Validation of rigid registration of mammographic images

Mammographic image registration is an important step in the analysis of differences between the left and right breast in order to detect bilateral asymmetry, which is an early sign of breast cancer. We have used rigid registration to align left and right mammographic images as a first step for comparing corresponding regions in the left and right breast. Registration performance was tested on a dataset consisting of normal and asymmetric cases obtained from three different mammographic image databases: mini-MIAS, DDSM and our own database containing digital mammographic images. For the validation of mammographic image registration, we have used several measures: sum of squared differences (SSD), normalized cross correlation (NCC), mutual information (MI), normalized mutual information (NMI), structural similarity index (SSIM) and universal image quality (UIQI) measure. Results showed that rigid registration works well on all three databases and that it might be a good first step in analysing bilateral asymmetry.