Evaluation of Morphometric Diagnosis Accuracy in Thyroid Cytology

Decision making in pathology is changing toward quantitative criteria. Quantitative measurements obviously can demonstrate minimal differences, not readily identifiable to human vision. Morphometry is the quantitative description of geometric characteristics of objects. This study aims at testing the feasibility of this new technique using available instruments and software and selecting useful parameters and their thresholds for discrimination of benign versus malignant lesions of thyroid. The cases were selected based on surgical pathology reports of fine needle aspiration thyroid specimens from the pathology department of Dr. Shariati hospital between October 2003 to March 2004. Patients whose surgical pathologic reports were unavaible were excluded. At least 100 cells, selected by a zone guided manner, were measured for each patient. Totally 67 patients were enrolled for the study, 36 with benign pathology diagnosis and 31 with malignant lesions. In a pilot study to test reproducibility of measurements, a single cell was selected and measured 20 times. Coefficient of variation remained below 2% for all measurements, except for nuclear diameter which has been excluded from analysis. Benign cells were different from malignant ones in mean nuclear perimeter (Peri), mean nuclear area (Area), nuclear area to perimeter ratio (A/P), cluster diameter, mean gray level (in 256 gray levels) and standard deviations of Peri, Area and A/P (P-values < 0. 05) but not in CV of nuclear area, intercellular standard deviation of gray levels, intracellular CV of gray levels and its standard deviation. The receiver Operator Characteristic curve analysis was performed for parameters with significant difference and sensitivities and specificities calculated for different cut-off values. The highest sensitivity was reached for nuclear perimeter, nuclear area, cluster diameter and nuclear area to perimeter ratio equal to 93%.55 and for cluster diameter equal to 89%.55%. The results are comparable to classic cytology practice, and this technique is recommended screening of thyroid FNA specimens.