CT Image Reconstruction

It explains CT reconstruction features of CT reconstruction Software TomoShop®, such as fast data processing, high quality tomographic image processing, adopting various types of CT system and so on.

1. CT scan system

The left diagram below shows normal type of Cone beam X-ray CT hardware(CBCT). The sample is set on the turning table that rotates. The sample is rotated and the scan is taken from 1 to 360 degrees (sometime less than 180), then the projection image data is saved in the hardware system. In other type of cone beam X-ray CT hardware, X-ray source and FDP(camera) are rotating instead of the sample. The set of projection image data is reconstructed as 3D volume data(slice data) by the CT reconstruction data, and CT user inspect or observe the 3D volume data while NDI is conducting.

The process of image reconstruction

Ordinal CT reconstruction takes the steps which showed the right diagram above. Our CT reconstruction software: TomoShop® also follow this steps for reconstruction, however it also including the special functions that is stated in below.

  • Corresponding to variety of file format of import data.
  • Noise reduction on the projection data.
  • Various types of artifact reduction function
  • Large sized back projections process could be taken with fast processing.
  • There are algorithm for corresponding variety of types of CT scan.

2. High-speed image reconstruction by PC with GPU(CUDA architect)

TomoShop® could reduce those long processing time by using CUDA GPU (GeForce, Quadro, etc.).

To illustrate the high-speed image reconstruction by TomoShop®, the next paragraphs show the example of result from TomoShop®.

Example of high-speed image reconstruction:

  • The purpose: Showing high-speed image reconstruction by using TomoShop®
  • Operation: Do image reconstruction and create 3D image (the pixel 512 x 512 x 512) by FDK method [2] from the projection data of 360 sheets (the pixel 512×512) that are taken by standard circular orbit cone-beam CT apparatus.
    • PC: Notebook type
    • OS: Windows 10Pro (64bit)
    • CPU: Intel® Core™ i7-363QM @2.4GHz
    • Memory: 16.0GB
    • GPU: NVIDIA GeForce GTX 680M
    • Computation time of the back-projection process + Reconstruction Filter calculation + Back projection calculation: 1.5 seconds ☞ 0.5 – 0.75 seconds (by using NVIDIA 2nd generation architecture “Ampere” GPU; RTX3080 or later)

Ultra-fast reconstruction (with multi-GPU with large sized data)

When reconstructing CT image data in considerably large size, it may take too much time with just one GPU (graphic board) not enough power.

TomoShop(from ver3.5) is capable of cone beam CT reconstruction using multi-GPU. It is possible to perform CT reconstruction at super high-speed by using a plurality (up to 16) of GPUs (Graphic boards) of the same model number. (*TomoShop Use NVIDIA’s CUADA GPU)

When CT reconstruction is performed using the FDK[2] method with respect to X-ray projection data obtained with a standard circular orbital cone beam CT apparatus (CCT), the calculation time and implementation environment are as follows.

  • OS: 64-bit Windows 10 Pro
  • CPU and memory: Intel Core i7-6850K 3.60 GHz, 128 GB RAM.
  • GPU: Dual GeForce GTX1080, 8GiB GDDR 5X, CUDA Version 8.0
  • Captured image size: 2000 x 2000 pixels
  • Number of shot images: 720
  • Reconstructed image size: 2000 x 2000 x 2000 voxels
  • Comparison of reconstruction time (Log calculation + Weight calculation + Filter calculation + Back projection calculation)

Single GPU ☞ 52 seconds ☞ 17 – 26 seconds (by using NVIDIA 2nd generation architecture “Ampere” GPU; RTX3080 or later)
Dual GPU ☞ 30 seconds ☞ 10 – 15 seconds (by using NVIDIA 2nd generation architecture “Ampere” GPU; RTX3080 or later)

Note: The multi-GPU function is an optional function. Therefore, extra charge is required. (In the case of the F3 edition series, it costs 600,000 yen (excluding tax) for dual GPUs. If you use more GPUs, it costs 200,000 yen (excluding tax) for each GPU.)

It’s even faster with GPU (RTX3080 or later) of NVIDIA’s 2nd generation Ampere architecture. (2 to 3 times faster!)

☞ Go to TomoShop® (CT reconstruction software) page.

3. High quality tomographic image

The solution for cupping artifacts on the reconstruction image

Beam Hardening Phenomenon which referring to the effect of selective X-ray attenuation and scatter from polychromatic X-ray beams, are one of the typical problem with using CT system.
For example, Cupping artifacts are occurred on the reconstruction image by beam hardening phenomenon, and it prevents obtaining real shape of the target sample and it affects the accuracy of inspection that is the crucial problem with using CT.

However, TomoShop® has methods that could reduce those cupping artifacts.
(see the image below　↓)

  • Cupping artifacts (Beam-hardening) is appeared by using ordinal CT reconstruction methods. (Left)
  • Cupping artifacts (Beam-hardening) is reduced by using TomoShop®’s CT reconstruction method. (Right)

☞ Go to Artifact Reduction page.

☞ Go to TomoShop® (CT reconstruction software) page.

 

Selection of image reconstruction algorithm depending on target objects

Currently, FDK is the most popular and the most favor for algorithm of image reconstruction.

However, selection of image reconstruction algorithm is crucial for obtaining high quality images. Especially, when the target object has complex structures with many materials, the reconstructed result does not have strong contrast among its luminance, FDK algorithm less powerful tool for algorithm or image reconstruction.

If there is a case for inspecting the objects that are not easy for FDK tool, TomoShop® has two new powerful tools, CFDK[1] method and TFDK[1,3] method with filtering for reducing cone-beam artifact. These are more powerful tools than FDK.
(See the images below. ↓)

 

※CFDK[1] method, TFDK[1,3] method are only mounted with F4 Edition.

☞ Go to TomoShop® (CT reconstruction software) page

 

4. Corresponding to various movements of various CT equipment

TomoShop® can be adopted most type of cone-beam CT.

1. Standard/Partial scan, Omni scan, Offset scan

Geometric design of the device depends on the purpose of use and inspection areas, cone-beam CT system, the detector arrangement and the scanning trajectory is changed accordingly.

Each cone beam CT system has different geometric design of devices depends on purpose of use and inspection areas, so that the detector arrangement and the scanning trajectory is changed according to the purposes of usages.
For example, standard cone beam CT system can use FDK[2] method for reconstruction. On the other hand, digital tomosynthesis such as breast tomosynthesis will not be use FDK[2] method due to different placement of the detector.

TomoShop® has Omni back projection algorithm [1] that be able to correspond various trajectories geometric shooting systems such as circle, orbital, polygonal and elliptical orbit. TomoShop® also has half-scan reconstruction function. 

Offset scan data reconstruction is also available.

2. Partial reconstruction

TomoShop ® can partially reconstruct a target object by indicating the areas and partially processing back-projection.

The actual non-destructive inspection is not always necessary to check all the imaging range of the object. It is very often necessary to inspect only one part or few parts of the imaging range.

In TomoShop ®, user can set the parameter of the inspection area of the object that users trying to test. TomoShop® does image reconstruction of specified area with very fast speed.

3. Zoom reconstruction

TomoShop ® creates reconstructions of subjects in larger sized or smaller sized up on users’ request.

In Tomography, the standard voxel size of the object is determined by the distance to the detector, the distance from the X-ray source to the center of rotation, and the distance from X-ray source to the detector.

In comparison, the voxel size given by the user, TomoShop® can reconstruct the target object.

☞ Go to TomoShop® (CT reconstruction software) page

 

4. Oblique cone beam CT reconstruction (Planar CT, PCT)

The detail page ☞ Click here!

☞ Go to TomoShop® (CT reconstruction software) page.

 

5. Helical CT Reconstruction

Helical CT system is popular in Health and industrial divisions. TomoShop® can correspond to helical CT.

☞ Go to TomoShop® Helix Edition Series page.

 

6. Multi-detector CT

TomoShop® can be used for the multi-detector CT. The multi-detector CT can be used for the large-sized samples.

 

7. Multiple X-ray Sources with Multi-detector CT

TomoShop® can be used for the multiple x-ray source CT or the multiple x-ray sources + multi-detector CT. The multiple x-ray source CT or the multiple x-ray source + multi-detector CT can be used for the large-sized samples.

 

8. Horizontal Translations (straight line) Track System

Horizontal Translations (straight line) Track System is more commonly used major system than circular path system for automatic inspection line for electric parts such as printed circuit boards. TomoShop can be adopted to Horizontal Translations (straight line) Track System.

☞ See HT Edition Series

Straight Track Parallel Fan-beam Scanning System

Straight Line Trajectory Cone-beam Scanning System

9. Iterative CT Reconstruction

In recent years, Iterative CT reconstruction method is drawing attention as a new CT reconstruction against Back-projection CT reconstruction method. The advantage of Iterative CT reconstruction method is reconstruction results with high image quality can be obtained even with few projected images(raw data). Particularly, in CT for medical use and animals, exposure to the patient is regarded as a problem, so Iterative CT reconstruction method can keep the exposure rate low by shortening the number of shots and CT scan time has been attracting attention in recent CT technologies. Iterative CT reconstruction method is also drawing attention in other fields, since it is reducing scanning time, which has the advantage of getting results in short time with high quality images.

Advantages:

  • High quality CT images can be obtained with few projection images (raw data).
  • The CT scan time is shorter than the conventional CT scan time.
  • It is possible to realize a low exposure rate and shorten the CT scan time.

TomoShop implements the CT reconstruction function of Iterative CT reconstruction tool. The iterative CT reconstruction tool implemented in TomoShop has been developed with the goal of producing high quality images by using advance information.

TomoShop (from ver 4.0) offers this original high-quality Iterative CT reconstruction tool and GPU(CUDA)to provide higher quality CT images in a shorter time.

Note: Iterative CT reconstruction tool is an optional function of F3 and F4 edition series. (A separate fee (JPY2,000,000) will be charged.)

10. Phase-contrast X-ray Imaging CT(Phase CT) reconstruction

TomoShop can reconstruct CT projection image data from phase-contrast x-ray imaging technology (Phase CT).
☞ Please see “Phase-contrast X-ray Imaging (PCI)” for details on phase-contrast x-ray imaging (PCI) technology and phase CT reconstruction.

11. Supports reconstruction of various imaging data (other than X-ray CT)

TomoShop is software originally developed for reconstructing projected image data taken by X-ray CT. However it can also be used for reconstructing image data taken with other devices to provide high-quality results.

For example, it is used in the reconstruction of the following image data.

  • TEM / STEM (electron microscope imaging data)
  • Neutron CT
  • Proton CT
  • Photograph image (gray scale) 
  • Video camera data (gray scale) 
    Etc.

Therefore, please feel free to contact us if you would like to reconstruct the image data taken by other than these X-ray CTs.

[1] 李　美花、工藤博幸, CUDAによるコーンビームCT画像再構成の高速化とツールキット開発, 映像情報メディカル, Vol. 40, No. 13, pp. 1194-1198, 2008年12月.

[2] L.A. Feldkamp, L.C. Davis and J.W. Kress, Practical Cone-Beam Algorithm, J. Opt. Doc. Am. A 1, pp. 612-619, 1984.

[3] M. Grass, T. Koehler and R. Proksa, 3D Cone-Beam CT Reconstruction for Circular Trajectories, Phys. Med. Biol. Vol. 45, No. 2, pp. 329-347, 2000.

※All images in this page are belong to Midorino Research Co., Ltd