This section covers the processing aspect of acquiring an image.
Physics
A CT image is made up of pixels along a greyscale. What determines the level of grey is the density of the material, also expressed as the linear attenuation coefficient, and this is represented numerically by the Hounsfield Units (also called the CT number). The Hounsfield units are set so that water measures 0 and everything else is relative to this.
μt = attenuation coefficient of tissue
μw = attenuation coefficient of water

Each detector in the CT scanner samples a line of the patient and the sum total of the attenuation of the material passed through along the beam path is calculated. As the gantry rotates the detectors receive beams at different angles so, in the end, we have a series of values of summed linear attenuation coefficients from different angles. Now, these need to be processed to form an image.
Typical Hounsfield unit values
Tissue | CT number (HU) |
Bone Liver White matter Grey matter Intravascular blood Fresh clotted blood Muscle Kidney CSF Water Fat Air | +1000 40 to 60 20 to 30 37 to 45 30 to 45 70 to 80 10 to 40 30 15 0 -50 to -100 -1000 |
Written by radiologists, for radiologists with plenty of easy-to-follow diagrams to explain complicated concepts. An excellent resource for radiology physics revision.
Post-Processing
Backprojection

There are a few main issues with backprojection:
1. Too few projections cause artefacts in the image as there are too few directions of summed LACs to accurately represent the image. Typically 2000 projections are used.
2. Even with a large number of projections the edges of structures are not well delineated due to the averaging out of values and there is blurring caused by the backprojection technique. This is corrected with filtered backprojection.
Iterative Reconstruction
This is generally a more time-consuming method but is proving useful for low dose CT studies.
It involves several steps:
- Filtered backprojection is initially performed to assign a number value to all pixels in the matrix.
- The computer then calculates what it expected the detectors to have received based on the image generated THEN works out the difference between the actual detector measurements and the calculated measurements. It then uses this information to generate an updated image.
- This continues through multiple iterations, each time bringing the calculated values closer and closer to the true values.
If you want further information on iterative reconstruction and backprojection a good website is: http://www.dspguide.com/ch25/5.htm
Σ Summary
- Image is made up of pixels of varying grey, the shade of which is assigned a “Hounsfield Unit” (also called “CT number”) which is compared to a look-up-table to give the greyscale.
- The x-ray beam and detectors rotate around the subject sampling rows at different angles. Each row is coded as a single summed attenuation value.
- The attenuation values are then processed to produce the image mainly via two techniques
- Backprojection: The summed attenuation values are averaged out over the row. With several projections it comes closer to actual image. There are some weaknesses:
- Too few projections cause artefacts
- Blurred images – solved by filtered backprojection
- For multislice scanners filter interpolation is used in which all projections within a certain axial slice are summed and averaged.
- Iterative reconstruction: Filtered backprojection is initially performed to assign a number value to all pixels in the matrix. The computer then calculates what it expected the detectors to have received based on the image generated and compares this to the actual detector measurements, adjusting the image values to bring them closer to the true values.
- Almost exclusively used now.
- Weakness: Calculations are lengthy
- Strength: Reducing CT dose