NetForum uses cookies to ensure that we give you the best experience on our website. If you continue to use the site, we'll assume that you are happy to receive these cookies on the NetForum website. Read about our cookies.
NetForum Community
Learn. Share. Optimize.
Log in | Sign up now | Submit content | Contact
Go to similar content

A calibration for non-linear artifact reduction: EVOEye

White Paper
Philips CT Clinical Science Philips Healthcare • USA

Haining Sun, Philips and PN MS, Shenyang, China

Abstract
EVOEye provides a method of improving CT image quality through a novel systematic calibration algorithm. This algorithm is based on the traditional water-equivalent X-ray polychromatic correction method and works by assuming that X-ray related components are perfect and that no factors impact the X-ray. It places all uncertain or unknown factors into one “black box.” A calibration curve is produced by taking into consideration X-ray polychromatic factors, differences in filter materials, detector insensibility, mechanical deflection, and uniformity of detectors. This curve can quickly calibrate raw CT data to reduce non-linear artifacts, such as shading artifacts, cupping artifacts, and ring artifacts in soft tissue. Compared with previous calibration methods, the black box assumption takes into consideration mechanical misalignment and electronic noise. It can reduce assembly complexity, as well as remove ring artifacts and other non-uniformity artifacts to reduce the difficulty and complexity of post-processing ring removal. This calibration method operates in pre-processing mode and does not increase normal reconstruction time. Results show that EVOEye helped improve low-contrast image quality from 4 mm at 3% HU to 2 mm at 3% HU.

Introduction
CT image quality, especially for low-contrast exams, is essential for diagnostic confidence. Many factors affect CT image quality, such as cupping artifacts, ring artifacts, artifacts from non-uniformity band rings caused by non-ideal X-ray source polychromatic factors, non-ideal detector insensitivity, and varying uniformity of different detector modules.

Correction of X-ray polychromatic artifacts has been an area of active research since the beginning of CT. Device-based corrections and algorithm-based corrections are both under study. Devices such as aluminum filters, brass filters, and bow-tie filters were designed to ameliorate “soft” X-rays. Meanwhile, many artifact-reducing algorithms for beam hardening and rings have been presented by academic centers and research companies.1-8 Calibration methods are usually divided into two categories. One focuses on preprocessing techniques, e.g. water-equivalent calibration for soft tissue1 and the other on post-processing technique for hard tissue, e.g. bone calibration.2-5 There are also other post-processing methods that focus on ring artifacts.

The traditional water-equivalent correction method assumes the X-ray spectrum of the tube to be available and accurate. However, the real spectrum is not easy to obtain due to different X-ray filters, e.g. differentials in X-ray tube filters and bow-tie filters. Theoretical analysis helps guide determination of the spectrum.6,7 However, changes in tube materials, misalignment in the mechanical assembly, and other electrical noise cause the spectrum to be off, especially after a tube change.

Meanwhile, the assumed general X-ray spectrum does not take into account the difference among X-ray channels, and so ring artifacts, which are caused by detector insensibility and non-uniformity, are not able to be corrected.

Another method, similar to that of Yan,7,8 explored a postprocessing method to accurately calculate the spectrum, but was found to be time-consuming and dealt only with the image instead of the directly with the raw data.

In this paper, we present a method to directly calibrate raw data based on the traditional water-equivalent X-ray polychromatic correction method. It assumes X-ray related components are perfect and that no factors impact the X-ray. All uncertain or unknown factors are considered to be in a black box that produces a calibration curve to account for X-ray polychromatic factors, differences in filter materials, detector insensibility, mechanical deflection, and detector uniformity. In identifying these imperfections, EVOEye is able to calculate a more precise spectrum calibration curve. Image results from using this spectrum curve to calibrate raw data showed that lowcontrast image quality improved from 4 mm @ 3% HU to 2 mm @ 3% HU without increasing the time for image calibration and reconstruction.

Methodology
In third-generation CT systems, the X-ray tube projects a fan-shaped X-ray beam that passes through the object and is received by the detectors on the other side of the tube. The detectors measure this raw transmission data from every angle of view. Using these raw data, the CT system can reconstruct images.

CT reconstruction theory assumes that X-rays emitted by the tube are monochromatic, but in fact the spectrum is polychromatic. Therefore, many dark beam-hardening artifacts appear in the images. The traditional waterequivalent correction is described in Fig. 1.
Fig. 1 This equation indicates the relationship between raw data and material attenuation.
Fig. 1
This equation indicates the relationship between raw data and material attenuation.

Because the human body is composed primarily of water and soft tissue, the measured raw data (Rawmeasured) can be remapped to ideal raw data, that is, without X-ray polychromatic artifacts, by using a 4th order polynomial function curve (Fig. 2).
Fig. 2 The curve parameters can be obtained with minimum least-square fit using the known water attenuation and spectrum function.
Fig. 2
The curve parameters can be obtained with minimum least-square fit using the known water attenuation and spectrum function.

The curve parameters can be obtained with minimum least-square fit using the known water attenuation and spectrum function (SF).

SF plays an important role in calculating curve parameters. Tube manufacturers provide initial information on SF, but for CT systems in clinical use, filter materials such as bow-tie filters, aluminum, or brass filters are often used to filter some “bad” X-rays. All filter materials change the initial SF. Therefore, the SF needs to be adjusted according to realworld situations. Separately considering all filter materials during the process of adjusting SF renders the problem too complicated to solve and so it remains indefinitely. Hence, we regard all filter materials and some unknown factors affecting the X-ray as a black box as shown in Fig. 3, and we directly calculate the materials spectrum.
Fig. 3 The ideal map for the black box description accounts for other filters such as tube filters, bow-tie filters, and other materials between the X-ray and the detectors.
Fig. 3
The ideal map for the black box description accounts for other filters such as tube filters, bow-tie filters, and other materials between the X-ray and the detectors.

Note that for a channel, the SF and the length of the filter material are attributes of the CT system and so should be considered as independent of the phantom put on the field of scan (FOS). Different phantoms should result in the same SF and the same filter material. Given the initial SF, measured phantom parameters and experimental raw data, (assuming that filter material attenuation is known), we can calculate the length of filter materials using the equation in Fig. 4. If the SF is accurate, different phantom experiments will yield the same filter material length (LF).
Fig. 4
Fig. 4

Raw(n,w)
denotes raw data for the nth channel when phantom w is scanned. Sinit (n,e) denotes the initial spectrum function of the nth channel at e energy level. μP (e) and μF (e) are the attenuation of the phantom and filter material at e energy level respectively. For simplicity we assumed only one filter material such as aluminum is considered in the black box. LP (n,w) and LF (n,w) denote the length of phantom and filter material on the nth channel when phantom w is scanned. LP (n,w) can be calculated through the geometry parameters.

All experiments were carried out using the Philips MX16EVO CT scanner. The experimental phantom size is 7 inches. Water phantom No.1 was put into the FOS and static scanned with following scan settings: 120 kVp, 175 mAs, 1 second, slice width: 7.5 mm. The initial SF was chosen as the dashed line shown in Fig. 5. We calculated filter material lengths using the equation in Fig. 4 with the raw data measured from two different phantoms. The different channel spectrum parameters were shown in Fig. 6. Using the adjusted spectrum, the 4th order polynomials in the equation in Fig. 2 were calibrated. Results and more discussion will be shown in the next section.
Fig. 5 Two spectra from channel 280. The dashed line refers to the initial spectrum while the continuous line refers to the adjusted spectrum. Note that different channels may have different initial and final spectrums.Fig. 6 Calculation of difference in spectrum parameters.
Fig. 5
Fig. 6
Two spectra from channel 280. The dashed line refers to the initial spectrum while the continuous line refers to the adjusted spectrum. Note that different channels may have different initial and final spectrums.
Calculation of difference in spectrum parameters.

Results and discussion
The following paragraphs depict the results of the preceding experiments.

Another water phantom was scanned at setting: 120 kVp, 175 mAs, slice width: 7.5 mm. As shown in Fig. 7, the right image Fig. 7 (b) was reconstructed by traditional water-equivalent method using the new spectrum, and the left one in Fig. 7 (a) was done without the spectrum adjusting. Clearly the cupping artifacts and ring artifacts in image (a) were significantly reduced as shown in image (b).

Catphan phantom low contrast was also scanned with the same scan settings: 120 kVp, 250 mAs, slice width: 9 mm. Compared with Fig. 8 (b), the low contrast improved from 4 mm @ 0.3% to 2 mm @ 0.3%.
Fig. 7 Water Phantom CT images (120 kVp, 175 mAs, 7 mm, FOV: 15 cm) with and without EVOEye algorithm displayed at a level and width of 0 HU and 50 HU respectively. The cupping artifact in (a) was significantly reduced as shown in (b).Fig. 8 Catphan Phantom images low contrast part (120 kVp, 250 mAs, 9 m, FOV: 25 cm) with and without EVOEye algorithm. The results show that (a) 4 mm @ 0.3% improved in (b) 2 mm @ 0.3%.
Fig. 7
Fig. 8
Water Phantom CT images (120 kVp, 175 mAs, 7 mm, FOV: 15 cm) with and without EVOEye algorithm displayed at a level and width of 0 HU and 50 HU respectively. The cupping artifact in (a) was significantly reduced as shown in (b).
Catphan Phantom images low contrast part (120 kVp, 250 mAs, 9 m, FOV: 25 cm) with and without EVOEye algorithm. The results show that (a) 4 mm @ 0.3% improved in (b) 2 mm @ 0.3%.

Conclusion
We presented a method to improve low-contrast image quality. This EVOEye method synthesized all spectrum affecting materials as a black box and calculated the spectrum function based on an inverse monotonic relationship between the difference of filter material length and the adjusted spectrum parameter, using a static small water phantom and the original spectrum adjusted for a given channel. New spectrum functions, which replaced the old functions in the traditional water-equivalent approach, were used to calibrate the raw data. Compared to the traditional method, the new method obtained better images. Meanwhile, X-ray polychromatic artifacts and nonuniform detector ring artifacts were significantly reduced.

Acknowledgment
The authors would like to acknowledge the support of colleagues in the CT R&D Department of PN MS, Shenyang, China.

References
  1. G.T.Herman, “Correction for beam hardening in Computed Tomography.” Phys. Med. Biol., vol. 24, no. 1, pp. 81-106, 1979. 
  2. P.K.Kijewski and B.E.Bjarngard, “Correction for beam hardening in computed tomography.” Med. Phys., vol. 5, no. 3, pp. 209-214, 1978. 
  3. C. H. Yan and R.T. Whalen, “Modeling of polychromatic attenuation using computed tomography reconstructed images.” Med.Phys., vol. 26, no. 4, pp. 631-642, 1999. 
  4. P.M.Joseph and C.Ruth, “A method for simultaneous correction of spectrum hardening artifacts in CT images containing both bone and iodine.” Med.Phys., vol. 24, no. 10., pp. 1629-1634, 1997. 
  5. R.Raupach, “Method for correcting for beam hardening in a CT image.” U.S. Patent 6,600,801, Jul. 29, 2003. 
  6. C.Ruth and P.M.Joseph, “Estimation of a photon energy spectrum for a computed tomography scanner.” Med.Phys., vol. 24, no. 5, pp. 695-702, 1997. 
  7. C.H.Yan and R.T.Whalen, “Reconstruction algorithm for polychromatic CT imaging: Application to beam hardening correction.” IEEE. Trans. Medical Imaging,.vol. 19, no. 1, pp. 1-11, 2000. 
  8. C.H.Yan, R.T.Whalen, Sandy Napel “Method for beam hardening correction in quantitative computed x-ray tomography.” U.S. Patent 6,324,240, Nov. 27, 2001.


This content has been made possible by NetForum Community.
Share this on: Share your link in twitter Share your link in facebook Share your link on LinkedIn Print Rate this article: Log in to vote

 
Rating:
Votes:
0
Views:
11600
Added:
Jan 5, 2012

Rate this:
Log in to vote
 

White Paper
MX 16-slice / MX16 EVO
artifacts, Body, image quality, Interventional, low contrast volume, Musculoskeletal, Neuro, Oncology, Pediatric
 

Clinical News
Best Practices
Case Studies
Publications and Abstracts
White Papers
Web seminars and Presentations
ExamCards
Protocols
Application Tips and FAQ
Training
Try an Application
Business News
Case Studies
White Papers
Web Seminars and Presentations
Utilization Services
Contributing Professionals
Contributing Institutions
Become a Contributor