Multiple logistic regression analysis of tumor shape features for ultrasound breast cancer diagnosis

The implementation of ultrasound image diagnosis greatly depends on the experience of physicians. Therefore, the diagnostic process is highly depends on the physician, and is also lacking the experience sharing media and reference standard in the diagnostic process. In order to establish the experience-sharing platform, in this research, we developed a computer-aided diagnosis system based on the tumor shapes. To establish multivariate analysis model of tumor shape features for ultrasound of breast cancer diagnosis, we evaluated 123 ultrasound images of pathologically proven solid breast tumors (76 infiltrative ductal carcinomas and 47 fibroadenomas). The shapes of tumors were classified and coded by experienced radiologists, and a multiple logistic regression algorithm was used to classify the tumor as benign or malignant. The accuracy of this model for classifying malignancies was 92.7%, the sensitivity was 89.5%, and the specificity was 97.9 %. According to the experience of the radiologists and the regression analysis algorithm, we established a computer-aided diagnosis system applied to ultrasound of solid breast tumors. This system helps differentiate solid breast tumors with relatively high accuracy and can be of significant aid for inexperienced operators to avoid misdiagnoses.