Chemistry: analytical and immunological testing – Biological cellular material tested
Reexamination Certificate
2000-10-18
2004-12-28
Soderquist, Arlen (Department: 1743)
Chemistry: analytical and immunological testing
Biological cellular material tested
C436S173000, C702S019000, C600S410000
Reexamination Certificate
active
06835572
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Technical Field of the Invention
The present invention relates to the use of magnetic resonance spectroscopy, and more particularly to such use for determining pathology, vascularization and nodel involvement of a biopsy of breast tissue.
2. Description of the Related Art
Within this application several publications are referenced by arabic numerals within parentheses. Full citations for these and other references may be found at the end of the specification immediately preceding the claims. The disclosures of all of these publications in their entireties are hereby incorporated by reference into this application in order to more fully describe the state of the art to which this invention pertains. Clinical evaluation, mammography and aspiration cytology or core biopsy (triple assessment) is undertaken on women presenting with breast lesions in most Western countries. Clinical assessment of palpable breast lumps is unreliable (1, 2). Impalpable lesions are usually discovered by screening or diagnostic mammnography, which has a reported sensitivity of 77-94% and a specificity of 92-95% (3). Cytological assessment of fine needle aspiration biopsies (FNAB) has sensitivities ranging from 65-98% and specificities ranging from 34-100% (4) depending on the skill of the person performing the aspiration and the expertise of the cytopathologist.
Following surgical excision of the lesion a time consuming process of preparation and pathological assessment of the specimen determines the nature of the tumour and the prognostic features associated with it.
SUMMARY OF THE INVENTION
Magnetic resonance spectroscopy (MRS) is a modality with a proven record in the diagnosis of minimally invasive malignant lesions (
5
-
11
). MR spectra of small samples of tissue or even cell suspensions enable the reliable determination of whether the tissue of origin is malignant or benign. Often MRS is able to detect malignancy before morphological manifestations are visible by light microscopy (
8
).
The potential of proton MRS from FNAB specimens to distinguish benign from malignant breast lesions has been demonstrated previously (
12
). At that time the MRS method relied on visual reading to process spectra and calculate the ratio of the diagnostic metabolites choline and creatine. This spectral ratio allowed tissue to be identified as either benign or malignant. In a small cohort of 20 patients within that study it also distinguished high grade ductal carcinoma in situ (DCIS) with comedonecrosis or microinvasion from low grade DCIS. Despite the limitation of visual inspection, which could only assess those spectra with a signal to noise ratio (SNR) of greater than 10, the visual method resulted in a diagnosis of malignant or benign with a sensitivity and specificity of 95 and 96%.
FIG. 1
shows malignant and benign spectra with good SNR while
FIG. 2
shows spectra with poor SNR.
Twenty percent of the spectra were discarded because low aspirate cellularity yielded inadequate SNR. In the initial study visual analysis used only two of fifty or more available resonances (
6
). Thus, potentially diagnostic and prognostic information in the remaining spectrum may have been ignored.
A 3-stage, robust statistical classification strategy (SCS) has been developed to classify biomedical data and to assess the full MR spectrum obtained from biological samples. The robustness of the method has been demonstrated previously with the analysis of proton MR spectra of thyroid tumours (
13
), ovarian (
14
), prostate (
9
), and brain tissues (
15
). The present invention applies SCS to assess proton MR spectra of breast aspirates against pathological criteria in order to determine the correct pathology on samples with sub-optimal cellularity and SNR and to determine if other diagnostic and prognostic information is available in the spectra.
The inventors have determined that SCS on MRS from breast FNAB is more reliable than visual inspection to determine whether a lesion is benign or malignant, and that a greater proportion of spectra is useful for analysis. Furthermore, spectral information obtained from MRS on FNAB of breast cancer specimens predicted lymph node metastases (overall accuracy of 96%) and vascular invasion (overall accuracy of 92%).
The invention provides a method for obtaining a statistical classifier for classifying spectral data from a biopsy of breast tissue to determine the classification of a characteristic of the breast tissue, comprising:
(a) locating a plurality of maximally discriminatory subregions in magnetic resonance spectra of biopsies of breast tissue having known classifiers of a characteristic,
(b) cross-validating the spectra by selecting a portion of the spectra, developing linear discriminant analysis classifiers from said first portion of spectra, and validating the remainder of the spectra using the classifiers from the first portion of the spectra, to obtain optimized linear discriminant analysis coefficients,
(c) repeating step (b) a plurality of times, each time selecting a different portion of the spectra, to obtain a different set of optimized linear discriminant analysis coefficients for each of said plurality of times;
(d) obtaining a weighted average of the linear discriminant analysis coefficients to obtain final classifier spectra indicating the classification of the characteristic based on the spectra; and
(e) comparing spectra from a biopsy of breast tissue of unknown classification of a characteristic to the final classifier spectra to determine the classification of the characteristic of the breast tissue.
The invention provides an apparatus for obtaining a statistical classifier for classifying spectral data from a biopsy of breast tissue to determine the classification of a characteristic of the breast tissue, comprising:
(a) a locator for locating a plurality of maximally discriminatory subregions in magnetic resonance spectra of biopsies of breast tissue having known classifiers of a characteristic of breast tissue,
(b) a cross-validator for selecting a portion of the spectra, developing linear discriminant analysis classifiers from said first portion of spectra, and validating the remainder of the spectra using the classifiers from the first portion of the spectra, to obtain optimized linear discriminant analysis coefficients, said cross-validator selecting, developing and validating a plurality of times, each time selecting a different portion of the spectra, to obtain a different set of optimized linear discriminant analysis coefficients for each of said plurality of times, and
(c) an averager for obtaining a weighted average of the linear discriminant analysis coefficients to obtain final classifier spectra indicating the classification of the characteristic based on the spectra,
whereby spectra from a biopsy of breast tissue of unknown classification of a characteristic may be compared to the final classifier spectra to determine the classification of the characteristic of the breast tissue.
The invention provides a method for determining the classification of a characteristic of breast tissue, comprising:
obtaining magnetic resonance spectra of a biopsy of breast tissue having unknown classification of a characteristic and comparing the spectra with a classifier, said classifier having been obtained by:
(a) locating a plurality of maximally discriminatory subregions in the magnetic resonance spectra of biopsies of breast tissue having known classifications of a characteristic of the breast tissue,
(b) cross-validating the spectra of (a) by selecting a portion of spectra, developing linear discriminant analysis classifier from said first portion of spectra, and validating the remainder of the spectra using the classifications from the first portion of the spectra, to obtain optimized linear discriminant analysis coefficients,
(c) repeating step (b) a plurality of times, each time selecting a different portion of the spectra, to obtain a different set of optimized linear discriminant analysis coefficients for each of said plurality
Mountford Carolyn E.
Russell Peter
Smith Ian C. P.
Somorjai Rajmund L.
Institute for Magnetic Resonance Research
Phillips Peter J.
Soderquist Arlen
LandOfFree
Magnetic resonance spectroscopy of breast biopsy to... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Magnetic resonance spectroscopy of breast biopsy to..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Magnetic resonance spectroscopy of breast biopsy to... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3298754