Please download the latest release (Version 1) at our registration/download page.
DigiBreast is a numerical breast phantom designed for 3D multi-physics simulations and validations of model-based image reconstruction algorithms for mammographically compressed breasts. The development of this phantom was described in Deng2015 with the original intent of testing a structure-prior guided image reconstruction algorithm for combined mammography and diffuse optical tomography (DOT) imaging. This digital breast phantom contains generic information such as 3D breast shapes and internal anatomical structures. We believe such breast phantom can address the needs for simulation-based validations for a wide range of model-based imaging modalities. Potential utility of this digital phantom include, but not limited to, simulations of breast deformation, 2D and 3D x-ray breast imaging, and tomographic imaging of a compressed breast using optical, microwave, thermal and electrical impedance methods.
A unique aspect of this digital breast phantom is the inclusion of a realistic 3D glandularity map measured through a dual-energy x-ray mammography system, provided by Philips Healthcare. In comparison, conventional numerical breast phantoms represent various breast tissue constituents, i.e. the fibroglandular and adipose tissue, by piece-wise-constant regions (using a binary segmentation algorithm). Such representation removes the fine spatial details in the breast anatomical images, and results in loss of information. Statistical, or fuzzy segmentation methods avoids such information loss, and provides spatially-varying tissue volume fraction maps. In our previous works Fang2010, we have reported a joint x-ray/DOT image reconstruction algorithm utilizing a spatially varying tissue compositional model to improve DOT image resolution. This method was further characterized in Deng2015.
The DigiBreast phantom has limitations. While the breast shape is in 3D, the internal tissue compositional maps were derived from 2D x-ray measurements, thus, have an overall "cylindrical" shape along the sagittal direction. We, however, believe this approximation has negligible impact to most potential applications which deal with a mammographically compressed breast. This is because most of these methods utilizing a parallel-plate based measurement scheme and such scheme has an anisotropic spatial resolution - the horizontal/axial has significantly higher resolution than in the vertical/sagittal direction. Therefore, the focus in most of these imaging modalities are in the axial/horizontal view instead of the sagittal view. This limitation can be overcome in the future when 3D x-ray spectral imaging becomes available.
DigiBreast ├── AUTHORS.txt # Acknowledgement of contributions ├── README.txt # This file ├── data # DigiBreast data in MATLAB .mat format │ └── DigiBreast.mat # DigiBreast main data │ └── DigiBreast_source.mat # DigiBreast source data ├── script # All related MATLAB scripts │ ├── digibreast_meshrefine.m # Creating the refined meshes at a given ROI │ ├── digibreast_savejson.m # Saving DigiBreast data in JSON and UBJSON │ ├── digibreast_priors.m # Creating tissue compositional maps │ ├── digibreast_lesionprofile.m # Creating a Gaussian-spherical tumor prior │ └── digibreast_tablelookup.m # Utility to lookup the optical properties └── LICENSE_BSD.txt # License file
DigiBreast.mat is a MATLAB mat-file that containing all essential components of the 3D digital breast phantom used in the simulation study as presented in Deng2015 paper, including variables ForwardMesh, ReconMesh, LesionCentroids, and OpticalProperties. This phantom is built on the source images included in DigiBreast_source.mat, and a 2 cm slab was added toward the chest wall.
DigiBreast_source.mat is a MATLAB mat-file that contains the anonymized and down-sampled (to 1 mm pixel resolution) 2D images of the original mammogram, glandularity and thickness maps. All original images are clinical measurements of a real patient breast in cranio-caudal view (CC view) using a Philips dual-energy mammographic system -- MicroDose SI. The MAT-file includes variables Mammogram, Glandularity, ThicknessMap, and Registration. Users can choose to use our readily-built 3D DigiBreast phantom in DigiBreast.mat, or to use the 2D source images offered in DigiBreast_source.mat to customize their own 3D realistic breast phantoms.
% example: % to generate a Gaussian lesion profile that represents the volume fractions % of a 5mm FWHM lesion located within the adipose vicinity as shown in %[Deng2015] on the forward mesh node=ForwardMesh.node; centroid=LesionCentroids.adipose; fwhmsize=5; lesionprofile=digibreast_lesionprofile(node,centroid,fwhmsize); figure; plotmesh([ForwardMesh.node lesionprofile],ForwardMesh.elem,'z=15','linestyle','none'); colorbar;
% example: % to generate the refined mesh used in [Deng2015] (see Table 1 for details) mesh=ForwardMesh; mesh_refined=digibreast_meshrefine(mesh,LesionCentroids.adipose,10,0.1); plotmesh(mesh_refined.node,mesh_refined.elem,'z=15','facecolor','w'); reconmesh_refined=digibreast_meshrefine(ReconMesh,LesionCentroids.adipose,10,1); % interpolation of glandularity maps in the refined mesh mesh.value=[ForwardMesh.glandularity.truth ForwardMesh.glandularity.dualgaussian]; mesh_refined=digibreast_meshrefine(mesh,LesionCentroids.adipose,10,0.1); figure; subplot(121); plotmesh([mesh_refined.node mesh_refined.value(:,1)],mesh_refined.elem,'z=15',... 'linestyle','none');colorbar subplot(122); plotmesh([mesh_refined.node mesh_refined.value(:,2)],mesh_refined.elem,'z=15',... 'linestyle','none');colorbar
% example: % to generate 2-compositional normal tissue priors using glandularity map % derived from dual gaussian segmentation algorithm priors=digibreast_priors(ForwardMesh.glandularity.dualgaussian); figure; subplot(211); plotmesh([ForwardMesh.node priors.normal(:,1)],ForwardMesh.elem,'z=15','linestyle','none'); title('Adipose tissue volume fractions');colorbar; subplot(212); plotmesh([ForwardMesh.node priors.normal(:,2)],ForwardMesh.elem,'z=15','linestyle','none'); title('Fibroglandular tissue volume fractions');colorbar; % to generate 3-compositional normal and lesion tissue priors using the % same glandularity map derived from dual gaussian segmentation algorithm lesionprofile=digibreast_lesionprofile(ForwardMesh.node,LesionCentroids.adipose,5); priors=digibreast_priors(ForwardMesh.glandularity.dualgaussian,lesionprofile); figure; subplot(311); plotmesh([ForwardMesh.node priors.lesion(:,1)],ForwardMesh.elem,'z=15','linestyle','none'); title('Adipose tissue volume fractions');colorbar; subplot(312); plotmesh([ForwardMesh.node priors.lesion(:,2)],ForwardMesh.elem,'z=15','linestyle','none'); title('Fibroglandular tissue volume fractions');colorbar; subplot(313); plotmesh([ForwardMesh.node priors.lesion(:,3)],ForwardMesh.elem,'z=15','linestyle','none'); title('Lesion tissue volume fractions');colorbar;
% example: Sp_fib=digibreast_tablelookup(OpticalProperties,'fib','s_power');
With the provided fibroglandular tissue volume fraction map (variable Glandularity within DigiBreast_source.mat), users can freely build your own breast phantom by multiplying the optical properties of fibroglandular, adipose, and cancerous tissues to the volume fractions at each pixel/node.
Tissue type | HbO (μM) | HbR (μM) | μs’ (mm−1) | |
---|---|---|---|---|
at 690 nm | at 830 nm | |||
Adipose | 13.84 | 4.81 | 0.851 | 0.713 |
Fibroglandular | 18.96 | 6.47 | 0.925 | 0.775 |
Malignant | 20.60 | 6.72 | 0.957 | 0.801 |
The DigiBreast phantom main and source data is in the public domain. The MATLAB scripts under the "script" sub-folder have a BSD license. See LICENSE_BSD.txt for details.
Some of the scripts included in this package requires the installation of the "iso2mesh" and "JSONLab" toolboxes. To download these toolboxes:
If you use this DigiBreast phantom main or source data in your publication, please cite the phantom version number (currently Version_1) to avoid conflict to any further updates of this mesh. We are also appreciated if you can cite the Deng2015 paper below.
[Deng2015] B. Deng, D.H. Brooks, D.A. Boas, M. Lundqvist, and Q. Fang, "Characterization of structural-prior guided optical tomography using realistic breast models derived from dual-energy x-ray mammography," Biomedical Optics Express 6(7): 2366-79 (2015).
[Fang2010] Q. Fang, R.H. Moore, D. B. Kopans, D.A. Boas DA, “Compositional-prior-guided image reconstruction algorithm for multi-modality imaging,” Biomedical Optics Express, 1(1), 223-235 (2010)
[Fang2011] Q. Fang, J. Selb, S.A. Carp, G. Boverman, E.L. Miller, D.H. Brooks, R.H. Moore, D.B. Kopans and D.A. Boas, "Combined optical and X-ray tomosynthesis breast imaging," Radiology 258(1): 89-97 (2011).