Quantitative Analysis of Flow Microfluorometric Data for Screening Gynecologic Cytology Specimens

Gynecologic cytology specimens that included the entire spectrum of cervical cytology classification were stained with a combination of propidium iodide and fluorescein isothiocyanate, then analyzed using a flow microfluorometer to measure nucleic acid and protein content, respectively. Numerous descriptors of the resulting two parameter distribution (nucleic acid versus protein) were defined. These descriptors included assessment of the presence or absence of abnormal cells. They also included measures of the staining intensity and dispersion of the normal squamous cell population and the intensity of inflammatory response in the cell population. Relative percentages of inflammatory and epithelial cells were demonstrated to effect the screening performance of this system only in borderline lesions. Decision tree algorithms allowed optimization of the selected parameters for screening logic of normal-abnormal decisions on a specimen-by-specimen basis. In addition, quantitative definitions of specimen adequacy were determined. Appropriate controls for batch staining of specimens were evaluated. These results of applying pattern recognition techniques to flow microfluorometer multiparameter data demonstrate that considerably more information about cell populations and subpopulations can be extracted than heretofore possible.