General Departmental Seminar Series
A Bayesian Network for Mammographic Diagnosis
Charles Kahn, Medical College of Wisconsin
Friday, November 3, 2000, 4:00 pm
1325 Computer Science & Statistics Center
1210 W. Dayton St.
Breast cancer is the second-leading cause of cancer mortality in American women. Screening mammography can increase the likelihood of early detection and cure, but it can be difficult to differentiate benign and malignant lesions. We developed a Bayesian network model to assist radiologists in the diagnosis of breast cancer. Bayesian networks are graphical representations of joint probability distributions, and allow reasoning under uncertainty. Our model, MammoNet, incorporates five patient-history features, two physicial findings, and 15 mammographic features. MammoNet is a potentially useful tool that can improve the diagnostic performance of non-expert radiologists and radiology residents.
Back to General Departmental Seminar Series