Adaptive reconstruction of phased array NMR imagery
First Claim
Patent Images
1. A method reconstructing a phased array nuclear magnetic resonance image comprising the steps of:
- a) using N coils, N>
1, collecting N sets of nuclear magnetic resonance data, each of said data sets having a multitude of sample values;
b) from said data sets, forming N complex images {Cj, j=1, . . . , N} of an image space, each of said N complex images formed from data from a single coil; and
,c) for a chosen pixel within an image space, forming a reconstructed pixel formed by the steps of;
1) mathematically creating a sample correlation matrix Rs by averaging complex cross products of the images {Cj, j=1, . . . N} over two or more pixel locations in the image space,2) establishing a noise array correlation matrix, Rn,3) computing a preliminary matrix P=Rn-1 ×
Rs,4) from P, establishing an Eigen de-composition P'"'"', consisting of a set of Eigenvectors {Vj, j=1, . . . , N} and associated Eigenvalues {λ
j, j=1, . . . , N},5) from P'"'"', establishing a vector m being the Eigenvector Vk whose associated Eigenvalue λ
k has the maximal magnitude among all the Eigenvalues (/λ
k />
/λ
j /, j=1, . . . , N),6) creating a pixel weight vector m* formed as a complex conjugate value of m, and,7) forming the reconstructed pixel I where I= ##EQU2##
2 Assignments
0 Petitions
Accused Products
Abstract
A method to model the NMR signal and/or noise functions as stochastic processes. Locally relevant statistics for the signal and/or noise processes are derived directly from the set of individual coil images, in the form of array correlation matrices, by averaging individual coil image cross-products over two or more pixel locations. An optimal complex weight vector is computed on the basis of the estimated signal and noise correlation statistics. The weight vector is applied to coherently combine the individual coil images at a single pixel location, at multiple pixel locations, or over the entire image field of view (FOV).
-
Citations
20 Claims
-
1. A method reconstructing a phased array nuclear magnetic resonance image comprising the steps of:
-
a) using N coils, N>
1, collecting N sets of nuclear magnetic resonance data, each of said data sets having a multitude of sample values;b) from said data sets, forming N complex images {Cj, j=1, . . . , N} of an image space, each of said N complex images formed from data from a single coil; and
,c) for a chosen pixel within an image space, forming a reconstructed pixel formed by the steps of; 1) mathematically creating a sample correlation matrix Rs by averaging complex cross products of the images {Cj, j=1, . . . N} over two or more pixel locations in the image space, 2) establishing a noise array correlation matrix, Rn, 3) computing a preliminary matrix P=Rn-1 ×
Rs,4) from P, establishing an Eigen de-composition P'"'"', consisting of a set of Eigenvectors {Vj, j=1, . . . , N} and associated Eigenvalues {λ
j, j=1, . . . , N},5) from P'"'"', establishing a vector m being the Eigenvector Vk whose associated Eigenvalue λ
k has the maximal magnitude among all the Eigenvalues (/λ
k />
/λ
j /, j=1, . . . , N),6) creating a pixel weight vector m* formed as a complex conjugate value of m, and, 7) forming the reconstructed pixel I where I= ##EQU2## - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10)
-
-
11. An imaging method for reconstructing an image from a phased array nuclear magnetic resonance image having N coils, N>
- 1, each coil generating a data set, said method comprising the steps of;
a) from said data sets, forming N complex images {Cj, j=1, . . . N} of an image space, each of said N complex images formed from data from a single coil; and
,b) reconstructing a pixel within said by; 1) mathematically creating a sample correlation matrix Rs by averaging complex cross products of the images {Cj, j=1, . . . N} over two or more pixel locations in the image space, 2) establishing a noise array correlation matrix, Rn, 3) computing a preliminary matrix P=Rn-1 ×
Rs,4) from P, establishing an Eigen de-composition P'"'"', consisting of a set of Eigenvectors {Vj, j=1, . . . , N} and associated Eigenvalues {λ
j, j=1, . . . , N},5) from P'"'"', establishing a vector m being the Eigenvector Vk whose associated Eigenvalue λ
k has the maximal magnitude among all the Eigenvalues (/λ
k /≧
/λ
j /, j=1, . . . , N),6) creating a pixel weight vector m* formed as a complex conjugate value of m, and, 7) forming the reconstructed pixel I where I= ##EQU3## - View Dependent Claims (12, 13, 14, 15, 16, 17, 18, 19, 20)
- 1, each coil generating a data set, said method comprising the steps of;
Specification