Learning vector quantizer in the investigation of thyroid lesions

OBJECTIVE: To investigate the capability of the learning vector quantizer ( LVQ) in the discrimination of benign from malignant thyroid lesions. STUDY DESIGN: The study was performed on May-Grunwald-Giemsa-stained smears taken by fine needle aspiration (FNA). Using a custom image analysis syste m, 25 features that describe the size, shape and texture of approximately 1 00 nuclei were measured from each case. Statistical features were extracted from each case and a linear regression analysis was performed to detect th e statistically significant features. The cases were distributed by categor y, as follows: 100 cases of goiter and follicular adenomas, 11 cases of fol licular carcinoma, 35 cases of papillary carcinoma, 24 cases of oncocytic a denoma, 8 cases of oncocytic carcinoma and 20 cases of Hashimoto thyroiditi s. About 30% of the cases from each class were used for training two LVQ cl assifiers. The remaining 139 cases, out of a total of 198, were used as the test set. A classifier was used to discriminate into Jour classes and a se cond into two classes. RESULTS: The application of LVQ neural networks allows good discrimination between benign and malignant lesions (O.A. = 97.8). However, reliable discr imination of the cytologic types of the lesions was not obtained. CONCLUSION: These results indicate that the use of neural networks combined with image morphometry may offer useful information on the potential for m alignancy of thyroid lesions and may improve the diagnostic accuracy of FNA of the thyroid gland, especially in cases of follicular neoplasms classifi ed as suspicious for malignancy and in cases of oncocytic tumors.