Robert Michael Lapp, Presenter: Nothing to Disclose

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

Willi A. Kalender PhD, Abstract Co-Author: Nothing to Disclose

To allow for the adaptation of spatial resolution and image noise in real time in small-animal imaging (SAI).

Image reconstructions in SAI with typical reconstructed volumes of 1024³ voxels are time consuming with processing times of typically 30-60 minutes (720 projections, detector size 1024², Intel Xeon 2x3 GHz, 3 GB RAM). Also, the data set sizes of 2 GB demand large storage capacities. Multiple reconstructions with different reconstruction kernels to influence image characteristics such as noise and resolution are therefore becoming exceedingly inconvenient. An image-based convolution approach is presented which allows for the modification of noise and resolution of the images in real time while maintaining the nearly isotropic resolution of the Feldkamp-type reconstruction. The technique estimates short 3D convolution kernels from reconstruction kernels in frequency domain which are then used to generate image characteristics equivalent to direct reconstructions. The kernels are separated for performance reasons and applied in x-, y- and z-direction. To evaluate the performance of the method, data were acquired from a SAI scanner (Tomoscope 30s, VAMP GmbH, Germany) and volumes were reconstructed. The filtering technique was applied for various reconstruction kernels, and images were compared to direct reconstructions. We quantitatively evaluated image characteristics such as spatial resolution and image noise, and calculated difference images.

Filtered images matched direct reconstructions with typical deviations of +/- 3 HU in image noise level and 5% of the FWHM of the point spread function. Processing times were in the range of 75-250 ms per image and thus around 75-250 s for a complete volume depending on the length of the convolution kernels. The fast processing times for single slices or for selected subvolumes also allow for an interactive manipulation and provide the possibility to achieve optimal noise/resolution settings depending on the anatomical region and diagnostic needs.

The proposed technique abolishes the need for multiple reconstructions for many applications and allows for an interactive balancing between resolution and noise.

Lapp, R, Kachelriess, M, Kalender, W, Variable Isotropic Resolution Filtering in Small Animal Imaging.  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4417770.html