DDSM Converter

Convert DDSM LJPEG and generate CSV annotation file.

DDSM (Digital Database of Screening Mammogram)

Source: http://www.eng.usf.edu/cvprg/Mammography/Database.html

Dependencies

• Python 3+
• OpenCV-Python
• Pandas
• Numpy

pip install -r requirements.txt

How to use

1.Build jpeg

cd jpegdir; make

2.Run convert.py:

python3 convert.py --input path/to/ddsm/folder --output folder/path/to/store/output --format png --correction

Argparse:
--correction apply correction to avoid black images and noisy image (recommended).
--normalization apply opencv normalization.
--format file extension (opencv supported format only).

Example I/O

Input directory (tree):

path/to/ddsm/folder
└── cases
    ├── benigns
    │   └── benign_01
    │       └── case0029
    │           ├── C-0029-1.ics
    │           ├── C_0029_1.LEFT_CC.LJPEG
    │           ├── C_0029_1.LEFT_CC.OVERLAY
    │           ├── C_0029_1.LEFT_MLO.LJPEG
    │           ├── C_0029_1.LEFT_MLO.OVERLAY
    │           ├── C_0029_1.RIGHT_CC.LJPEG
    │           ├── C_0029_1.RIGHT_MLO.LJPEG
    │           └── TAPE_C_0029_1.COMB.16_PGM
    ├── cancers
    │   └── cancer_01
    │       └── case3086
    │           ├── B-3086-1.ics
    │           ├── B_3086_1.LEFT_CC.LJPEG
    │           ├── B_3086_1.LEFT_CC.OVERLAY
    │           ├── B_3086_1.LEFT_MLO.LJPEG
    │           ├── B_3086_1.LEFT_MLO.OVERLAY
    │           ├── B_3086_1.RIGHT_CC.LJPEG
    │           ├── B_3086_1.RIGHT_MLO.LJPEG
    │           └── TAPE_B_3086_1.COMB.16_PGM
    └── normals
        └── normal_10
            └── case3660
                ├── B-3660-1.ics
                ├── B_3660_1.LEFT_CC.LJPEG
                ├── B_3660_1.LEFT_MLO.LJPEG
                ├── B_3660_1.RIGHT_CC.LJPEG
                ├── B_3660_1.RIGHT_MLO.LJPEG
                └── TAPE_B_3660_1.COMB.16_PGM

Output directory (tree):

folder/path/to/store/output
├── benigns
│   ├── C_0029_1.LEFT_CC.jpg
│   └── C_0029_1.LEFT_MLO.jpg
├── benigns.csv
├── cancers
│   ├── B_3086_1.LEFT_CC.jpg
│   └── B_3086_1.LEFT_MLO.jpg
├── cancers.csv
└── normals
    ├── B_3660_1.LEFT_CC.jpg
    ├── B_3660_1.LEFT_MLO.jpg
    ├── B_3660_1.RIGHT_CC.jpg
    └── B_3660_1.RIGHT_MLO.jpg

CSV:

NAME,FEATURE,SEVERITY,X1,Y1,X2,Y2,HEIGHT,WIDTH

Reference

• nicholaslocascio/ljpeg-ddsm
• fjeg/ddsm_tools