DCS Summer 2012 Research Awards -- Project Description

Algorithms for Medical Imaging

Faculty name: Ken Jackson
Research area: Scientific Computing
Campus address: BA 4228
Campus phone: 416-978-7075
Email address: krj [at] cs.toronto.edu
Number of students: 1-2
Skills required:
  • Strong math skills:
    • calculus (2 years)
    • linear algebra
  • Strong CS background:
    • numerical computing
    • strong programming skills

Brief project description:

One of the problems that our group has recently studied is how to cope with "beam-hardening" in computerized tomography (CT). This arises because most CT image reconstruction algorithms assume that the x-rays used for the CT scan all have the same energy. This makes the mathematical problem of reconstructing the image much simpler. However, in reality, the x-rays used in CT scans do not all have the same energy. This mismatch between the assumptions of the algorithm and the true physics of the problem results in "beam-hardening" artifacts in some images. This produces streaking and blurring in some images and generally reduces the image quality.

We have developed a CT scan reconstruction algorithm that takes into account the poly-energetic nature of the x-rays used in CT scans. One of the things that we would like to do is explore the parallelization of this algorithm. In particular, we would like to develop a version of the algorithm to run on parallel systems based on GPUs (Graphics Processing Units). Such systems have become very popular recently because they are a very cost-effective way to solve problems of this type.

Back to the index.