Image analysis – Image enhancement or restoration
Reexamination Certificate
1999-07-14
2002-05-21
Lee, Thomas D. (Department: 2724)
Image analysis
Image enhancement or restoration
C382S260000
Reexamination Certificate
active
06393157
ABSTRACT:
DESCRIPTION
1. Field of the Invention.
The invention is in the field of image processing performed on digital images, more particular on medical images such as mammographic images.
More particularly the invention relates to the automatic detection of objects or structures which have physical significance within a complex or cluttered digital image, such as a digitised medical image (e.g. a digital mammogram used for detecting breast cancer, or else a film-based mammogram which has been digitised so as to be suitable for analysis). In the case of mammography, the objects to be detected are radiological opacities which might be considered by radiologists to be suspiciously abnormal. More generally, the objects could be any suspicious radiological opacity (e.g. in chest X-rays) or could be a discrete feature in any other form of digital imaging (e.g. an object of interest in a sonar scan).
As indicated above, the invention has a wide range of potential application. The invention has been made within the specific context of breast imaging, and in particular as part of an attempt to construct a system which can automatically detect suspicious features within a breast image and alert a radiologist to the presence of those features. Radiologists are interested in a number of different abnormal features: this invention relates to the detection specifically of radiological opacities, which can include radiological categories such as ill-defined masses and stellate or spiculated masses or lesions.
Such an invention functions by obtaining a representation of a scene or image which can be manipulated and transformed by a computer system. The invention thus comprises an image acquisition part, a computation part, and an output part which either presents results of the computer manipulation or passes those results onto a further computer system for further analysis. The methods applied within the computer subsystem are particularly critical to the success of the overall system.
The following description relates to the computer's manipulations.
2. Description of the Prior Art
For the purposes of this description we shall define an image to be a representation of a physical scene, in which the image has been generated by some imaging technology: examples of imaging technology could include television or CCD cameras or X-ray, sonar or ultrasound imaging devices. The initial medium on which an image is recorded could be an electronic solid-state device, a photographic film, or some other device such as a photostimulable phosphor. That recorded image could then be converted into digital form by a combination of electronic (as in the case of a CCD signal) or mechanical/optical means (as in the case of digitising a photographic film or digitising the data from a photostimulable phosphor). The number of dimensions which an image could have could be one (e.g. acoustic signals), two (e.g. X-ray radiological images) or more (e.g. nuclear magnetic resonance images).
The general problem to be tackled is to analyse such an image so as to identify, or help identify, the brightness fluctuations in the image which result from the presence of discrete physical structures within the original scene. This is relatively easy in cases where an image comprises only a uniform background brightness with a single localised fluctuation in brightness caused by an object, or where an image contains brightness fluctuations due to more than one discrete object but which are widely separated on the image. The aim of this invention is to help analyse more complex images, where the brightness fluctuations due to the object of interest are superimposed on a background of variable brightness, or where a number of objects are either physically close together or even overlapping. The latter problem is particularly acute in images which are two-dimensional projections of a three-dimensional scene, such as radiological images, and we shall tend to refer to the latter in the subsequent discussion. We shall refer to this type of image as being “cluttered”.
A typical example of a cluttered image, and one for which this invention was specifically targeted, is mammographic images, in which the normal structure within a breast creates a complex background of structures, which have highly variable X-ray opacity, size and shape and which frequently overlap and intersect.
A powerful approach to helping to separate out these various structures has become known as multiresolution or multiscale analysis. The aim is to segregate structures in an image into ranges of sizes. In that way, structures in the size-range of interest can be enhanced or individually identified. This is an active research field, and a number of techniques have been proposed and used for carrying out multiscale analysis.
Multiresolution Analysis and Multiresolution Pyramids
The principle usually applied to multiresolution analysis is to smooth, or convolve, an image with a chosen function. If the smoothing function is a low-pass filter, this has the effect of supressing or eliminating high-frequency Fourier components. A set of multiresolution images can be constructed by smoothing the image with a succession of such functions, whose size increases by some factor (usually a factor 2) between each step. The differences between successive steps, or resolution levels, contain information on structures whose size are approximately in the range between the sizes of the filters that were used to generate the difference image. In general, we should expect that lower-resolution data can be represented by a coarser rate of sampling, in which case we would have constructed a multiresolution pyramid (e.a. Burt P J & Adelson E H. 1983. IEEE Transactions on Communications. 31, 532-540).
Wavelet Analysis
An improved approach to generating a multiresolution analysis or pyramid has become known as wavelet analysis (Mallat S G. 1989a. IEEE Transactions on Pattern Analysis and Machine Intelligence. 11, 674-693 & Mallat SG. 1989b, Transactions of the American Mathematical Society, 315. 6-87, Daubechies I., 1992. “Ten Lectures on Wavelets”, CBMS-NSF Regional Conference Series in Applied Mathematics. Society for Industrial & Applied Mathematics, Philadelphia, Pa. Chui C K, 1992 (ed). “Wavelet Analysis and its Applications”, Vol. I & II, Academic Press), in which the smoothing functions that link each resolution level are carefully chosen to achieve optimum results for the application. If the smoothing function is chosen so that it obeys a scaling law.
φ
(
x
)
=
2
∑
n
h
n
φ
(
2
x
-
n
)
then a multiresolution pyramid may be constructed such that no information is lost in constructing the pyramid. The difference images between successive resolution levels may themselves be encoded as a set of sample points to be convolved with a set of basis functions which are translates of a function known as the wavelet. The total number of sample points in the resulting pyramid may be no more than the number of sample points in the original image—a useful feature where data volume may he a consideration—and the process may be inverted, starting with the multiresolution pyramid to reconstruct exactly the original data.
Having produced the wavelet coefficients for an image, they may then be used to create a lossy compressed image in data-compression applications, or to analyse the structures in an image by measuring the amplitude of the wavelet coefficients at resolution levels of particular interest.
Examples in mammography include the detection of microcalcifications (e.g. Clarke L P. Kallergi M, Qian W. Li H-D, Clark R A & Silbiger M L. 1994. Cancer Letters. 77, 173-181—Laine A F Schuler S. Fan J & Huda, 1994. IEEE Translation on Medical Imaging, 13, 725-740) or in providing a method of texture analysis which forms part of a method for detecting abnormal masses (e.g. Wei D. Chan H-P Helvie M A, Sahiner B. Petrick N. Adler D D & Goodsitt M M, 1995. Medical Physics. 22, 1501-1513).
Disadvantages of Wavelet Analysis
The previous approach using wavelets is not p
Brinich Stephen
Hoffman, Warnick & D'Alessandro LLC
Lee Thomas D.
Merecki John A.
Particle Physics and Astronomy Council
LandOfFree
Method for automatically detecting objects of predefined... 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 automatically detecting objects of predefined..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for automatically detecting objects of predefined... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2893820