A comparative study of the inter-reader variability of breast percent density estimation in digital mammography: potential effect of reader's training and clinical experience

The variability of breast percent density (PD%) estimation from digital mammography (DM) images was evaluated using measurements from readers with different training and clinical experience Post-processed DM images (PremiumViewTM, GE Healthcare)from 40 women were analyzed Breast PD% estimation was performed using the Cumulus software (Ver 4.0, Univ Toronto) Two groups of readers were considered, one with clinical (i.e., radiologists) and one with non-clinical training (i.e., physicists) Consistency of PD% was analyzed using the Pearson correlation coefficient (r) and ANOVA Inter-reader agreement was higher among clinical (r=0.91, p<0.001), than non-clinical readers (r=0.83, p<0.001) Intra-reader consistency after repeated reads was on average equally high for both groups (r=0.91, p<0.001) Our results suggest that the reader's experience and training has an effect on the obtained PD% measures The higher correlation among the clinically trained readers could be attributed to their extensive exposure to post-processed DM images and their knowledge of breast anatomy.