Thorsten Heuser, Presenter: Nothing to Disclose

Marcus Brehm, Abstract Co-Author: Nothing to Disclose

Ludwig Ritschl, Abstract Co-Author: Nothing to Disclose

Stefan Sawall DIPLENG, Abstract Co-Author: Nothing to Disclose

Marc Kachelriess PhD, Abstract Co-Author: Nothing to Disclose

To improve image quality in x-ray computed tomography (CT) in cases of missing or corrupt data.

CT image quality often suffers from artifacts due to missing or corrupt data. Numerous approaches to reduce these artifacts have been published. These approaches typically use inter- or extrapolation techniques specific for the kind of artifact investigated and in many cases tend to introduce new artifacts. Often, however, prior data are available which can, potentially, be used to better compensate for the missing or corrupt data and thereby to help reduce artifacts without introducing new artifacts. These prior data may be images from a different scan of the same patient, e.g. a planning CT, or images of a similar patient taken from a patient data base. We propose a prior-based artifact correction (PBAC) algorithm, a generalized correction method for prominent artifacts in CT resulting from missing or corrupt data. To compensate for differences in patient shape and position PBAC performs a non-rigid registration to match the prior with the patient, and it accounts for differences in the CT values by histogram matching. The registered prior is forward projected and its projections are used to smoothly inpaint the missing or corrupt data regions into the patient projections. Image reconstruction of the obtained projection data results in the corrected image. PBAC is evaluated using several data sets measured with a clinical spiral cone-beam CT device (Somatom Definition Flash, Siemens Healthcare, Forchheim, Germany).

Compared to the uncorrected CT images PBAC reduces the artifacts by 79% in case of metal pedicle screws in a thorax scan, by 99% for a hip patient with 12.2 cm truncation, and by 67% in case of a head scan with a limited scan angle of 150°. In all cases PBAC significantly outperforms the conventional correction methods which achieve artifact reduction values of only 42%, 90%, and 5%, respectively.

PBAC is a highly effective method to correct for CT artifacts resulting from missing or corrupt data by making use of prior knowledge. It significantly increases the image quality and preserves the patient-specific anatomy to allow for reliable medical diagnosis.

PBAC is relevant for clinical CT which often suffers from metal artifacts, and it is relevant for flat detector CT which additionally suffers from truncation or limited angle artifacts.

Heuser, T, Brehm, M, Ritschl, L, Sawall, S, Kachelriess, M, Prior-based Artifact Correction in Computed Tomography.  Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL. http://archive.rsna.org/2013/13015657.html