Affordable Access

A pattern classification system for automated cervical cytologic screening based on flow microfluorometric analysis.

Authors
Type
Published Article
Journal
Analytical and quantitative cytology
Publication Date
Volume
1
Issue
1
Pages
61–66
Identifiers
PMID: 397790
Source
Medline
License
Unknown

Abstract

A data analytic technique is described for use in automated cervical cytology. The method entails the application of a two-dimensional Fourier transform to histogram data obtained by flow microfluorometry and the subsequent use of the Fourier coefficients as parameters for pattern classification. Analyses were performed on 186 samples including material from 62 positive and 124 negative cases. Cell suspensions were stained with propidium iodide and fluorescein isothiocyanate, and red and green cytofluoresence were measured simultaneously. The two-dimensional histogram data were normalized, the Fourier transform was applied, and a multivariate classifier based on 30 coefficients was assembled using a training set of half the original series. Performance was then assessed on the remaining cases. Overall accuracy was 79.6%, with a false-positive rate of 14.8% and a false-negative rate of 31.2%. The potential applicability of this approach as the basis of a practical screening system is discussed.

Statistics

Seen <100 times