensors, the second diffusion tensors interpolated between the first diffusion tensors, an image of the first and second diffusion tensors having a spatial resolution higher than the scan resolution, the interpolation of the second diffusion tensors from the first diffusion tensors being by Riemannian weighted averaging using a rotation group in three dimensions as an orientation space and a shape space having the first diffusion tensors with a same orientation as elements, wherein the interpolation comprises computing shapes in the shape space in closed form;
calculating, by the magnetic resonance imaging system, biomarkers of the patient from the image of the first and second diffusion tensors; and
displaying an image of the biomarkers.
2. The method of claim 1 wherein generating comprises generating with measurements along six or more gradient directions.
3. The method of claim 2 wherein generating comprises generating with measurements along forty or more gradient directions.
4. The method of claim 2 wherein generating comprises fitting first diffusion models as orientation distribution functions to the measurements.
5. The method of claim 1 wherein interpolating comprises at least doubling the spatial resolution.
6. The method of claim 1 wherein interpolating comprising minimizing a sum of distances in the orientation and shape spaces.
7. The method of claim 6 wherein minimizing the sum comprises minimizing the sum of a distance of the first diffusion tensors in the shape space and a distance of the first diffusion tensors in the orientation space, with non-negative weights for the Riemannian weighted averaging applied using these distances.
8. The method of claim 1 wherein the computing of the shapes comprises computing shapes of the second diffusion tensors as a matrix exponential of a sum of weighted matrix logarithms of some of the first diffusion tensors and the corresponding alignment variables and their transposes.
9. The method of claim 1 wherein interpolating comprises iteratively computing a solution, the iterative computation alternating between aligning the first diffusion tensors with a current estimate of the second tensor and updating the orientation of that current estimate.
10. The method of claim 1 wherein computing the shapes in closed form comprises computing the shapes of the second diffusion tensors in a constant and a finite number of operations.
11. The method of claim 1 wherein calculating comprises calculating the fractional anisotropy as a biomarker.
12. The method of claim 1 wherein calculating comprises calculating the biomarkers at the spatial resolution greater than the scan resolution.
13. The method of claim 1 wherein calculating comprises calculating the biomarker as a scalar map, and wherein displaying comprises displaying the image of the scalar map.