Fast and accurate computation of system matrix for area integral model-based algebraic reconstruction technique

Shunli Zhang, Dinghua Zhang, Hao Gong, Omid Ghasemalizadeh, Ge Wang, Guohua Cao

Research output: Contribution to journalArticlepeer-review


Iterative algorithms, such as the algebraic reconstruction technique (ART), are popular for image reconstruction. For iterative reconstruction, the area integral model (AIM) is more accurate for better reconstruction quality than the line integral model (LIM). However, the computation of the system matrix for AIM is more complex and time-consuming than that for LIM. Here, we propose a fast and accurate method to compute the system matrix for AIM. First, we calculate the intersection of each boundary line of a narrow fan-beam with pixels in a recursive and efficient manner. Then, by grouping the beam-pixel intersection area into six types according to the slopes of the two boundary lines, we analytically compute the intersection area of the narrow fan-beam with the pixels in a simple algebraic fashion. Overall, experimental results show that our method is about three times faster than the Siddon algorithm and about two times faster than the distance-driven model (DDM) in computation of the system matrix. The reconstruction speed of our AIM-based ART is also faster than the LIM-based ART that uses the Siddon algorithm and DDM-based ART, for one iteration. The fast reconstruction speed of our method was accomplished without compromising the image quality.

Original languageEnglish (US)
Article number113101
JournalOptical Engineering
Issue number11
StatePublished - Nov 1 2014


  • Siddon algorithm
  • algebraic reconstruction technique
  • area integral model
  • computed tomography
  • image reconstruction

ASJC Scopus subject areas

  • Atomic and Molecular Physics, and Optics
  • General Engineering


Dive into the research topics of 'Fast and accurate computation of system matrix for area integral model-based algebraic reconstruction technique'. Together they form a unique fingerprint.

Cite this