Thyroid carcinoma diagnosis based on a set of karyometric parameters of follicular cells

BACKGROUND. In the works of various authors, performed with the help of mor phometry, an assessment of cell changes was carried out, as a rule, by one sign of their atypia (e.g. area of cell nuclei). Moreover, the selective an alysis of morphometric data for certain nosologic forms was performed. Such an approach allowed researchers to develop only additional diagnostic crit eria of thyroid carcinoma. At the same time, the use of a set of quantitati ve signs of cell atypia as a criterion for tumor malignancy made it possibl e to improve oncologic diagnosis. In this study, which is a first step in t he development of a quantitative approach to thyroid carcinoma diagnosis, a system of karyometric parameters was developed with the help of morphometr y of cytologic material of patients with major nosologic forms of thyroid p athology (papillary and follicular carcinoma, follicular adenoma, autoimmun e thyroiditis, and nodular colloid and diffuse toxic goiter). The significa nce of each parameter in thyroid carcinoma diagnosis was evaluated. METHODS. Thyroid cells of autopsy, operative, and aspiration biopsy materia l were the object of the study. One hundred fifty-two preparations with his tologically verified diagnosis were studied. Light morphometry was performe d with the help of a semiautomatic computer analyzer of images. One hundred fifty thyroid nuclei were assessed in each preparation. RESULTS. Among many karyometric parameters, six were selected: mean area of thyrocyte nuclei, ratio of frequencies, and number of classes on histogram s of distribution of their area, coefficient and a free term of the regress ion equation to a scattered diagram of perimeter and nuclear area, and the discriminant index. Boundary values of these parameters between the maligna nt and benign pathology were united into an expert system for thyroid carci noma diagnosis. Clinical trials of the system were performed. On the basis of the obtained results, the informativity (accuracy, sensitivity, specific ity) of its karyometric parameters was assessed. The formula for calculatio n of the weighting coefficient for each parameter was derived, the diagnost ic index being determined by the sum of their values. CONCLUSIONS. An expert system was created based on a set of karyometric par ameters reflecting the regularities of pathologic changes in thyrocyte nucl ei with malignant and benign pathology. The informativity and weighting coe fficients included into the system of parameters have been assessed, which made it possible to perform the probabilistic diagnosis of thyroid carcinom a. (C) 2001 American Cancer Society.