Recent studies showed that it is not to be overlooked that the disagreement between radiologists when it comes to their diagnosis of screening mammogram with BI-RADS. There are some factors, such as patient's age and breast density, well-recognized that influences the performance of screening mammography. In this project, we further include other factors and analysis the effect their effects, especially features related with radiologists, such as their years of experience. Furthermore, based on those information, machine learning models are trained to denoise the labels for biopsy unrelated exams.