A Computerized System to Assess Axillary Lymph Node Malignancy from Sonographic Images

摘要: A computational approach to classifying axillary lymph node metastasis in sonographic images is described. One hundred five ultrasound of nodes from patients with breast cancer were evaluated (81 benign and 24 malignant), each was manually segmented, delineating both the whole internal hilum surfaces. Normalized signed distance transforms computed segmented boundaries structures, pixel then assigned coordinates a 3-D feature space according pixel's intensity, its boundary boundary. Three-dimensional histograms over accumulated for by summing all pixels, bin counts served as predictor inputs support vector machine learning algorithm. Repeated random sampling 80/25 train/test splits used estimate generalization performance generate receiver operating characteristic curves. The optimal classifier had an area under curve 0.95 sensitivity specificity 0.90 0.90. Our results indicate feasibility nodal staging computerized analysis.