Image analysis – Applications – Biomedical applications
Reexamination Certificate
2000-03-02
2003-04-01
Patel, Jayanti K. (Department: 2625)
Image analysis
Applications
Biomedical applications
C250S587000
Reexamination Certificate
active
06542628
ABSTRACT:
BACKGROUND OF THE INVENTION
The invention concerns the automatic detection of small-sized elements of interest mingled with noise in a digital radiographic image.
It has a particularly interesting application in the medical field, in which analysis of the fixation of calcium salts in the tissues of the organism is undertaken. These micro-calcifications are of small size, and it is often difficult to distinguish them from the noise in an image, where microcalcifications and noise are represented at the same time in the form of small spots.
The effective treatment of a breast cancer necessitates a medical follow-up from the first signs of cancer. These signs are present in the form of calcium salts fixed on the tissues of the mammary gland. They are generally detected on an X-ray of the mammary gland, in the course of which they appear on an image in the form of minuscule bright spots (light spots).
Their small dimensions render their detection difficult and they are sometimes missed.
Entirely digitized mammography acquisition chains are known, intended for the detection of microcalcifications. Said detection necessitates the development of image processing algorithms in order to differentiate the light spots due to said microcalcifications from those due to noise.
In general, in the prior art, the detection of microcalcifications is carried out from a film that is digitized. Filtering can then be carried out, followed by application of a threshold on the image.
One can cite the method of estimation of noise level from the image itself, developed by Nico Karssemeijer, “Adaptative Noise Equalization and Recognition of Microcalcification Clusters in Mammograms,” in “State of the Art in Digital Mammographic Image Analysis,” published by K. W. Bowyer, S. Astley,
World Scientific,
1994. But this method leads to an ambiguity of distinction between the elements representing noise and the elements representing microcalcifications, for it is based on scanned film images and does not take the acquisition parameters into account.
Another method was presented by Michael Brady, Ralph Highman and Brasil Shepstone, “A representation for mammographic image processing,”
Medical Image Analysis,
Vol. 1, No. 1, 1996, pp. 1-18, in which a process of image generation on X-ray film is modeled. But this process does not include a noise model.
In all the systems of the prior art, the selection of a threshold is often difficult, since the noise is dependent on the image content. The size of the microcalcifications not being limited by a minimum value, their detection results from a compromise between the detection of many noises as microcalcifications and a neglect of small-sized micro-calcifications.
BRIEF SUMMARY OF THE INVENTION
The invention is directed to a solution to the problem of determination of the threshold in order to extract the microcalcifications from the noise.
In an embodiment of the invention an estimate is made of the probability for a pixel or an elementary zone to be a false negative, that is, the probability for a pixel representing a theoretical microcalcification not to be detected, in order to afford the user the choice of an acceptable number of false negatives.
In an embodiment of the invention a method of improved detection of elements of interest by means of a detection system in a digital radiographic image of an object, acquired on an acquisition chain.
According to a general characteristic of the invention, the method comprises two phases. The first is a calibration phase, in which one determines from a mathematical model of the acquisition chain and object a plurality of images corresponding to a plurality of gray input levels representing a set of parameters of the object and a set of parameters of the acquisition chain variable over a range of predetermined values. One then elaborates, from the system of detection and plurality of images, a mathematical model of detection giving a theoretical number of elements detected as a function of a set of parameters of the acquisition chain, a useful output signal of the detection system and a gray background level of the image considered, representing one of the sets of parameters of the object.
Each gray level presented on input of the mathematical model of the acquisition chain and of the object advantageously represents a set of parameters of the object, such as the thickness and composition of a breast on a mammogram.
The mathematical model of detection thus established represents the performances of the detection system relative to a set of parameters variable over a range of given values. In other words, the theoretical response of the detection system is determined as a function of all the configurations realizable in the course of an X-ray. The first phase is preferably carried out only once for any detection system.
The second phase is carried out for each acquisition of a digital radiographic image. It is a phase of use in which the detection system is applied on the digital radiographic image. Then, for each elementary zone of that image, one introduces as input data in the mathematical model of detection the useful output signal of the detection system, a set of parameters of the acquisition chain taken among the predetermined values and used to obtain said digital radiographic image, and a gray background level calculated on said digital radiographic image, so as to determine a theoretical number of elements detected. One then selects among the elementary zones of the digital radiographic image acquired those whose theoretical number of detected elements satisfies a predetermined criterion.
REFERENCES:
patent: 4947323 (1990-08-01), Smith
patent: 5515450 (1996-05-01), Takeo et al.
patent: 5627907 (1997-05-01), Gur et al.
patent: 5633511 (1997-05-01), Lee et al.
patent: 6137898 (2000-10-01), Broussard et al.
patent: 4-86820 (1992-03-01), None
patent: 11508815 (1997-09-01), None
patent: 11501538 (1999-02-01), None
patent: 11-56828 (1999-03-01), None
Muller Serge
Rick Andreas
Azarian Seyed
Chaskin Jay L.
GE Medical Systems S.A.
Patel Jayanti K.
LandOfFree
Method for detection of elements of interest in a digital... 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 detection of elements of interest in a digital..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for detection of elements of interest in a digital... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3068509