Loading...

Joint Segmentation and Motion Estimation of Cardiac Cine MR Image Sequences

Eslami, Abouzar | 2010

517 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 41101 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Jahed, Mehran
  7. Abstract:
  8. In this study a variational framework for joint segmentation and motion estimation is provided for inspecting heart in Cine MRI sequences. In the first work, a functional including Mumford-Shah segmentation and optical flow based dense motion estimation is proposed. However, the motion estimation technique is replaced with warping estimation in the second work to reach in more regular and smooth motion field based on tracking of cardiac boundaries. Both of these functionals are then approximated by using the phase-field method to make them suitable for extracting Euler-Lagrange equations. Numerical solution to the optimization problem, when the time and image spaces are discretized by finite element method, is proposed and its implementation considerations, e.g. memory usage and computational complexity are discussed. Several simulation runs are used to test the convergence and the parameter sensitivity of the method and to show the efficacy of the proposed method in specific conditions such as edge weakness in some frames and intensity fluctuation. In order to validate the proposed method, a pair of tagged and non-tagged Cine MR Image sequences is taken from a dog’s heart. The motion filed resulted from processing the non-tagged Cine MRI by our method is compared with the sensed motion field in tagged data using HARP method. Local and global measures of their distance are reported to investigate the motion estimation error of the proposed method both over a heart cycle and over different regions of heart. The proposed method is further applied to a comprehensive set of clinical data and the results are compared with conventional cascade methods of segmentation and motion estimation methods of cardiac Cine MRI. For clinical experiments the cardiac boundaries were delineated by and expert to provide a ground truth so that the extracted and tracked boundaries resulted from our method be compared with. Based on the results, joint segmentation and motion estimation outperforms previously reported cascade approaches as a consequence of taking the advantage of interdependency of segmentation and motion estimation. The main idea of joint segmentation and motion estimation is to exploit the correlation of subsequent frames (usually 5) after motion estimation to attain better segmentation and simultaneously utilizing the segmentation result to improve motion estimation efficacy. So, the minimizer of the joint tasks functional includes optimum motion field and optimum edge set when considering both spatial and temporal discontinuities in processing the Cine MRI sequence
  9. Keywords:
  10. Segmentation ; Motion Estimation ; Optical Flow ; Cardiac Cine Magnetic Resonance Imagin (MRI) ; Joint Tasks Processing ; Edge Tracking ; Segmentation

 Digital Object List

 Bookmark

No TOC