| Resumo: | Computer-Aided Systems can assist differentiation and classification of breast benign and malignant lesions and enhance the performance of breast cancer diagnosis. Breast lesions are strongly correlated with their shape: benign lesions present regular shapes, although malignant lesions tend to present irregular shapes. Therefore, the use of quantitative measures, such as fractal dimension (FD), can help characterize the smoothness or the roughness of the lesion shape. The main purpose of this work is to assess if the concomitant use of FD measure (contour FD) with a proposed FD measure (area FD) can improve the classification of lesions according to the BIRADS (Breast Imaging Reporting and Data System) scale and lesion type. Both FD measures were calculated through the box-counting method, directly from manually segmented lesions, and after applying a region growing/erosion algorithm. The last FD measure is based on the normalized difference between the FD measures before and after the application of the region growing/erosion algorithm. Results indicate that the contour FD is a useful measure in the differentiation of lesions according to the BIRADS scale and type, although, in some situations, errors occur. The combined use of contour FD with the four proposed FD measures can improve the classification of lesions.
|
|---|