Application Of Ultrasound Image Feature In Breast Cancer Detection And Classification

Objective Breast cancer is the first leading cause of cancer death for women all over the world.Since the cause of the disease remains unknown, early detection and diagnosis is the key forbreast cancer control, and it can increase the success of treatment, save lives and reduce cost.Ultrasound imaging is one of the most frequently used diagnosis tools to detect and classifyabnormalities of the breast.The study aimed at exploring early diagnostic value of ultrasonicimaging features between breast benign and malignant masses, and in order to improve thediagnostic accuracy.Methods Two hundred and sixteen cases of breast masses diagnosed pathologically wereanalyzed retrospectively, one hundred and forty-four cases were benign, and the remainingseventy-two cases were malignant. The ultrasonographic parameters of benign and malignantmasses were recorded and compared using binary logistic regression analysis and Fisherdiscriminant analysis. The ROC curve was drawn and the areas under curves were calculated.Database was established by Epi Data3.1and analyzed by SPSS19.0.Results the binary logistic regression analysis showed boundary of benign nodules wassignificantly clearer than malignant nodules(P＜0.05). Malignant nodules manifested moreclearly crab foot or burr，without envelope compared with benign nodules(P＜0.01), rear echochange is echo enhancement or no change and attenuation, two groups of detection rate had asignificant difference (P＜0.05). Multiple factors discriminant analysis showed benign massappeare to have regular shape，smooth border，homogeneous internal echoes，no rear attenuation.While the malignant tumors usually have irregaler shape and border,low echoic, no attenuation.The ipsolateral axillary lymph nodes were enlarged. There were4variables entered the equationwhen using Stepwise Methods. The function of forecasting model was:Y=-6.670+1.105X3+2.673X4+0.277X6+0.380X7. The sensitivity and specificity of self-validatedof training sample was93.1%and96.5%, repectively. The sensitivity and specificity ofcross-validated of training sample was90.3%and95.8%, repectively, while The fingings oftesting sample showed the sensitivity and specificity was88.5%and90.5%, repectively. Thearea under the ROC curve was0.988.Conclusions The present study have shown there were a association between the morphologyfeatures of the ultrasound imaging and classification of breast mass. Ultrasound imaging was one of diagnosis tools to detect and distinguish benign and malignant breast tumors. Logisticregression model and discriminant analysis model was a valuable and beneficial means for breastcancer detection and classification...