Method for functional brain imaging from magnetoencephalographic data by estimation of source signal-to-noise ratio
First Claim
1. A method of making a functional image of a subject'"'"'s brain from magnetoencephalographic measurements, said method comprising the steps of:
- (a) simultaneously collecting a plurality of magnetoencephalographic data signals from a plurality of sensors surrounding said brain;
(b) selecting an array of voxels relative to said plurality of sensors, said voxels defining a region of interest within said brain;
(c) determining one of a correlation matrix and a covariance matrix for said data signals;
(d) determining an uncorrelated noise variance matrix for said sensors;
(e) determining, for each of said sensors, a predicted signal value attributable to a theoretical source of unit strength at each of said voxels;
(f) determining, for each of said voxels, by inverse solution of said matrices and said predicted signal values, a source power, being the mean-square source current dipole moment;
(g) determining, for each of said voxels, an uncorrelated noise variance;
(h) determining, for each of said voxels, a function of said voxel source power and said voxel uncorrelated noise variance;
(i) for each of said voxels, converting said function into a false-color or gray-scale functional image of source activity;
(j) coregistering said functional image with a predefined anatomical image; and
, (k) displaying said coregistered images.
9 Assignments
0 Petitions
Accused Products
Abstract
An improved method, termed “statistical synthetic aperture magnetometry” (SSAM) of transforming magnetoencephalographic (MEG) measurements into corresponding three-dimensional images of the electrophysiological activity within the brain. The computed images are static, representing the time-integrated brain activity over a selected period. By selecting the time periods and frequency bands of interest, the SSAM method selectively images brain activity relating to different types of brain pathology or to cognitive events. Unlike prior art methods, the SSAM method compensates for the growth of ionic signal source strength estimates with depth into the head, resulting, in part, from the declining sensitivity of the MEG sensors. This is achieved by computing and displaying functions of the ratio of source strength to its noise for each element comprising the image. That is, a functional image is determined by an array of voxels where each voxel is based upon a function of source signal-to-noise ratio (SNR) rather than the source strength, alone. By using functions of SNR to represent source activity, the SSAM method achieves more accurate and higher resolution source localization. Each voxel is represented as a function of the ratio of a source power estimate to a source noise variance estimate. Such functions are found to be maximum at the true locations of sources, whereas plots of source power alone (as in prior art methods), show maxima which appear deeper and more diffuse in the brain than they in fact are.
-
Citations
29 Claims
-
1. A method of making a functional image of a subject'"'"'s brain from magnetoencephalographic measurements, said method comprising the steps of:
-
(a) simultaneously collecting a plurality of magnetoencephalographic data signals from a plurality of sensors surrounding said brain;
(b) selecting an array of voxels relative to said plurality of sensors, said voxels defining a region of interest within said brain;
(c) determining one of a correlation matrix and a covariance matrix for said data signals;
(d) determining an uncorrelated noise variance matrix for said sensors;
(e) determining, for each of said sensors, a predicted signal value attributable to a theoretical source of unit strength at each of said voxels;
(f) determining, for each of said voxels, by inverse solution of said matrices and said predicted signal values, a source power, being the mean-square source current dipole moment;
(g) determining, for each of said voxels, an uncorrelated noise variance;
(h) determining, for each of said voxels, a function of said voxel source power and said voxel uncorrelated noise variance;
(i) for each of said voxels, converting said function into a false-color or gray-scale functional image of source activity;
(j) coregistering said functional image with a predefined anatomical image; and
,(k) displaying said coregistered images. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13)
(a) performing said steps 1(a) through 1(h) to determine an active source power to noise variance ratio (a)ρ
θ
while said brain performs an activity task;
(b) performing said steps 1(a) through 1(h) to determine a control source power to noise variance ratio (c)ρ
η
while said brain performs a control task;
(c) for each of said voxels, deriving a ratio of said active and control source power to noise variance ratios;
(d) converting said ratio of said active and control source power to noise variance ratios into a false-color or gray-scale functional image of source activity;
(e) coregistering said functional image with a predefined anatomical image; and
,(f) displaying said coregistered images.
-
-
13. A method as defined in claim 1, further comprising the steps of:
-
(a) performing said steps 1(a) through 1(h) to determine an active source power and an active noise variance while said brain performs an activity task;
(b) performing said steps 1(a) through 1(h) to determine a control source power and a control noise variance while said brain performs a control task;
(c) for each of said voxels, deriving a function containing the ratio of the difference between said active and control source powers to the sum of their noise variance ratios;
(d) converting said function containing said ratio into a false-color or gray-scale functional image of source activity;
(e) coregistering said functional image with a predefined anatomical image; and
,(f) displaying said coregistered images.
-
-
14. A method of making a functional image of a subject'"'"'s brain from magnetoencephalographic measurements, said method comprising the steps of:
-
(a) simultaneously collecting a plurality of magnetoencephalographic data signals from a plurality of M sensors surrounding said subject'"'"'s brain;
(b) selecting K time-sampled portions of said collected magnetoencephalographic data signals;
(c) deriving a covariance matrix C of elements Cij, where;
(i) (ii ) i,j=1,2,3, . . . , M;
(iii) k=1,2,3, . . . , K;
(iv) mik is the response of the ith sensor during time sample k;
(v) mjk is the response of the jth sensor during time sample k;
(vi) (d) selecting an array of voxels defining a region of interest within said brain;
(e) determining an uncorrelated noise variance value for each of said sensors;
(f) determining a weighting coefficient
for each of said voxels, where G is Green'"'"'s function, μ
is a regularization parameter, and T denotes the matrix transpose;
(g) determining, for each of said voxels, at a selected target θ
, the mean-square source moment Sθ
2=[Gθ
T(C+μ
Σ
)−
1Gθ
]−
1;
(h) determining, for each of said voxels, at said respective targets θ
, a noise variance σ
θ
2=Wθ
TΣ
Wθ
;
(i) determining a function of source power and said noise variance for each of said voxels;
(j) converting said function into a false-color or gray-scale functional image of source activity;
(k) coregistering said functional image with a predefined anatomical image; and
,(l) displaying said coregistered images. - View Dependent Claims (15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29)
(a) performing said steps 14(a) through 14(h) to determine an active source power to noise variance ratio (a)ρ
θ
while said subject'"'"'s brain performs an activity task;
(b) performing said steps 14(a) through 14(h) to determine a control source power to noise variance ratio (c)ρ
θ
while said subject'"'"'s brain performs a control task;
(c) for each of said voxels, deriving a ratio of said active and control source power to noise variance ratios;
(d) converting said ratio of said active and control source power to noise variance ratios into a false-color or gray-scale functional image of source activity;
(e) coregistering said functional image with a predefined anatomical image; and
,(f) displaying said coregistered images.
-
-
29. A method as defined in claim 14, further comprising the steps of:
-
(a) performing said steps 14(a) through 14(h) to determine an active source power and an active noise variance ratio while said subject'"'"'s brain performs an activity task;
(b) performing said steps 14(a) through 14(h) to determine a control source power and a control noise variance ratio while said subject'"'"'s brain performs a control task;
(c) for each of said voxels, deriving a function of the ratio of said active and control source power difference to the active and control noise variance sum;
(d) converting said function of said active and control source power and noise variances into a false-color or gray-scale functional image of source activity;
(e) coregistering said functional image with a predefined anatomical image; and
,(f) displaying said coregistered images.
-
Specification