OBJECTIVE: To assess the performance of morphometric parameters calculated from breast lesion contours at ultrasonography, in the quantification of morphological features and in the distinction among BI-RADS categories 2, 3, 4 and 5. MATERIALS AND METHODS: The present casuistry included 40 cases, with orthogonal imaging of patients submitted to surgery. Based on the lesions segmentation, the following five morphometric parameters were calculated: area ratio, overlap ratio, normalized residual value, circularity and depth-to-width ratio. Linear discriminant analysis was applied to select the best parameters, and ROC curve was utilized as figure of merit. RESULTS: Overlap ratio was able to statistically distinguish breast lesions classified as BI-RADS 3 from others classified as BI-RADS 4 (a = 5%; p = 0.015). Additionally, such parameter presented the best performance in the differentiation between malignant and benign breast lesions. CONCLUSION: These results indicate that morphometric analysis of breast lesions at ultrasonography seems to be helpful in distinguishing patients who should undergo biopsy from those to be followed-up only with imaging methods.
Breast ultrasonography; BI-RADS; Computer-assisted image processing; Computerassisted image interpretation
