A new collection of preprocessed digital mammograms

Abstract

We contribute with a publicly available repository of digital mammograms including both raw and preprocessed images. The study of mammographies is the most used and effective method to diagnose breast cancer. It is possible to improve quality of images for more accurate predictions of radiologists, by applying some preprocessing techniques. In this work we introduce a method for mammogram preprocessing. Our method includes the following processes: reduction of the work area, bit conversion, denoising using the adaptive median filter, contrast enhancement using histogram equalization, and image compression using histogram shrinking. Practical experiments were conducted on raw images in DICOM format from the Mammography Clinic at Hospital de Alta Especialidad Juan Graham Casasús located in Tabasco, Mexico. Results were evaluated by medical doctors. © Springer-Verlag 2013.