Researchers model breast CT compression, deformation

Tuesday, December 1 | 11:30 a.m.-11:40 a.m. | SSG16-07 | Room S502AB

Researchers are working toward automated lesion segmentation in breast image analysis, but targeting breast compression and localization is the first step -- and the subject of this presentation from the University of Michigan.

"The purpose of the study is to model breast compression and target localization using breast CT volumes," Ravi Samala, PhD, told AuntMinnie.com in an email.

First, they had to study breast compression and deformation using biomechanical models, generating high-resolution models from breast CT, using a finite element method to model breast compression.

Because breast CT is performed without compression, compensation was made in the gravity loading of the breast CT volume, and the data were compressed until they reached the thickness of a corresponding digital mammography view. The pilot study modeled 10 breast CT volumes and corresponding digital mammograms.

The uncompressed 3D breast CT method offers the possibility of modeling breast compression and deformation accurately with an error of less than 1 cm, the group concluded.