Image reconstruction (CT)

A.Prof Frank Gaillard and Andrew Murphy et al.

Image reconstruction in computed tomography is a rapidly evolving industry, the race to produce an efficient yet accurate image reconstruction method while keeping scan dose to a minimum has defined improvements in CT over the past decade.

The mathematical problem that CT image reconstruction is trying to solve is to compute the attenuation coefficients of different x-ray absorption paths (ray sum) that are obtained as a set of data (projection).

Reconstruction algorithms

There are various algorithms used in CT image reconstruction, the following are some of the more common algorithms utilized in commercially available CT today. 

  • iterative algorithm without statistical modelling
    • used originally by Hounsfield however not commercially used due to computer limitations
    • will use an assumption and will compare to the assumption with its measured data. Then will continue to make iterations until the two data sets are in agreement.
  • iterative algorithm with statistical modelling
    • iterative reconstruction with statistical modelling that takes into account either
    • optics (x-ray source, image voxels and detector)
    • noise (photon statistics)
    • physics (data acquisition)
    • object (radiation attenuation)
  • back projection
    • not used in the clinical setting, as it is unable to produce sharp images
    • known for its distinctive artefact that resembles a star
  • filtered back projection (convolution method)
    • still widely used in CT today
    • utilizes a convolution filter to alleviate the blurring associated with back projection
    • fast, however, has several limitations including noise and artefact creation
Physics and imaging technology: CT
Share article

Article information

rID: 51829
Section: Physics
Synonyms or Alternate Spellings:

Support Radiopaedia and see fewer ads

Updating… Please wait.

 Unable to process the form. Check for errors and try again.

 Thank you for updating your details.