Image analysis – Applications – Biomedical applications
Reexamination Certificate
1999-12-22
2003-11-25
Au, Amelia M. (Department: 2623)
Image analysis
Applications
Biomedical applications
C382S275000
Reexamination Certificate
active
06654485
ABSTRACT:
FIELD OF THE INVENTION
This application is related to image processing and more particularly to a method for separating image information of interest from aggregate image data.
BACKGROUND OF THE INVENTION
In the art of image processing, it will often be the case that image information of interest is intermixed with, or even masked by other image information which is not of interest. Thus an objective in such cases will be a separation of the image information of interest from the information which is not of interest. An important image processing application in which this objective may come into play is found in the field of non-invasive imaging of anatomical structures and physiological processes. An exemplary such application is Magnetic Resonance Imaging (“MRI”), and particularly the MRI subclass known as functional Magnetic Resonance Imaging (“fMRI”) which has been shown to be particularly useful in the identification of parts of the brain associated with specific cognitive processes. Hereafter fMRI will be used both to illustrate the problems solved by the invention and as a preferred embodiment of the invention. It will be understood however that the invention is broadly applicable to the processing of image data generally, particularly to such images generated in a medical or clinical environment.
As will be known to those skilled in the art, functional Magnetic Resonance Imaging of the brain holds great promise as a tool to elucidate the functioning of the human brain [See, for example, applications of fMRI described in S. Ogawa et al., “Intrinsic signal changes accompanying sensory stimulation: Functional brain mapping with magnetic resonance imaging”,
Proc. Natl. Acad. Sci. USA
, 89 5951 (1992) and K. K. Kwong et al., “Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation”,
Proc. Natl. Acad. Sci. USA
, 89 5675 (1992)]. Changes in the oxygen content in cerebral blood causes small but detectable changes in an MR image. Since the oxygen content of the blood is known to be locally dependent on brain activity, detection of such changes in oxygen content provides a particularly reliable indicator of moment-to-moment brain function. There are, however, problems that limit the full utility of this technique. A major problem is that signal levels for changes in the image related to function are fairly small, and could well be lost among unwanted sources of image fluctuation. Such undesirable variations in the image include incoherent noise, approximately periodic fluctuations arising from physiological sources—e.g., cardiac and respiratory cycles, and motion of the experimental subject. Previous data analysis techniques have generally relied on statistical tests of significance to extract spatial maps of brain regions showing correlations with a stimulus timecourse. [See for example, P. A. Bandettini et al., “Processing Strategies for Time-Course Data Sets in Functional MRI of the Human Brain”,
Magnetic Resonance in Medicine
, 30 161 (1993)].
The conventional data analysis approaches generally have three drawbacks:
(i) they do not take into account the complete structure of the signal and the noise, thus preventing optimal detection of the signal;
(ii) such methods essentially produce maps of static brain regions, and it is difficult to assess the full spatio-temporal nature of the image signal in this way; and
(iii) an analysis methodology based on looking for changes correlated to the stimulus time course prevents detection of events that do not appear in synchrony with the stimulus.
As is known, techniques have been suggested in attempts to overcome the above drawbacks. To alleviate (ii), maps have been produced showing correlations with several shifted versions of the stimulus time course [See, E. A. DeYoe et al., “Functional Magnetic Resonance Imaging of the Human Brain”,
J. Neuroscience Methods
, 54 171 (1994)]. To avoid drawback (iii), the use of Principal Component Analysis (or equivalently Singular Value Decomposition) has been proposed to extract coherent changes in the signal that are not necessarily locked to the stimulus time course. [See J. R. Baker et al., “Statistical Assessment of Functional MRI Signal Change”, 2 626
, Proceedings of the Society of Magnetic Resonance, Second Meeting
Aug. 6-12, 1994, San Francisco, Calif.] However, such proposals have not led to a comprehensive solution to the full problem.
SUMMARY OF THE INVENTION
A methodology is presented, and a system for carrying out that methodology, for processing image data to extract image information of interest from aggregate image information containing other interfering information. Application of the methodology also leads to significant compression in the data, which reduces storage requirements. The methodology involves using an augmented linear decomposition of the image data, followed by frequency domain analysis and decomposition using multiple window techniques. The frequency domain analysis is augmented by subtraction of nuisance signals coherent with auxiliary time series which are simultaneously monitored. In carrying out the methodology of the invention, raw image data is formed into a matrix with spatial coordinates of that data along one dimension of the matrix and time along another dimension. That image data matrix is thereafter augmented in both the spatial coordinate and time dimensions by the addition of space and time shifted versions of the input data. The resultant matrix is then linearly decomposed into a sum of outer products of static images and time courses. Components of the image data corresponding to a baseline image, image information introduced by non-regular extraneous factors, and image information related to random noise are removed from the data by suppression of appropriate terms of the linear decomposition. The image data so reduced may contain both image information of interest and image data related to regularly occurring extraneous factors. The image data of interest is removed from such reduced aggregate image data by frequency domain filtering of such reduced aggregate data.
REFERENCES:
patent: 5260791 (1993-11-01), Lubin
patent: 6073041 (2000-06-01), Hu et al.
patent: 6091361 (2000-07-01), Davis et al.
patent: 6094050 (2000-07-01), Zaroubi et al.
patent: 6144873 (2000-11-01), Madore et al.
patent: 6289232 (2001-09-01), Jakob et al.
patent: 6321200 (2001-11-01), Casey
patent: 6370416 (2002-04-01), Rosenfeld
Mitra Partha Pratim
Thomson David James
Au Amelia M.
Bhatnagar Anand
Lucent Technologies - Inc.
LandOfFree
Image processing system and method does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Image processing system and method, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image processing system and method will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3183499