Skip to content

randomize_contours

Generate randomize contours given and image, label after resampling using MIRP processing library

First, we interpolate the inputs to a isotropic spacing, then get the supervoxel pertubations using the MIRP methods, as defined in:

Zwanenburg, A., Leger, S., Agolli, L. et al. Assessing robustness of radiomic features by image perturbation. Sci Rep 9, 614 (2019). https://doi.org/10.1038/s41598-018-36938-4

Parameters:

Name Type Description Default
input_itk_volume str

path to itk compatible image volume (.mhd, .nrrd, .nii, etc.)

 required
input_itk_labels str

path to itk compatible label volume (.mha)

 required
output_dir str

output/working directory

 required
resample_pixel_spacing float

voxel size in mm

 required
resample_smoothing_beta float

smoothing beta of gaussian filter

 required

Returns:

Name Type Description
dict

metadata about function call
Source code in src/luna/radiology/cli/randomize_contours.py 
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
def randomize_contours(
    input_itk_volume: str,
    input_itk_labels: str,
    resample_pixel_spacing: float,
    resample_smoothing_beta: float,
    output_dir: str,
):
    """Generate randomize contours given and image, label after resampling using MIRP processing library

    First, we interpolate the inputs to a isotropic spacing, then get the supervoxel pertubations using the MIRP methods,
    as defined in:

    Zwanenburg, A., Leger, S., Agolli, L. et al. Assessing robustness of radiomic features by image perturbation. Sci Rep 9, 614 (2019). https://doi.org/10.1038/s41598-018-36938-4

    Args:
        input_itk_volume (str): path to itk compatible image volume (.mhd, .nrrd, .nii, etc.)
        input_itk_labels (str): path to itk compatible label volume (.mha)
        output_dir (str): output/working directory
        resample_pixel_spacing (float): voxel size in mm
        resample_smoothing_beta (float): smoothing beta of gaussian filter

    Returns:
        dict: metadata about function call
    """
    logger.info("Hello, processing %s, %s", input_itk_volume, input_itk_labels)
    settings = Settings()

    print(settings)

    resample_pixel_spacing = np.full((3), resample_pixel_spacing)

    settings.img_interpolate.new_spacing = resample_pixel_spacing
    settings.roi_interpolate.new_spacing = resample_pixel_spacing
    settings.img_interpolate.smoothing_beta = resample_smoothing_beta

    # Read
    image_class_object = read_itk_image(input_itk_volume, "CT")
    roi_class_object_list = read_itk_segmentation(input_itk_labels)

    # Crop for faster interpolation
    image_class_object, roi_class_object_list = crop_image(
        img_obj=image_class_object,
        roi_list=roi_class_object_list,
        boundary=50.0,
        z_only=True,
    )

    # Interpolation
    image_class_object = interpolate_image(
        img_obj=image_class_object, settings=settings
    )
    roi_class_object_list = interpolate_roi(
        img_obj=image_class_object, roi_list=roi_class_object_list, settings=settings
    )

    # Export
    image_file = image_class_object.export(file_path=f"{output_dir}/main_image")

    # ROI processing
    roi_class_object = combine_all_rois(
        roi_list=roi_class_object_list, settings=settings
    )
    label_file = roi_class_object.export(
        img_obj=image_class_object, file_path=f"{output_dir}/main_label"
    )

    roi_class_object_list, svx_class_object_list = randomise_roi_contours(
        img_obj=image_class_object, roi_list=roi_class_object_list, settings=settings
    )

    roi_supervoxels = combine_all_rois(
        roi_list=svx_class_object_list, settings=settings
    )
    voxels_file = roi_supervoxels.export(
        img_obj=image_class_object, file_path=f"{output_dir}/supervoxels"
    )

    for roi in combine_pertubation_rois(
        roi_list=roi_class_object_list, settings=settings
    ):
        if "COMBINED" in roi.name:
            roi.export(
                img_obj=image_class_object, file_path=f"{output_dir}/pertubations"
            )

    print(image_file, label_file)

    # Construct return dicts
    properties = {
        "itk_volume": image_file,
        "itk_labels": label_file,
        "mirp_pertubations": f"{output_dir}/pertubations",
        "mirp_supervoxels": voxels_file,
    }

    return properties