Print
Univis
Search
 
FAU-Logo
Techn. Fakultät Website deprecated and outdated. Click here for the new site. FAU-Logo

Dr. Volker Daum

Alumnus of the Pattern Recognition Lab of the Friedrich-Alexander-Universität Erlangen-Nürnberg

To make the information contained in different imaging modalities and multiple acquisitions useable for the improvement of medical diagnosis as well as (semi-)automatic segmentation and classification.

Current projects and research interests

Nonrigid Medical Image Registration

 

The aim of registration algorithms is to match images or volumes, such that they can be compared in the same frame of Reference. This is necessary for longitudonal studies, inter patient comparisons (mono-modal) and the combination of different imaging modalities (multi-modal) like PET and CT. In nonrigid registration each pixel is allowed to move independently to some extent, while a so-called regularizer ensures that the deformation remains locally smooth.

Animation of a nonrigid registration process, on an MR T2 image pair. Top left: deforming moving image; top right: contour overlay with reference image; bottom: x and y component of the deformation field
Animation of a nonrigid registration process, on an MR T1-T2 image pair. Top left: deforming moving image; top right: contour overlay with reference image; bottom: x and y component of the deformation field

PET-MR Attenuation Correction

 

In PET imaging the patient is injected with a radioactive tracer substance, that, due to the metabolism in the body, accumulates in regions of interest, like cancerous tissue. The radiation emanating from the patient can be measured by a detector. The measured radiation is influenced by two things:

  1. The tracer concentration in a body region
  2. The tissue the radiation has to pass through on its way to the detector, which attenuates the radiation

In PET imaging one only wants to measure the tracer concentration, not the attenuation. The measured data therefore has to be corrected for the attenuation (attenuation correction). Normally this is done with the help of an attenuation map generated by a CT or a transmission scan. In hybrid PET-MR machines however only an additional MR imaging sequence is performed, but MR measurements do not give direct information about attenuation.

The aim of this project is, therefore, to explore ways to generate attenuation maps from MR images. One way to tackle this problem is by atlas registration. An atlas attenuation map (for example another patient's CT) is nonrigidly registered to the patient MR, thus generating a pseudo-CT that can be used for the attenuation correction.

 

From top left to bottom right: atlas CT, patient MR, ground truth patient CT, atlas registered pseudo-CT

Full Body PET Difference Imaging

 

Difference imaging is helpful in assessing the changes between two datasets. In PET imaging for instance this can help to quickly identify spots of tumor growth or recession. Normally the physician would have to examine the both full body PET scans for any changes, which can be a rather lengthy and demanding process. In a difference, ideally only the points where the activity has changed due to a change in patient physiology will show.

In practice several other effects would show in a straightforwardly generated difference image. The first effect is the difference in overall tracer concentration. This can be taken care of by a normalization that ensures that overall tracer concentration is similar in the normalized images. The second big effect is patient motion. As the scans that have to be subtracted are taken at different points in time, there is no way to assure that the patient position in the machine is exactly as it was during the first scan. This is solved by retrospective image registration (rigid and nonrigid), that accounts for patient motion. As in the examinations we work with all scans performed are PET-CT we use the CT for registration to get maximum registration accuracy and to be less influenced in the registration, by the changes in tumor activity.