Research Abstracts Online
January - December 2011
University of Minnesota Twin Cities
College of Science and Engineering
of Electrical and Computer Engineering
PI: Ahmed H. Tewfik
Real-Time Cardiac Surface Tracking Using Sparse Representations and Sequential State-Estimation Techniques
Minimally invasive surgeries such as catheter-based radio frequency ablation of atrial fibrillation require more visual and contextual information for higher success rates. Real-time three-dimensional tracking of the inner and outer surfaces with sub-millimeter precision is required. In this project, the researchers are developing a real-time three-dimensional cardiac motion tracker using Kalman filtering/state-space model approach using low-density imaging of the cardiac surface.
The researchers recently developed a novel approach to track three-dimensional organs in real time from subsamples using spherical harmonics (SPHARM) transform and subspace clustering. Pre-operative volumetric MRI/CT scans of organs are converted into three-dimensional surfaces and shape-modeled using spherical harmonic transform. Active organ deformation spaces in the SPHARM domain are identified using subspace clustering and accurate real-time three-dimensional reconstruction of organs is performed from low-density three-dimensional samples under the least squares formulation. For this project, the researchers are extending this approach for organs exhibiting natural deformations, such as the heart, using sequential state estimation techniques to incorporate natural temporal model redundancy into our spatial subspaces to allow more-accurate tracking of cardiac surfaces under high under-sampling.
Vimal Singh, Graduate Student