유방촬영검사의 디지털 데이터베이스를 이용한 컴퓨터보조진단

Abstract

The purpose of this study was to propose robust schemes in terms of mass diagnosis and to objectively describe the confounding effect on cancer risk. For the experiment of mammographic masses, 249 cases of malignant (case group) and 307 cases of benign (control group) from the Digital Database for Screening Mammography (DDSM) were selected. Each radiological feature was categorized and then performed for the multiple logistic regression analysis to reveal the interaction to the probability of a malignant tumor. For 95% confidence interval and high p-value, these analyses significantly showed that the effects on incidence rate of malignant tumor were complexly associated in the order of margins, age, size, shape and breast tissue density. Our method may be useful for identifying breast cancer in mammography and developing computer-aided diagnosis as a solution under the PACS environment.