Imaging Research Laboratories presents:

 
2 related talks by:

 
/_Kenan Prof. Stephen Pizer_/

of UNC & Morphormics, Inc.

 
with the overall title:

 
*"**Segmentation and 3D visualization from medical images using skeletal models**"*

 

         Robarts Research Institute, 2^nd floor conference room

 

         Monday, June 14^th @ 4:00pm ? 5:00pm

			*&*

         Tuesday, June 15^th @ 11:00pm ? 12:00pm

(The second talk is designed on the assumption that listeners have also attended the first)

 
 
*/Talk 1/* (Monday at 4pm): /Segmentation via posterior optimization of skeletal models and quantile function appearance models/. Co-authors: Edward Chaney, Eli Broadhurst.

 
Abstract: Topics to be covered: Skeletal object models, estimating probability distributions and shape spaces on skeletal models, estimating probability distributions on appearance via quantile functions, and segmentation by initialization, posterior optimization, and refinement. Application to and evaluation of a clinically available system for segmentation of male pelvic structures from CT for radiation treatment planning.

 
*/Talk 2/* (Tuesday at 11am): /Prostate segmentation and 3D visualization for prostate biopsy guidance, from 3D TRUS after MRI-&-MRSI imaging/. Co-authors: Mark Foskey, Derek Merck, Julian Rosenman.

 
*/Abstract:/*/ /

Our purpose is to transfer local values on prostate malignancy suspicion determined from MRI & MRSI to a segmentation of the prostate at TRUS (biopsy) time, and to visualize the suspiciousness within the prostate at TRUS time to guide the biopsy needle. Segmenting prostates from 3D trans-rectal ultrasound images, taken during biopsy, given a prostate segmentation from the MRI, is done via posterior optimization using skeletal models, a shape space of prostate deformations between MR and TRUS imaging, and a TRUS quantile-function-based appearance model involving not only intensity features but also texture features. The 3D visualizations that we anticipate for this biopsy will leverage our methods of visualization of multiple sources of information using the object-relative coordinates given by our skeletal models. This visualization approach has been seen to be useful in radiation treatment planning, where radiation dose, CT images, and segmented objects need to be understood together.

*/ /*