Role of morphometry in evaluation of cytologically borderline breast lesions: a study of 70 cases.

Fine needle aspiration, an accepted minimally invasive diagnostic procedure for breast masses, leads to an indeterminate (borderline) diagnosis in a significant proportion of cases. This study was aimed at evaluation of the utility of image analysis and morphometry in the cytologic assessment of borderline breast cases. A 'study' group of 45 cases (including fibroadenoma, fibrocystic change and ductal carcinoma) and a 'test' group of 25 cytologically borderline cases were included. Nuclei from each of the groups (nuclei of monolayer sheet of fibroadenoma, naked nuclei of fibroadenoma, ductal cells nuclei of fibrocystic change and nuclei of carcinoma) were manually traced and analyzed using self-written image analysis software. Discriminant analysis was done on the 'study' group and the resultant discriminant functions were applied to the 'test' group. The percentage of nuclei correctly classified of the study as well as test group was noted. On step-wise discriminant analysis, various morphometric, textural parameters and color information parameters were selected. Of the original 'study' group, 98.49% of nuclei could be correctly classified. No overlapping of benign and malignant nuclei was observed in the 'study' group. When the resulting functions were tested on the 'test' group, all the cases were correctly classified as benign or malignant (compared with the final histologic diagnosis). This study suggests that application of computerized image analysis to cytologically borderline breast cases can help in the classification of these cases as benign or malignant. This may prevent unnecessary surgery in many of these cases.