README.md

DicomData

• findDCM.sh - Finds all DICOM files (excluding DICOMDIR files) in a directory and writes the absolute path to "DCM_files.txt"
• create_hdf5.py - Iterates over "DCM_files.txt" to find the last created file per protocol (right CC, left CC, right MLO, left MLO) per identification code, then writes each scan into an HDF5 file format called "DCM_data.hdf5".

DCM_data.hdf5 File Structure

• Root (File)

  • Identification codes (Groups)

    • Patient info (Dataset):

      • Array of integer and float values that describe:
        • Age
        • Height (m)
        • Type of exam
        • Breast size
        • Breast type
        • Particular conditions
        • Beauty marks
        • Exams compared with previous exam
        • Right CC scan acceptable
        • Left CC scan acceptable
        • Right MLO scan acceptable
        • Left MLO scan acceptable
        • Overall CC acceptable
        • Overall MLO acceptable
        • Overall acceptable
        • Weight (Kg)
        • BMI
      • Attributes dictionary uses the dataset's array index as keys and the above labels as values

    • RIGHT_CC, LEFT_CC, RIGHT_MLO, LEFT_MLO (Dataset):

      • 3 dimensional pixel array of the mammogram, (sheet #, x, y).

    • RIGHT_CC_LABELS, LEFT_CC_LABELS (Dataset):

      • Array of integer values that describe:

        • Whether the image can be evaluated

        • Whether part of the image is cut-off

        • Whether the line from the nipple to the chest is perpendicular to the edge of the image

        • Whether there is good visualisation of deep tissue

        • Whether the line from the nipple to the chest on the CC is less than or equal to 1 cm from the line from the nipple to the pectoral on the MLO

        • Whether the nipple is seen from the side

        • Whether there are any artefacts or superposition

        • Whether the skin of the breast is folded

      • Attributes dictionary uses the dataset's array index as keys and the above labels as values

    • RIGHT_MLO_LABELS, LEFT_MLO_LABELS (Dataset):

      • Array of integer values that describe:

        • Whether the image can be evaluated

        • Whether part of the image is cut-off

        • Whether there is good visualisation of deep tissue

        • Whether there is adequate pectoral muscle in the image

        • Whether there is a maximal view of the pectoral muscle

        • Whether the nipple is seen from the side

        • Whether the infra-mammary angle is well open and shown

        • Whether there are any artefacts or superposition

        • Whether the skin of the breast is folded

      • Attributes dictionary uses the dataset's array index as keys and the above labels as values