Database dedicated to X-ray breast imaging

Abstract

Breast cancer computational models are a key instrument used in the development and optimization of new breast imaging techniques, new realistic test models for X-ray breast dosimetry, as well as reconstruction and image improvement algorithms. This requires the availability in one place of a large number of different breast cancer models and X-ray images from test objects. This work summarizes the types of lesions and X-ray images stored in the MAXIMA database (http://maxima.tu-varna.bg/). The database consists of data and images related to the breast. More specifically, it contains X-ray images from various scientific studies carried with physical phantoms with varying properties and shapes (including anthropomorphic) obtained from different facilities. These include synchrotron facilities, research facilities with breast computed tomography (CT) and hospitals with different brands of CT and mammography machines. Furthermore, the database contains a number of anonymized X-ray patient images from tomosynthesis and mammograms, as well as whole body CT sets. Some of the images contain both benign and malignant formations, while others are lacking any kind of pathology. Besides the X-ray images the database also contains two types of breast computational lesion models. The first type is computational breast models, segmented from patient tomosynthesis images realized by using an in-house developed algorithm, while the second type of 3D models is created by applying mathematical algorithm based on a random walk approach. Currently, the database contains more than 70 images from breast tomosynthesis and 4 sets from whole body CT; 2 sets from scanned mastectomy cadavers, 50 segmented tumor models and 100 models generated by a mathematical algorithm. In addition, the database contains more than 50 different entries, which correspond to physical breast phantoms and step manufactured wedges. The developed database provides opportunities to researchers to study and work on improving techniques for early cancer detection.