Abstract: This invention teaches a method to identify cellular abnormalities which are associated with disease states. The method utilizes infrared (IR) spectra of cell samples which are dried on an infrared-transparent matrix and scanned at the frequency range from 3000-950 cm−1. The identification of samples is based on establishing a reference using a representative set of spectra of normal and/or diseased specimens. During the reference assembly process, multivariate techniques such as Principal Component Analysis (PCA) and/or Partial Least Squares (PLS) are used. PCA and PLS reduce the data based on maximum variations between the spectra, and generate clusters in a multidimensional space representing the different populations. The utilization of Mahalinobis distances, or linear regression (e.g., Principle Component Regression on the reduced data from PCA) form the basis for the discrimination.

Abstract: This invention discloses a method to identify premalignant and malignant stages of cervical cancer from an infrared (IR) spectrum of exfoliated cervical cells which are dried on an infrared transparent matrix and scanned at the frequency range from 3000-950 cm.sup.-1. The identification of samples is based on establishing a calibration using a representative set of spectra of normal, dysplastic and malignant specimens. During the calibration process, multivariate techniques such as Principal Component Analysis (PCA) and/or Partial Least Squares (PLS) are used. PCA and PLS reduce the data based on maximum variations between the spectra, and generate clusters in a multidimensional space representing the different populations. The utilization of Mahalinobis distances, or linear regression (e.g., Principle Component Regression on the reduced data from PCA) form the basis for the discrimination.

Abstract: This invention teaches a method to identify cellular abnormalities which are associated with disease states. In one aspect, the invention is a method to distinguish premalignant and malignant stages of cervical cancer from normal cervical cells. The method utilizes infrared (IR) spectra of exfoliated cervical cells which are dried on an infrared transparent matrix and scanned at the frequency range from 3000-950 cm.sup.-1. The identification of samples is based on establishing a calibration using a representative set of spectra of normal, dysplastic and malignant specimens. During the calibration process, multivariate techniques such as Principal Component Analysis (PCA) and/or Partial Least Squares (PLS) are used. PCA and PLS reduce the data based on maximum variations between the spectra, and generate clusters in a multidimensional space representing the different populations. The utilization of Mahalinobis distances, or linear regression (e.g.