A new mammography database of full-field digital images has been developed by a Portuguese group for public use by researchers. It is described in the February issue of Academic Radiology.
The INbreast database was developed by radiologists at the Centro Hospitalar de São João (CHSJ) in Porto with the objective of providing a new reference for future work relating to breast cancer imaging (Acad Radiol, February 2012, Vol. 19:2, pp. 236-248).
The core database consists of images from 115 mammography exams performed between April 2008 and July 2010 using a full-field digital mammography system (MammoNovation, Siemens Healthcare). It includes 90 cases of four images each and 25 cases of two images each of women who had a mastectomy.
The database includes examples of normal mammography images and mammographic images showing masses, calcifications, architectural distortions, asymmetries, and multiple findings. The images were annotated by an experienced mammographer and validated by a second breast specialist.
Seven types of annotations were made, according to lead author Inês C. Moreira, a master's degree candidate at the Universidade de Porto. They include asymmetry, calcification, cluster, mass, distortion, speculated region, and pectoral muscle. Accurate contours also are provided in Extensible Markup Language (XML) format.
The database also includes anonymized information about the patient, including family history, the age at the time of the mammogram, and BI-RADS classification. Biopsy results also are included for cases in which they were performed. In the core database, this included 56 cases, of which 45 proved to be malignant.
The developers of the INbreast database plan to expand it. It may be accessed here.
