Reconstructing Cone-Beam Projections

Dragonfly's CT Reconstruction module includes the bench-marked Feldkamp-David-Kress (FDK) algorithm for reconstructing 2D cone-beam projection data into 3D volumes, as well as iterative methods and pre-processing to improve image quality. In addition, geometry readers are provided for commercial CT scanners from Nikon, SkyScan, KA Imaging, North Star Imaging, YXLON, GE, and other leading manufacturers.

Choose Utilities > CT Reconstruction on the menu bar to open the CT Reconstruction dialog. Then choose RTK (Cone Beam) to access the reconstruction algorithms and pre-processing options for reconstructing cone-beam projections, as shown below.

CT Reconstruction dialog for cone-beam projections

Settings and Parameters for Cone-Beam Reconstructions

The following settings and parameters are available in Dragonfly's CT Reconstruction module for reconstructing cone-beam projection data.

Reconstruction Engine

The options in the Reconstruction engine box let you choose the reconstruction and acquisition parameters.

Reconstruction Parameters

The options in the Reconstruction parameters box, shown below, let you choose a reconstruction algorithm and iteration.

Reconstruction parameters

Algorithm… The bench-marked Feldkamp-David-Kress (FDK) algorithm is a filtered back-projection algorithm that applies a filter in the frequency domain before back-projecting the projection images.

Compute on GPU (using CUDA)… Lets you accelerate time-consuming reconstructions by computing on GPU. You should note that computing on GPU is required when using iterative methods.

Use iterative method… Lets you choose a forward projection method — CUDA Ray Cast or Joseph — for iteration and the number of iterations.

Advanced Parameters

The options in the Advanced Reconstruction Parameters dialog, shown below, let you choose filtering and the filter settings, as well as the settings for the forward projection method when using iteration.

Click the Advanced Parameters button to open the Advanced Reconstruction Parameters dialog, shown below.

Advanced Reconstruction Parameters

Advanced Reconstruction Parameters dialog

Advanced parameters for cone beam reconstructions
	Description
Filtering	Lets you choose a filter and the filter settings for the reconstruction. The quality of reconstructed images can be limited by several factors that result in poor spatial resolution, low contrast, and high noise levels. However, filtering can compensate for loss of detail in an image while reducing high-frequency noise. The filter chosen for a given image reconstruction task is mainly a compromise between the reduction of noise and the preservation of detail or contrast of the reproduced image. You should note that the Ramlak and Shepp-Logan filters are high pass filters that retain edges, while the Cosine, Hamming, and Hann filters are band pass filters. They are usually used to smooth images and remove extra edges from the image. The following filters, which are applied to data in the frequency domain, are available in the Filter drop-down menu: None… Filtering will not be applied. Cosine… The Cosine filter is obtained by multiplying the Ramlak filter by a cosine function. The advantage of the Cosine and Hann filters is that they reduce image noise. A disadvantage is that they do not preserve edges in the image. Hamming… The low pass Hamming filter provides a high degree of smoothing and has only a single parameter to describe its shape — the cut-off frequency. The only difference with the Hann filter is on the amplitude at the cut-off frequency. Hann… The Hann ('Hanning') filter is a relatively simple low-pass filter that is very effective in reducing image noise. However, it does not preserve edges well. Ramlak… The Ramlak filter is a high−pass filter that blocks out low frequencies, which can cause blurring. If there are areas in the image where the signal changes rapidly a high−pass filter sharpens the edges and gives better contrast. A disadvantage of Ramlak and other high−pass filters is that they let pass high frequencies and as a consequence also noise. Shepp-Logan… The Shepp-Logan filter belongs to the family of low pass filters and produces the least smoothing and has the highest resolution. Note The Shepp-Logan filter is recommended for the most common cases as it reduces noise by at least 10% while the resolution is almost 96% of a 'Ramlak' filtered image. You must set the following when filtering is applied: Frequency cut… Lets you set cut-off frequency of the selected filter. A value of '0' will disable the filter. Note The value of the cut-off frequency determines how the filter will affect both image noise and resolution. A high cut-off frequency will improve the spatial resolution and detail, but the image will remain noisy. A low cut-off frequency will increase smoothing, but will degrade image contrast in the final reconstruction. Truncation padding… This parameter sets the amount of padding — '0' means no padding, '0.5' means padding with 50% more data, while the max value of '1' results in 100% padding.
Forward Projection	Lets you set the values of the convergence factor for the selected forward projection method for iteration, as well as to choose to enforce positivity and disable displaced detector filter.

Acquisition Parameters

The options in the Acquisition parameters box, shown below, let you import the acquisition parameters from a text file or to manually enter the geometry.

Acquisition parameters

Acquisition parameters for cone-beam projections
	Description
CT manufacturer	Lets you choose the device manufacturer of the CT system used to acquire the projection dataset.
Import from file	Lets you choose the metadata file that contains the acquisition parameters. The list of files is automatically filtered based on the device manufacturer as follows: .xtekct… For datasets acquired with Nikon scanners. .json… For datasets acquired with KA Imaging scanners. .rtf… For dataset acquired with North Star Imaging scanners. Click the Import* button to open the Select a Geometry dialog or CT Acquisition Technique Sheet dialog. Click the Reset button to reset all of the original imported parameters. Note If the metadata file is not available, you can enter the acquisition parameters manually, as described below.
Source to detector	Indicates the selected source to detector distance.
Source offset	Indicates the selected source offsets.
Source to isocenter	Indicates the selected source to isocenter distance.
Detector offset	Indicates the selected detector offsets.
Min angle	Indicates the minimum angle.
Detector angle	Indicates the detector angle ranges.
Angle step	Indicates the angle step.
Clockwise steps	Lets you choose between clockwise and counter-clockwise steps.

Pre-Processing

The Pre-processing options let you choose to apply flat-field corrections, median filtering, and other corrections to improve image quality.

Pre-processing options

Pre-processing options for cone-beam projections
	Description
Flat Field Correction	If selected, lets you normalize projection data using the flat and dark field projections. In X-ray imaging, acquired projection images generally suffer from fixed-pattern noise, which is one of the limiting factors of image quality. In conventional flat field correction, projection images without the sample are acquired with and without the X-ray beam turned on, which are referred to as flat fields and dark fields respectively. Flat field… Lets you select the flat-field image taken with the X-ray beam turned on, but without the sample. Fixed pattern noise is removed with the assumption that the detector response did not change over the CT scan time. Dark reference… Lets you select the dark-field image taken before the X-ray beam was turned on.
Median	Lets you choose to apply median filtering, at a selected kernel size and shape, to help smooth noisy data.

How to Reconstruct a Cone-Beam Projection

Refer to the following instructions for information about reconstructing cone-beam projections acquired from Nikon, SkyScan, YXLON, and other CT scanner manufacturers.

See Reconstructing Parallel-Beam Projections for information about reconstructing parallel beam projections.

Load your cone-beam projection dataset (see Importing Image Files).
Choose Workflows > CT Reconstruction on the menu bar.
The CT Reconstruction dialog appears.
Select your projection dataset in the Projections dataset drop-down menu.
Select RTK (Cone Beam) as the reconstruction package in the drop-down menu.
The reconstruction engine options, pre-processing options, and output dimensions appear in the dialog.
Do one of the following:
- Choose the manufacturer of the CT scanning system used to acquire the data in the CT scanner manufacturer drop-down menu.
- Click the Import button.
  If you selected Nikon or KA Imaging as the manufacturer, navigate to and open the file that contains the geometry information for the loaded dataset in the Select a Geometry File dialog.
  If you selected North Star Imaging as the manufacturer, navigate to and open the rich-text formatted file (*.rtf extension) that contains the geometry information for the loaded dataset in the CT Acquisition Technique Sheet dialog.
- Enter the geometry information manually.
Choose the reconstruction parameters (see Reconstruction Parameters).
Click the Advanced Parameters button and then choose the advanced reconstruction parameters, optional (see Advanced Parameters).
Choose the required pre-processing options (see Pre-Processing).
Review the output dimensions, optional (see Output Dimensions).
Click the Compute Preview button, recommended, and then review of results of the one-slice reconstruction.
- If the result is acceptable, continue to the next step.
- If the result is unacceptable, modify the reconstruction parameters and/or pre-processing options and then generate another preview.
  Note You can load the settings of the best preview by clicking the Import Inputs from Preview button and then selecting the best preview in the Import from Preview dialog.
Do one of the following:
- Click the Reconstruct and Load button to compute and load the reconstructed dataset.
- Click the Reconstruct and Save button to compute and save the reconstructed dataset. You will then need to choose a location for the reconstructed dataset in the Export Reconstruction to File dialog.