engleski

Parotid Gland Tumors: Correlation Between Routine Cytology and Cytomorphometry by Digital Image Analysis

Background: The objective of this study was to compare qualitative cytomorphology and morphometric characteristics of parotid gland tumor cells, with the aid of a computer-assisted system of image analysis. Methods: Routine qualitative cytologic and quantitative morphometric results from 64 parotid gland tumors were compared. Ultrasound (US)-guided fine-needle aspiration (FNA) specimens were taken from 54 patients. Eleven conventionally used morphometric parameters were studied: area, perimeter, convex area, convexity, maximal and minimal radius, length, breadth, form factor (FF), elongation factor and nuclear-cytoplasmatic (N/C) ratio. Two newly introduced nuclear form factors were also measured: area symmetry factor and perimeter symmetry factor. Results: The following nuclear morphometric parameters were significantly different between malignant and benign tumors: area, perimeter, convex area, convexity, maximal and minimal radius, length, breadth, FF, elongation factor, area symmetry factor and perimeter symmetry factor. To increase the usefulness and practicality of an image analysis in clinical setting, cutoff values of each nuclear morphometric parameters were calculated. Comparing the cutoff values and (receiver operating characteristic) ROC curves the following nuclear morphometric parameters were found most useful in separating benign from malignant tumors: area, perimeter, convex area, maximal radius, length and FF. With a cutoff value for nuclear area determined by an ROC curve (cutoff >61.411 lm2, sensitivity = 67.0%, specificity = 84.4%, PPV 69.3%, NPV 83.0) the tumors were separated into malignant (with averages that were higher than the cutoff) and benign (with smaller averages). Cutoff values for nuclear outline, maximal radius, convex area and length were >29.701 lm (sensitivity = 69.5.8%, specificity = 84.5%, PPV 70.2%, NPV 84.1%), >5.314 lm (sensitivity = 67.0%, specificity = 86.9%, PPV 72.8%, NPV 83.4%), >63.077 lm2 (sensitivity = 66.3%, specificity = 85.6%, PPV 70.1%, NPV 82.9%), and >10 038 lm (sensitivity = 69.9%, specificity = 82.0%, PPV 67.1%, NPV 83.9%), respectively, with averages that were higher than the cutoff in malignant tumors. Cutoff value for FF was £0.875 (sensitivity = 74.0%, specificity = 71.5%, PPV 57.7%, NPV 84.0%), with averages that were closer to the value of 1 (value for a circle) in benign tumors. The following whole cell morphometric parameters were significantly different between malignant and benign tumors: minimal and maximal radius, convexity, breadth, FF, and elongation factor. N/C ratio was significantly higher in malignant tumors. Conclusions: The quantitative morphometric analysis is a useful tool in the cytological differentiation between benign and malignant parotid gland tumors. Computerized image analysis may add to morphological evaluation by turning qualitative data into quantitative values.

Parotid Gland Tumors; Cytology; Cytomorphometry; Digital Image Analysis

nije evidentirano