Computer-aided Diagnosis Scheme for Detection of Mass with Bilateral Subtraction Technique in Mamograms

抄録

Our purpose of this study was to develop a Computer-aided Diagnosis (CAD) scheme for detection of masses by a bilateral subtraction technique which subtracts normal breast tissues between right and left breasts. In our CAD scheme, breast regions were first segmented from each of right and left mammograms. Pixel values were transformed to uniform the histograms based on the pixel values of right and left breasts. The nodular structures such as mass were enhanced by the analysis of Hessian matrix. Right and left breasts were aligned more accurately by two steps of the alignment based on skin lines and the alignment based on normal breast structures as blood vessels and mammary glands. The candidates for mass were then detected by the difference in normal breast tissues between right and left breasts. The detection performance of our CAD scheme was evaluated by using 256 mammograms which were selected randomly from DDSM (Digital Database for Screening Mammography in the University of South Florida). Our CAD scheme achieved the sensitivity of 93.4% (239/256) at 22.2 (5683/256) false-positives per image.