Abstract
Objective : Morphometric analyses of fine needle aspiration smears from thyroid lesions were performed. The importance of the measured variables was evaluated and logistic regression equations were formulated to estimate the histology.
Study Design : Specimens from 28 papillary carcinoma (PC) s, 19 follicular carcinoma (FC) s, 23 follicular adenoma (FA) s, and 29 adenomatous goiter (AG) s were examined, and the nuclear area and its standard deviation (SD) within samples, nuclear perimeter, major/minor axis ratio, nuclear overlapping, and the number of nuclear layers were compared. The variables were selected to estimate the histology using a backward selection procedure. The probabilities for the histology were estimated using the medians and quartiles of the selected variables.
Results : The values of all the measured variables were higher in the malignant lesions than in the benign lesions. PCs showed the largest mean values for the nuclear area, its SD, the nuclear perimeter, the major/minor axis ratio, and nuclear overlapping. FCs showed the largest mean value for the number of nuclear layers. The nuclear area, major/minor axis ratio and nuclear overlapping were selected for the estimation of the histology of PC and AG, and the number of nuclear layers was selected for that of FC and FA. The quartiles gave the probability of 93% for PC, 98% for AG, 65% for FC, and 45% for FA.
Conclusion : Logistic regression equations could provide high estimated probabilities for the diagnosis of PC and AG.
