Differential diagnosis of follicular tumor by expert systems based on a set of quantitative features of thyrocyte nuclei and aggregates.

ABSTRACT

OBJECTIVE: To develop expert systems for classification of follicular thyroid tumor at a preoperative stage. STUDY DESIGN: Fine needle aspiration biopsy of the thyroid gland with a histologic conclusion of follicular cancer and follicular adenoma were the object of the morphometric study. General sample size was 4500 nuclei and 3000 aggregates. RESULTS: Quantitative regularities of pathologic changes in thyrocyte nuclei and aggregates in follicular cancer and follicular adenoma were revealed. Threshold values and weighting coefficients of quantitative features of thyrocyte nuclei and aggregates characterizing cancer made the basis of the two expert systems. Expert systems included standard 2-D S-matrix containing threshold values of nuclei and aggregates in cancer and their weighting coefficients as well as 1-D scientific X-matrix designed for filling with quantitative features of the studied object. The diagnosis was verified by the value of a diagnostic index by means of comparing feature values in the corresponding elements of S- and X-matrices. After that, a diagnostic index was calculated taking into account the features' weighting coefficient. CONCLUSION: The developed expert systems based on a set of quantitative features of thyrocyte nuclei and aggregates will allow assessing the malignant potential of a follicular thyroid tumor at a preoperative stage.