Tutorial

This page steps through running the HiP-CT registration pipeline.

import logging

import matplotlib.pyplot as plt

from hipct_reg.data import get_reg_input
from hipct_reg.helpers import (
    get_central_pixel_index,
    get_central_point,
    get_pixel_transform_params,
    resample_zoom_image,
    show_image,
)
from hipct_reg.registration import run_registration

logging.basicConfig(level=logging.INFO)

Get data

The registration pipeline makes use of simpleITK, both for the actual registration and for inputting the images. Using it for image input allows a smaller volume than the whole image volume to be used to register, while keeping track of its offset from the origin of the original (full) data volume.

hipct-reg takes a RegistrationInput object as input to the registration pipeline. This class stores the two images, the (pixel) indices of a common point in them, and their names.

get_reg_input provides an easy way to get the registration input for two datasets:

zoom_name = "LADAF-2021-64_heart_VOI-01_6.51um_bm18"
overview_name = "LADAF-2021-64_heart_complete-organ_19.89um_bm18"
zoom_point = (2240, 1770, 1546)
full_point = (4040, 5136, 502)

reg_input = get_reg_input(
    zoom_name=zoom_name,
    zoom_point=zoom_point,
    overview_name=overview_name,
    overview_point=full_point,
    zoom_level=0,
    overview_level=0,
)

Run registration

This is all the input we need - we can just pass this to run_registration to run the full registration pipeline:

transform, reg_metric = run_registration(reg_input)

This returns two objects - the registered transform, and the registration metric. The transform maps from the physical coordinates of the zoom image to the physical coordinates of the overview image. For HiP-CT data these are just the voxel coordinates multiplied by the image resolution.

To get the parameters for a pixel-to-pixel mapping from the zoom image to the full-organ image we can use get_pixel_transform_params:

pixel_transform_params = get_pixel_transform_params(reg_input, transform)
print(pixel_transform_params)

Plot results

# Get physical point in centre of full organ image
central_point = get_central_point(reg_input.overview_image)

fig, axs = plt.subplots(nrows=1, ncols=3, constrained_layout=True, figsize=(10, 5))

# Before
central_idx = get_central_pixel_index(reg_input.zoom_image)
show_image(reg_input.zoom_image, axs[0], central_idx[2])
axs[0].set_title("zoom before registration")

# After
zoom_resampled = resample_zoom_image(reg_input, transform)
resampled_idx = zoom_resampled.TransformPhysicalPointToIndex(central_point)
show_image(zoom_resampled, axs[1], resampled_idx[2])
axs[1].set_title("zoom after registration")
axs[1].grid(color="black")

# Full-organ image
central_idx = get_central_pixel_index(reg_input.overview_image)
show_image(reg_input.overview_image, axs[2], central_idx[2])
axs[2].set_title("Full-organ")
axs[2].grid(color="black")


plt.show()