Surgery – Diagnostic testing – Detecting nuclear – electromagnetic – or ultrasonic radiation
Reexamination Certificate
2000-12-15
2002-07-16
Lateef, Marvin M. (Department: 3737)
Surgery
Diagnostic testing
Detecting nuclear, electromagnetic, or ultrasonic radiation
C600S473000, C600S477000, C356S302000, C356S303000
Reexamination Certificate
active
06421553
ABSTRACT:
FIELD OF THE INVENTION
This invention relates generally to the use of spectral data for categorizing materials. More particularly, the invention relates to the use of spectral data for classifying tissue by discriminant analysis.
1. Field of the Invention
This invention relates generally to the use of spectral data for categorizing materials. More particularly, the invention relates to the use of spectral data for classifying tissue by discriminant analysis.
2. Background of the Invention
Information obtained from spectral analysis is useful in such diverse fields as geology, chemistry, medicine, physics and biology. For example, it is known that cells and tissues emit characteristic spectra in response to light stimulation. The nature of those spectra is indicative of the health of the tissue. Thus, a cancerous tissue has associated with it an emission spectrum that is different than the emission spectrum of a corresponding healthy tissue. Similarly, emission spectra are used in attempts to differentiate between the mineral content of two or more geologic samples.
In any spectral analysis of a sample, the condition of which is unknown, an attempt is made to associate the sample with a reference sample that represents a known state. Commonly, spectral values such as amplitude measured in response to excitation, are used to place a test sample on a range of spectral values obtained from samples of known condition. The object is to identify a best fit between the test sample and a sample the condition of which is known. While such an analysis is straightforward in principle, it is often the case that a spectrum recorded from a test specimen does not match any known spectrum, or appears to have features present in more than one known spectrum. In addition, there are often multiple features of interest or of consequence in a single spectrum, which makes any comparison problematic. Thus, it is not always clear where a test sample should be placed with a range of known samples. the result is that the spectra of many samples cannot be assigned a condition by comparison to known samples without significant ambiguity.
SUMMARY OF THE INVENTION
In spectral analysis, different types of spectra provide distinct and useful kinds of information. In conventional spectroscopic analysis, one may employ such diverse methods and techniques as optical spectra spanning the infrared, visible and ultraviolet based on the interactions of materials with light (or electromagnetic radiation), spectroscopic methods based on electromagnetic resonance, such as electron spin resonance (esr) and nuclear magnetic resonance (nmr) spectroscopy, x-ray spectroscopic methods based on crystal structures, Moessbauer spectroscopy based on nuclear resonance, and Raman spectroscopy based on molecular dynamics. Often, to characterize a specimen as completely as possible, all of the above methods are employed, irrespective of the results of a particular method of analysis, and the totality of the results obtained are used to analyze or describe the specimen being examined. In the present inventive systems and methods, a first spectral result is used to determine whether a specimen being tested can be categorized, and if the initial result warrants further analysis, as when a categorization cannot be made in a definitive manner, a further specific analysis is performed in an attempt to obtain suitable categorization information. In some instances, it is impossible to reach a definitive conclusion.
The invention provides methods of comparing spectral data from a test sample with spectral data corresponding to known conditions. In particular, the invention provides methods for assigning at least one condition to a test sample based upon a relationship between at least two metrics, each representative of the similarity of the test sample to samples having known conditions, based upon a comparison of spectral data from the test sample to a constellation of reference data from the samples having known conditions. For a known condition, one can record and analyze reference spectra from a number of specimens having the known condition to provide a number of reference data points characteristic of that known condition. N-dimensional data points obtained from reference spectra for a plurality of specimens exhibiting the same condition tend to fall in a cluster, or constellation.
In one aspect, the invention provides a method for determining a condition of a test specimen. A preferred method comprises obtaining at least one optical spectrum from a test specimen, the optical spectrum being characterized by (N) quantitative features, each of the (N) quantitative features corresponding to one of (N) pre-selected wavelengths, wherein (N) is an integer greater than 1. Preferred methods further comprise determining a first metric between a point in N-dimensional space corresponding to the (N) quantitative features of the optical spectrum from the test specimen and a point in N-dimensional space characteristic of a first constellation of reference data that includes a first plurality of points in N-dimensional space, each of which corresponds to (N) quantitative features of a reference optical spectrum at the (N) pre-selected wavelengths, the first constellation of reference data being representative of a first known condition. Preferred methods further comprise determining a second metric between the point in N-dimensional space corresponding to the (N) quantitative features of the optical spectrum from the test specimen and a point in N-dimensional space characteristic of a second constellation of reference data that includes a second plurality of points in N-dimensional space, each of which corresponds to (N) quantitative features of a reference optical spectrum at the (N) pre-selected wavelengths, the second constellation of reference data being representative of a second known condition. Preferred methods further comprise assigning one of the first and the second conditions to the test specimen, based at least in part on a relationship between the first and the second metrics.
In one embodiment, the first metric and the second metric are each selected from the group consisting of a square root of a sum of the squares of the differences in coordinates of points in the N-dimensional space, a Mahalanobis distance, a Bhattacharyya distance, and a probability.
In another embodiment, methods of the invention further comprise analyzing biological tissue. Preferred methods comprise obtaining at least one optical spectrum from a tissue sample obtained form a patient, and determining a first metric by comparing optical features of the spectrum obtained from the tissue sample to a first constellation of data points representing spectral characteristics of a first condition; determining a second metric by comparing optical features of the spectrum obtained from the tissue sample to a second constellation of data points representing spectral characteristics of a second condition; and assigning one of the two conditions to the test sample based upon a relationship between the first and second metrics. In a preferred embodiment, the optical spectra obtained are selected from the group consisting of a fluorescence spectrum, a reflectance spectrum and a Raman spectrum. In one embodiment, the tissue specimen is human cervical tissue or a sample prepared from human cervical tissue. In one embodiment, methods of the invention are conducted in vivo. In a preferred embodiment, the first condition is normal health and the second condition is selected from the group consisting of moderate cervical intraepithelial neoplasia (CIN II) and severe cervical intraepithelial neoplasia (CIN III).
In one embodiment, the method further includes reporting that the optical spectrum is inconclusive with regard to the first known condition and the second known condition in response to each of the first and the second metrics exceeding a pre-determined metric value. In one embodiment, at least one of the points in N-dimensional space characteristic of a first constellation
Costa Peter J.
Hui Kwong
Nordstrom Robert J.
Lateef Marvin M.
MediaSpectra, Inc.
Qaderi Runa Shah
Testa Hurwitz & Thibeault LLP
LandOfFree
Spectral data classification of samples does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Spectral data classification of samples, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Spectral data classification of samples will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2841064