Cancer Imaging Phenomics Toolkit (CaPTk)  1.6.1
Breast Cancer: Breast Density Estimation (LIBRA)

The Laboratory for Individualized Breast Radiodensity Assessment (LIBRA) is a software application for fully-automated breast density segmentation in full-field digital mammography (FFDM) [1].

LIBRA is available in single and batch mode.


  • Raw (i.e., "FOR PROCESSING") or vendor post-processed (i.e., "FOR PRESENTATION") FFDM images.
  • FFDM vendors currently supported: GE Healthcare and Hologic.
  • For single mode: Each image stored in a separate folder.
  • For batch mode: All images stored in a single folder.


  • Single mode:
    1. Load the FFDM image using the 'File -> Load -> Images' menu option.
    2. Launch LIBRA using the 'Applications -> LIBRA_SingleImage' menu option.
    3. LIBRA is executed and the segmentation mask is automatically loaded back to CaPTk.
    4. LIBRA outputs (described in the Outputs section below) are automatically saved in a temp folder under the CaPTk directory.
  • Batch mode:
    1. Launch LIBRA using the 'Applications -> LIBRA_BatchMode' menu option.
    2. The user is prompted to select the folder which contains the images for analysis.
    3. The user is then prompted to select an output folder where the density scores and density segmentation images .jpg files will be stored. Note: If the folder already exists and contains density estimates (e.g., from a previous run of LIBRA), it will append the new results to the results file.
    4. Lastly, the user will be asked whether they wish to store a processing log file and intermediate graphics in addition to the density segmentation outline which may be on interest for publication or visualization purposes, and are further described in the Outputs section below.
    5. LIBRA will then begin processing all the FFDM images; the progress of the software can be monitored via the related command-prompt window that opens when LIBRA starts running. When analysis is complete, a prompt appears asking the user if they want to open the results folder and if they want to perform additional analyses.


  • The following outputs are always generated:
    • Masks_<Image-Analyzed>.mat: A MATLAB datafile containing a structure array that stores the breast area (res.BreastArea), dense area (res.DenseArea), and area percent density estimates (res.PD_SVM) for the FFDM analyzed (res.dcm_fname). This structure array also stores the binary masks of the breast (res.BreastMask) and dense tissue segmentations (res.DenseMask), which may be useful for further processing and analysis.
    • Density.csv: A comma separated file (open-able by spreadsheet programs like Excel) that stores the breast area, dense area, and percent density estimates for each FFDM image analyzed, listed by file name. The Manufacturer, Laterality and View-Position of the mammogram are also provided for reference. (Note that the CSV file should be in ASCII format)
    • <Image-Analyzed-Filename>_density_segmentation.jpg: In the Result_Images sub-directory, the breast and density segmentation results are provided for each image analyzed. The breast is outlined in red, the density segmentation is in green. Note that for visualization purposes, the image is window-leveled to between the 5th and 95th percentile of the intensity values of the pixels within the breast region.
  • In single mode the following outputs are generated by default, while in batch mode they are only generated if the user specifies to save intermediate files:
    • LIBRA-logfile_<time-stamp>.txt: A log of the programs outputs during the session, this text file is time-stamped at the start of the LIBRA sessions so that multiple sessions writing to the same output folder each have their own unique log of events. The <time-stamp> takes the form of <Month>--<YYYY>_HH-MM-SS, in 24-hour format. .. _fig_log_file.
    • <Image-Analyzed-Filename>_Windowed_Original.jpg: In the Result_Images sub-directory, the analyzed image window-leveled to between the 5th and 95th percentile of the intensity values of the pixels within the breast region is also provided with out the breast density segmentation overlay for comparison purposes.
    • <Image-Analyzed>_density_imagesc.jpg: In the Result_Images sub-directory, the image clusters (grouped by colors) generated by the Fuzzy C-Means stage of the LIBRA algorithm is provided.
    • <Image-Analyzed>_intensity_histogram.jpg: In the Result_Images sub-directory, the breast intensity histogram (z-scored) and FCM-cluster centers are plotted in this image.

Please see the LIBRA manual for more details:

  • This application is also available as with a stand-alone CLI for data analysts to build pipelines around, using the following example command:
    LIBRA.exe --input C:/inputDICOMDir --output C:/outputDir


  1. B.M.Keller, D.L.Nathan, Y.Wang, Y.Zheng, J.C.Gee, E.F.Conant, D.Kontos, "Estimation of breast percent density in raw and processed full field digital mammography images via adaptive fuzzy c-means clustering and support vector machine segmentation", Med Phys. 39(8):4903-4917, 2012. DOI:10.1118/1.4736530