Functional Transformations¶

In [1]:

from pathlib import Path

from imgtools.io import read_dicom_auto
from imgtools.ops.functional import (
    bounding_box,
    centroid,
    clip_intensity,
    crop,
    crop_to_mask_bounding_box,
    image_statistics,
    min_max_scale,
    resample,
    resize,
    rotate,
    standard_scale,
    window_intensity,
    zoom,
)

from pathlib import Path from imgtools.io import read_dicom_auto from imgtools.ops.functional import ( bounding_box, centroid, clip_intensity, crop, crop_to_mask_bounding_box, image_statistics, min_max_scale, resample, resize, rotate, standard_scale, window_intensity, zoom, )

In [2]:

image_path = Path.cwd().parent.parent.parent / 'data/Head-Neck-PET-CT/HN-CHUS-052/08-27-1885-CA ORL FDG TEP POS TX-94629/3.000000-Merged-06362'

# mask path used later in the notebook 
mask_path = Path.cwd().parent.parent.parent / 'data/Head-Neck-PET-CT/HN-CHUS-052/08-27-1885-CA ORL FDG TEP POS TX-94629/1.000000-RTstructCTsim-PETPET-CT-87625'

example_scan = read_dicom_auto(image_path.as_posix())
example_image = example_scan.image

image_path = Path.cwd().parent.parent.parent / 'data/Head-Neck-PET-CT/HN-CHUS-052/08-27-1885-CA ORL FDG TEP POS TX-94629/3.000000-Merged-06362' # mask path used later in the notebook mask_path = Path.cwd().parent.parent.parent / 'data/Head-Neck-PET-CT/HN-CHUS-052/08-27-1885-CA ORL FDG TEP POS TX-94629/1.000000-RTstructCTsim-PETPET-CT-87625' example_scan = read_dicom_auto(image_path.as_posix()) example_image = example_scan.image

In [3]:

from imgtools.datasets import example_data_paths
from SimpleITK import ReadImage
paths = example_data_paths()

example_image = ReadImage(paths['duck'])
example_mask = ReadImage(paths['mask'])

from imgtools.datasets import example_data_paths from SimpleITK import ReadImage paths = example_data_paths() example_image = ReadImage(paths['duck']) example_mask = ReadImage(paths['mask'])

Helpers¶

A set of helper functions to aid in the demo.

In [4]:

import math
import numpy as np
import matplotlib.pyplot as plt
import SimpleITK as sitk


def visualize_3d_slices_adaptive(image: sitk.Image, axis: int = 0, step: int = 1, cmap='gray') -> None:
    """
    Visualizes every n-th slice of a 3D SimpleITK image along a specified axis in an adaptive grid layout.

    Args:
        image (sitk.Image): The 3D SimpleITK image to visualize.
        axis (int): The axis along which to slice (0: x, 1: y, 2: z).
        step (int): The step size for selecting slices. Defaults to 1 (every slice).
    """
    array = sitk.GetArrayFromImage(image)
    num_slices = array.shape[axis]
    selected_slices = range(0, num_slices, step)
    num_selected_slices = len(selected_slices)

    vmax = array.max()
    vmin = array.min()

    # Calculate adaptive grid size
    cols = math.ceil(math.sqrt(num_selected_slices))
    rows = math.ceil(num_selected_slices / cols)

    fig, axes = plt.subplots(rows, cols, figsize=(cols * 3, rows * 3))
    axes = axes.ravel()  # Flatten the axes for easy iteration

    for i, slice_index in enumerate(selected_slices):
        if axis == 0:
            slice_ = array[slice_index, :, :]
        elif axis == 1:
            slice_ = array[:, slice_index, :]
        elif axis == 2:
            slice_ = array[:, :, slice_index]

        slice_rotated = np.rot90(slice_, k = 2)

        axes[i].imshow(slice_rotated, cmap, vmax=vmax, vmin=vmin)
        axes[i].set_title(f'Slice {slice_index}')
        axes[i].axis('off')

    # Turn off unused axes
    for ax in axes[num_selected_slices:]:
        ax.axis('off')

    plt.tight_layout()
    plt.show()

visualize_3d_slices_adaptive(example_image, axis=2, step=5)

import math import numpy as np import matplotlib.pyplot as plt import SimpleITK as sitk def visualize_3d_slices_adaptive(image: sitk.Image, axis: int = 0, step: int = 1, cmap='gray') -> None: """ Visualizes every n-th slice of a 3D SimpleITK image along a specified axis in an adaptive grid layout. Args: image (sitk.Image): The 3D SimpleITK image to visualize. axis (int): The axis along which to slice (0: x, 1: y, 2: z). step (int): The step size for selecting slices. Defaults to 1 (every slice). """ array = sitk.GetArrayFromImage(image) num_slices = array.shape[axis] selected_slices = range(0, num_slices, step) num_selected_slices = len(selected_slices) vmax = array.max() vmin = array.min() # Calculate adaptive grid size cols = math.ceil(math.sqrt(num_selected_slices)) rows = math.ceil(num_selected_slices / cols) fig, axes = plt.subplots(rows, cols, figsize=(cols * 3, rows * 3)) axes = axes.ravel() # Flatten the axes for easy iteration for i, slice_index in enumerate(selected_slices): if axis == 0: slice_ = array[slice_index, :, :] elif axis == 1: slice_ = array[:, slice_index, :] elif axis == 2: slice_ = array[:, :, slice_index] slice_rotated = np.rot90(slice_, k = 2) axes[i].imshow(slice_rotated, cmap, vmax=vmax, vmin=vmin) axes[i].set_title(f'Slice {slice_index}') axes[i].axis('off') # Turn off unused axes for ax in axes[num_selected_slices:]: ax.axis('off') plt.tight_layout() plt.show() visualize_3d_slices_adaptive(example_image, axis=2, step=5)

Image Statistics¶

In [5]:

image_statistics(example_image)

image_statistics(example_image)

Out[5]:

ImageStatistics(minimum=0.0, maximum=255.0, sum=24488459.0, mean=24.488459, variance=3802.7554795607984, standard_deviation=61.666485870047744)

Image Transformations¶

Resample¶

In [6]:

print('Original Spacing:', example_image.GetSpacing())
print('Original Size:', example_image.GetSize())

resampled_image = resample(example_image, [1, 1, 1], output_size=[128, 128, 128])
print('Resampled Spacing:', resampled_image.GetSpacing())
print('Resampled Size:', resampled_image.GetSize())

visualize_3d_slices_adaptive(resampled_image, axis=2, step=5)

print('Original Spacing:', example_image.GetSpacing()) print('Original Size:', example_image.GetSize()) resampled_image = resample(example_image, [1, 1, 1], output_size=[128, 128, 128]) print('Resampled Spacing:', resampled_image.GetSpacing()) print('Resampled Size:', resampled_image.GetSize()) visualize_3d_slices_adaptive(resampled_image, axis=2, step=5)

Original Spacing: (1.0, 1.0, 1.0)
Original Size: (100, 100, 100)
Resampled Spacing: (1.0, 1.0, 1.0)
Resampled Size: (128, 128, 128)

Resize¶

In [7]:

print('Original Size:', example_image.GetSize())
resized_image = resize(example_image, [256, 256, 0])
print('Resized Size:', resized_image.GetSize())

visualize_3d_slices_adaptive(resampled_image, axis=2, step=5)

print('Original Size:', example_image.GetSize()) resized_image = resize(example_image, [256, 256, 0]) print('Resized Size:', resized_image.GetSize()) visualize_3d_slices_adaptive(resampled_image, axis=2, step=5)

Original Size: (100, 100, 100)
Resized Size: (256, 256, 100)

Zoom¶

In [8]:

zoomed_image = zoom(example_image, 2.0)

visualize_3d_slices_adaptive(zoomed_image, axis=2, step=5)

zoomed_image = zoom(example_image, 2.0) visualize_3d_slices_adaptive(zoomed_image, axis=2, step=5)

Rotate¶

In [9]:

size = example_image.GetSize()
center_voxel = [size[i] // 2 for i in range(len(size))]

rotated_image = rotate(example_image, center_voxel, [45, 45, 45])

visualize_3d_slices_adaptive(rotated_image, axis=2, step=5)

size = example_image.GetSize() center_voxel = [size[i] // 2 for i in range(len(size))] rotated_image = rotate(example_image, center_voxel, [45, 45, 45]) visualize_3d_slices_adaptive(rotated_image, axis=2, step=5)

Crop¶

In [10]:

cropped_image = crop(example_image, [50, 50, 50], [90, 90, 90])

visualize_3d_slices_adaptive(cropped_image, axis=2, step=5)

cropped_image = crop(example_image, [50, 50, 50], [90, 90, 90]) visualize_3d_slices_adaptive(cropped_image, axis=2, step=5)

Mask Transformations¶

In [16]:

mask = example_mask
visualize_3d_slices_adaptive(mask, axis=2, step=3, cmap=None)

mask = example_mask visualize_3d_slices_adaptive(mask, axis=2, step=3, cmap=None)

Centroid¶

In [12]:

centre_coords = centroid(mask)
print(centre_coords)

centre_coords = centroid(mask) print(centre_coords)

(50, 29, 76)

Bounding Box¶

In [13]:

box_coords = bounding_box(mask)
print(box_coords)

box_coords = bounding_box(mask) print(box_coords)

((45, 21, 67), (10, 17, 17))

Crop to Bounding Box¶

In [14]:

cropped_image, cropped_mask, crop_centre = crop_to_mask_bounding_box(example_image, mask)
print(crop_centre)

visualize_3d_slices_adaptive(cropped_mask, axis=2, step=1)
visualize_3d_slices_adaptive(cropped_image, axis=2, step=1)

cropped_image, cropped_mask, crop_centre = crop_to_mask_bounding_box(example_image, mask) print(crop_centre) visualize_3d_slices_adaptive(cropped_mask, axis=2, step=1) visualize_3d_slices_adaptive(cropped_image, axis=2, step=1)

[np.float64(50.0), np.float64(29.5), np.float64(75.5)]

Intensity Transformations¶

Clip Intensity¶

In [17]:

clipped_image = clip_intensity(example_image, lower=100, upper=255)

visualize_3d_slices_adaptive(clipped_image, axis=2, step=5)

clipped_image = clip_intensity(example_image, lower=100, upper=255) visualize_3d_slices_adaptive(clipped_image, axis=2, step=5)

Window Intensity¶

In [18]:

windowed_image = window_intensity(example_image, 1500, 0)

visualize_3d_slices_adaptive(windowed_image, axis=2, step=5)

windowed_image = window_intensity(example_image, 1500, 0) visualize_3d_slices_adaptive(windowed_image, axis=2, step=5)

Min-max Normalization¶

In [20]:

min_max_image = min_max_scale(example_image, minimum=0, maximum=5)

visualize_3d_slices_adaptive(min_max_image, axis=2, step=5)

min_max_image = min_max_scale(example_image, minimum=0, maximum=5) visualize_3d_slices_adaptive(min_max_image, axis=2, step=5)

Standardization¶

In [21]:

standardized_image = standard_scale(example_image)

visualize_3d_slices_adaptive(standardized_image, axis=2, step=5)

standardized_image = standard_scale(example_image) visualize_3d_slices_adaptive(standardized_image, axis=2, step=5)