Surgery – Diagnostic testing – Detecting nuclear – electromagnetic – or ultrasonic radiation
Reexamination Certificate
1998-08-24
2004-02-24
Lateef, Marvin M. (Department: 3737)
Surgery
Diagnostic testing
Detecting nuclear, electromagnetic, or ultrasonic radiation
C382S128000, C324S248000
Reexamination Certificate
active
06697660
ABSTRACT:
TECHNICAL FIELD
This invention is directed to a method of transforming magnetoencephalographic (MEG) measurements into corresponding three-dimensional images of the electrophysiological activity within the brain. Image blurring due to ambiguity of source depth is greatly reduced by computing and displaying a measure of the proportion of source strength to noise for each discrete element comprising the image, to achieve more accurate and higher resolution source localization.
BACKGROUND
Methods and instrumentation for generating functional images of human brain activity are of great importance for diagnosis of clinical brain disorders. Moreover, it is important that a functional imaging technology be non-invasive, providing a favourable risk-benefit ratio for patients and extending its application to fundamental neuropsychological research. “Functional brain imaging” implies a method that measures and displays brain activity, or some parameter relating to brain activity, as a function of a three-dimensional position within the head. When combined or fused with an anatomical image of the head, such as from a magnetic resonance image (MRI) or computed tomography (CT) scan, a functional brain image relates the level of brain activity to specific anatomical structures within the brain. Observing changes in the level of brain activity at specific loci helps persons skilled in the art understand how the brain operates in both health and disease conditions.
U.S. Provisional Application No. 60/072,340 filed Jan. 23, 1998 which is incorporated herein by reference, discloses an analytic method called “synthetic aperture magnetometry” (SAM), that is useful for transforming magnetoencephalographic (MEG) signals into corresponding functional brain images. However, if the SAM method is used to compute functional brain images, without using background or control state subtraction, the signal source strength estimates tend to get progressively stronger toward the center of the head. This results in an ambiguity of source depth, while resolving source features on a surface of constant depth. The progressive image “blur” with depth is not in agreement with acceptable neurophysiological data. However, when the SAM method is used to estimate differential images, comparing control and active brain states, the ambiguity in source depth is greatly reduced, but not eliminated. The differential mode is therefore the preferred mode of using the SAM method.
The present invention provides an improved method, termed “statistical synthetic aperture magnetometry” (SSAM). Like the SAM method, the SSAM method transforms 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 some selected period. Furthermore, by selecting frequency bands of interest, the SSAM method selectively images brain activity relating to different types of brain pathology or to cognitive events. The SSAM method uses the SAM method to compute both a source estimate and a noise estimate. However, unlike the SAM method, the SSAM method compensates for the growth of the source strength estimate with depth into the head. Such compensation is achieved by computing and displaying an image for which each element represents a function of the ratio of source strength to noise. That is, each image element (termed a “voxel” for volume images and “pixel” for planar image) is based upon the source signal-to-noise ratio (SNR) rather than the source strength, alone. By using SNR or a derivative function to represent source activity the SSAM method achieves more accurate and higher resolution source localization. The SAM method computes the root mean square (RMS) source estimate, on a voxel-by-voxel basis. However, it is also possible to use the SAM method to compute an estimate of the voxel uncorrelated noise. The SSAM method represents each image element as some 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 the SAM method), show maxima which appear deeper in the brain than in fact are.
The SSAM method lends itself to imaging both differential and non-differential brain activities. In differential mode, one compares the SNR of one state of brain activity (e.g., “active”) to another state (e.g., “control”); forming the appropriate combinations may compare two or more states of brain activity. These comparisons may be made either on a voxel-by-voxel basis, or by forming multiple-voxel statistical quantities. The comparison images may be expressed as a T (difference) or F (ratio) statistics. SSAM may also be used to transform MEG measurements into non-comparison functional brain images. The non-differential images are usually cast as z-statistics. The new functional images are computed from the statistics associated with each image element. The statistics used may be signal-to-noise ratio (SNR), functions of SNR, ratios of SNRs, differences, etc. When cast in the framework of existing z, T, or F-statistics, one can compute probability values representing the statistical significance of the activity. Thus, a functional image can be generated from the p-values.
The average MEG sensor measurement noise is estimated from the least-significant singular value of a measurement covariance matrix, following singular value decomposition (SVD), as this represents the spatial mode having the smallest signal power.
Unlike the SAM method, which maps root mean square (RMS) source power, the quantity that is mapped by the SSAM method is either the z-statistic, T-statistic, or a derived probability statistic of that power. The SSAM method results in a highly significant improvement in image quality over that attainable by the SAM method or by other currently available MEG image transforms. Specifically, the SSAM method improves image contrast and resolution, such that activity within deep structures of the brain can be observed separately from superficial sources. Thus, the SSAM method does not represent a trivial change in the units that are mapped from dipole power to a statistical representation of the dipole power. The SSAM method has been reduced to practice and has been demonstrated to image abnormal high and low frequency brain activity in cases of epilepsy, and can also map the brain areas relating to the production of speech.
SUMMARY OF THE INVENTION
In accordance with the preferred embodiment, the invention provides a method of making a functional image of a subject's brain from magnetoencephalographic measurements. A plurality of magnetoencephalographic data signals are simultaneously collected from a plurality of sensors surrounding the brain. An array of voxel coordinates is selected, relative to the sensors, such that the voxels define a region of interest within the brain. The covariance of the measured data signals is then determined, together with the sensors' uncorrelated noise variance. These quantities facilitate determination, for each voxel, a “source power”, being the mean-square source current dipole moment. The uncorrelated noise variance for each voxel is then determined, followed by a determination, for each voxel, of a function of the voxels' source power and uncorrelated noise variance. The function is then further processed into derivative statistics, or is converted directly into a false-color or gray-scale functional image of source activity. This functional image is then coregistered with a predefined anatomical image. The coregistered images can then be displayed.
Either before or after time-sampling, the collected MEG data signals are frequency domain filtered to exclude signal frequencies outside a selected frequency range. For example, the selected frequency range may be characteristic of a selected brain activity.
The source power to noise variance ratio for each image element
ρ
θ
=
S
θ
2
σ
θ
2
Allen Kenneth R.
CTF Systems Inc.
Lateef Marvin M.
Shaw Shawna J.
Townsend and Townsend / and Crew LLP
LandOfFree
Method for functional brain imaging from... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for functional brain imaging from..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for functional brain imaging from... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3318743