Characterizing the major sonographic textural difference between metastatic and common benign lymph nodes using support vector machine with histopathologic correlation

摘要: Sonographic texture analysis can reflect histopathological components and their arrangement in metastatic common benign lymph nodes. It is helpful differentiation between node lesions for target selection during biopsy of multiple nodes the strategy management. Two ultrasound systems, 107 sonographic regions interest (ROIs) metastases 174 ROIs nodes, were recruited study. Thirteen features derived from co-occurrence matrix used characterization above ROI images. Support vector machine (SVM) was as a classifier feature selector. The experimental results show that entropy gains best cross-validation accuracy 94.66% 87.73% both systems 1 2 classification disease. be further increased to 97.86% 100% by combination sum average There are significantly higher values than which due heterogeneous compositions larger cancer cells, lymphocytes, stroma contrast with simple inflammatory cells infiltration