RAI Labs Duke

Anthropomorphic Virtual Breast Phantoms

The first group of phantoms was publicly released on July 1, 2016! For more information, please see the following publication:

Erickson DW, Wells JR, Sturgeon GM, Samei E, Dobbins JT 3rd, Segars WP, Lo JY, "Population of 224 realistic human subject-based computational breast phantoms," Med Phys, 43: 23 (2016). PMC4684566 [ext. link]

Researchers who are interested in acquiring the data may email

Mammography images that were simulated from our virtual phantoms. Breast density increases from left to right.

Each of the Duke XCAT breast phantoms is a 3D computational model that is based on the dedicated breast CT scans of an actual human subject, which is why they are sometimes referred to as "patient-based phantoms." From those scans, the data undergoes a process of artifact correction, denoising, and segmentation. The phantoms are voxelized with voxel values corresponding to tissue classes. Given the resulting 3D volume, the user may assign voxel values corresponding to different imaging modalities, such as attenuation coefficients for radiography. Upon request, the breast phantoms may also be distributed in a mesh surface format. The phantoms are available in their original prone, uncompressed position as well as the standing, compressed position for FFDM/DBT. Other positions may also be available upon request.

There are two distinct cohorts of these phantoms. Cohort 1.x consists of up to 224 cases provided by John Boone Ph.D. of UC Davis, and those phantoms are now nearly complete and publicly available. Cohort 2.x is currently a work in progress with up to 140 cases provided by Steve Glick Ph.D. of Univ of Massachusetts / FDA, and those phantoms are still under active development but partial beta releases are possible. Unless otherwise noted, the rest of this document pertains only to Cohort 1.x.

This work was supported in part by NIH/NCI R01 CA134658, Principal Investigator Paul Segars. Continued development of these phantoms is co-directed by Paul Segars and Joseph Lo of the Carl E. Ravin Advanced Imaging Laboratories (RAI Labs), Department of Radiology, Duke University School of Medicine.