This application's broad, long-term objective is to lessen the disease burden of breast cancer by detecting it at an early and curable stage and by reducing the number of biopsies on benign lesions. Computer-aided diagnosis (CAD) refers to a diagnostic process in which a radiologist uses a computer analysis of a mammogram as a diagnostic aid to achieve more accurate interpretation. The hypothesis to be tested is that with optimization and clinical trial, CAD methods that classify breast lesions as malignant or benign can be used clinically.
The specific aims of this application are: (1) To compare two computer classification methods: one based on image features extracted by a computer and one based on the Breast Imaging Report and Data System (BI-RADS) lesion descriptors provided by radiologists; (2) To develop optimal strategies for radiologists to combine their diagnostic assessment with that of the computer; (3) To carry out a Phase II clinical trial and to plan a Phase III clinical trial. The significance and health-relatedness of CAD for breast lesion classification is that it can potentially help radiologists reduce the number of biopsies on benign lesions while maintaining or increasing the sensitivity of mammography. The significance and health-relatedness of this project is that it will increase clinical effectiveness of CAD through optimization, move CAD from laboratory research to clinical evaluation, and start a clinical trial process that will ultimately determine CAD's clinical efficacy. The research design is to optimize previously developed CAD methods and then to conduct a Phase II clinical trial. The methods to be used include lesion feature analysis, artif aboutcial neural networks (ANNs), receiver operating characteristic (ROC) analysis, observer study, mathematical modeling with respect to ideal observer performance, and Monte Carlo simulation.
Showing the most recent 10 out of 25 publications
