Quick Start

Start using NeCT to reconstruct CT images or 4DCT videos. For more advanced usage, look at the demos, the geometry configuration, and how to set up the configuration file.

Static CT4DCT

A static reconstruction using NeCT with medium quality, creating a CT image that is saved to a file.

Example

import numpy as np
import nect

geometry = nect.Geometry(
    DSD=1500.0,  # Distance Source Detector
    DSO=1000.0,  # Distance Source Origin
    nDetector=[256, 512],  # Number of detector pixels [rows, columns]/[height, width]
    dDetector=[1.75, 1.75],  # Size of detector pixels [row, columns]/[height, width]
    nVoxel=[256, 512, 256],  # Number of voxels [height, width, depth]/[z, y, x]
    dVoxel=[1.0, 1.0, 1.0],  # Size of voxels [height, width, depth]/[z, y, x]
    angles=np.linspace(0, 360, 49, endpoint=False),  # Projection angles
    mode="cone",  # Geometry mode (cone or parallel)
    angles_type="degrees",  # Angle units (degrees or radians)
)

volume = nect.reconstruct(
    geometry=geometry,
    projections=<projections-path>, # a single npy-file of shape [nProjections, height, width] 
                                    # or a directory of images
    quality="medium",
)
np.save("volume.npy", volume)

4DCT reconstruction with NeCT is done with higher quality. Afterwards, a video of the middle slice of the object is created through the whole time-series. The video is saved in the same folder as the reconstruction with config and the model weights.

Example

import numpy as np
import nect

geometry = nect.Geometry(
    DSD=1500.0,  # Distance Source Detector
    DSO=1000.0,  # Distance Source Origin
    nDetector=[256, 512],  # Number of detector pixels [rows, columns]/[height, width]
    dDetector=[1.75, 1.75],  # Size of detector pixels [row, columns]/[height, width]
    nVoxel=[256, 512, 256],  # Number of voxels [height, width, depth]/[z, y, x]
    dVoxel=[1.0, 1.0, 1.0],  # Size of voxels [height, width, depth]/[z, y, x]
    angles=np.linspace(0, 360, 49, endpoint=False),  # Projection angles
    mode="cone",  # Geometry mode (cone or parallel)
    angles_type="degrees",  # Angle units (degrees or radians)
)
projections = np.load(<projections-path>)

reconstruction_path = nect.reconstruct(
    geometry=geometry,
    projections=projections,
    quality="highest",
    mode="dynamic"
)
nect.export_video(reconstruction_path, add_scale_bar=True, acquisition_time_minutes=60)