Abstract

The objective was to investigate the texture in the axillary sentinel node (ASN) in 100 patients with breast cancer (BC), in order to explore its role in differentiating a normal node from a metastatic one. In the ASN image, texture categories were extracted using the MaZda software. The histopathological diagnosis of ASN was negative (non-metastatic) in 85 cases and positive (metastatic) in 15 cases. The mean values of 19 texture parameters belonging to the histogram, co-occurrence matrix, and wavelet, were significantly higher in the negative ASN compared to the positive ones. Using WEKA software and the RandomForest model, negative nodes were differentiated from positive nodes with a sensitivity of 97%, specificity of 82%, and accuracy of 86%. In conclusion, the texture of the ASN images offers quantitative, non-invasive information that predicts the histopathology of the nodes in patients with BC.