Image analysis – Applications – Biomedical applications
Reexamination Certificate
1998-04-13
2001-12-04
Patel, Jayanti K. (Department: 2723)
Image analysis
Applications
Biomedical applications
C128S920000, C436S043000, C600S300000
Reexamination Certificate
active
06327377
ABSTRACT:
TECHNICAL FIELD
This invention relates generally, as indicated, to cell classification and, more particularly, to a system for increasing the speed and accuracy of cervical smear analysis.
BACKGROUND OF THE INVENTION
The examination of a cervical smear by what often is referred to as a Pap test is a mass screening cytological examination which currently requires individual visual inspection by a person of virtually all of the approximately 100,000 cells on a typical slide. The test, therefore, suffers from a high false negative rate due to the tedium and fatigue associated with this requirement for exhaustive search.
Prompted by the clear commercial potential for automated cervical smear analysis, several attempts to this end have been made heretofore. These attempts have proven to be unsuccessful at least partly because they could not accommodate overlapping cells as are typically found in the Pap smear. To circumvent the classification problems created by overlapping cells, specialized “monolayer preparations” have been prepared. A monolayer preparation is a specially prepared smear in which, the cervical cells are centrifuged and filtered so that only a single layer of cells results. Besides serious cell preservation and cell transportation problems, the expense and time involved in the monolayer preparation precludes its use as a population screening substitute for the Pap smear.
Even when limited to the non-overlapping cell images provided by the monolayer preparation, prior art attempts at automated cytological classification have not been able to process cervical smear images at anything close to manual processing time. Many of these attempts at automated cytological classification have relied on feature extraction algorithms which attempt to select and to measure some feature within the image, e.g., the shape of the cell nucleus. Feature extraction algorithms have failed due to the inability to segment the image into the components which require measurement. One cannot measure nuclear size, for example, unless the image is segmented so that the cellular nuclei are identified. Template matching, in which an actual image (not a mathematical quantity) is compared with stored exemplar images also has not been successful since it is computationally intensive and the infinite variety of possible Pap smear images or scenes would require an excessive number of exemplar image comparisons. The distinction between feature extraction and template matching is outlined in the Collings reference on pages 1 through 5 while image segmentation techniques are discussed in Chapter 7 of the Gonzalez reference.
An example of the limitations of the prior art can be found in the 1987 reference entitled “Automated Cervical Screen Classification” by Tien et al, identified further below.
Background references of interest are, as follows:
Rumelhart, David E. and McClelland, James L., “Parallel Distributed Processing,” MIT Press, 1986, Volume 1;
Tien, D. et al, “Automated Cervical Smear Classification,” Proceedings of the IEEE/Ninth Annual Conference of the Engineering in Medicine and Biology Society, 1987, p. 1457-1458;
Hecht-Nielsen, Robert, “Neurocomputing: Picking the Human Brain,” IEEE Spectrum, March, 1988, p. 36-41; and
Lippmann, Richard P., “An Introduction to Computing with Neural Nets,” IEEE ASSP Magazine, April, 1987, p. 4-22.
BRIEF SUMMARY OF THE INVENTION
An object of the invention is to automate at least part of the classification procedure for cytological specimens.
Another consistent objective is to provide semi-automation in a cytological specimen classification apparatus and method, whereby at least part of the cell classification procedure may be carried out by a human being.
Consistent with the foregoing, an object of the present invention is to classify cytological specimens into categories, for example, categories of diagnostic significance, and, more particularly, to automate at least a part of such classification procedure.
As used herein the term “automated” means that at least part of the apparatus is automated; in the preferred embodiment a portion of the method is carried out by a person.
Briefly, according to one embodiment, the invention includes an initial classifier (sometimes referred to as a primary classifier) preliminarily to classify a cytological specimen and a subsequent classifier (sometimes referred to as a secondary classifier) to classify those portions of the cytological specimen selected by the initial classifier for subsequent classification.
According to one embodiment, the invention includes an initial classifier (sometimes referred to as a primary classifier) preliminarily to classify a cytological specimen, a subsequent classifier (sometimes referred to as a secondary classifier) to classify those portions of the cytological specimen selected by the initial classifier for subsequent classification, and a tertiary classification to determine characteristics of or to classify those portions of the cytological specimen that are selected by the subsequent classifier for further classification.
In one embodiment the primary classifier performs a low level morphological feature screening function on the entire image while the secondary classifier performs a high level pattern matching identification on those images not screened out by the primary classifier.
In an embodiment of the invention the primary classifier classifies specimens according to size criteria and integrated optical density.
In an embodiment the secondary classifier is a neural net.
In an embodiment the tertiary classifier may be a person.
In an embodiment the present invention performs its classification of a group of specimens within the period of time typically consumed for this task by careful manual screening (i.e., approximately six minutes/specimen) or faster.
In an embodiment the present invention performs its classification on cytological specimens which contain the numbers and types of objects other than single layers of cells of interest that are typically found in cervical Pap smears (e.g., clumps of cells, overlapping cells, debris, clumps of leucocytes, bacteria, mucus) .
In an embodiment the present invention performs the above-described classification on cervical smears for the detection of pre-malignant and malignant cells.
In an embodiment the present invention displays, e.g., on a monitor or other display medium, cells adjacent or near one or more exemplary cells having features distinctive of a certain cell classification, such as large dark nuclei for malignant or pre-malignant cells, to facilitate, by comparison, cell screening by a person.
In an embodiment the present invention performs its classification with smaller false negative error rates than those typically found in conventional manual cervical smear screening.
In an embodiment of the present invention classification of cytological specimens into medically significant diagnostic categories will be more reliable, i.e., will have lower false negative error rates, than present methods.
In an embodiment the cytological classification system of the present invention does not require a modification in the procedure by which cellular specimens are obtained from the patient, i.e., standard Pap smears are used for its input.
In an embodiment the cytological classification system of the present invention will permit reliable classification within processing time constraints that permit economically viable operation.
In an embodiment of the invention classification of a cytological specimen is made by a person, and subsequent automated (or semi-automated) classification of selected specimens, such as those primarily noted as negative by such person, or, if desired, of all specimens, then may be carried out.
In an embodiment an automated specimen transfer mechanism is provided to transport cytological specimens between a storage location and an examination location.
In an embodiment of the invention a marking system marks selected areas of a cytological specimen at which prescribed characteristics appear, such markin
Hall Thomas L.
Rutenberg Mark R.
Alston & Bird LLP
AutoCyte North Carolina, L.L.C.
Patel Jayanti K.
LandOfFree
Automated cytological specimen classification 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 Automated cytological specimen classification system and method, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Automated cytological specimen classification system and method will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2564379