The NIH/NIGMS
Center for Integrative Biomedical Computing

Identifying and Quantifying Shape Features of Human Left Ventricular Remodeling

Identifying and Quantifying Shape Features of Human Left Ventricular Remodeling

Raimond Winslow

Johns Hopkins University

Background

The goal of this collaboration is contribute to the set of tools and technologies available to study the effects coronary disease on the shapes of the left ventricle and, additionally, to develop quantitive methods for evaluating the extent of remodeling as a predictor of cardiovascular mortality.1

The LV shapes are derived from segmented multidetector computed-tomographic (MDCT) images. The technology in ShapeWorks could offer a powerful alternative the image-deformation-based methods described in.2

Preliminary results are quite promising. We have applied ShapeWorks to assign correspondences on an ensemble of 25 hearts, consisting of both normal controls and patients who have suffered a myocardial infarction. A parallel analysis indicates five significant modes of variability for this ensememble. A Hotelling-T test indicates significant shape differences between the normal controls and the patients.

This analysis of LV shape does present some challenges for the particle-based strategy in ShapeWorks. For instance, the high curvature region that separates the inside and outside of the ventrical represents a barrier to particle movement, and correspondence points on the wrong side of this barrier sometimes have difficulty establishing good positions using the gradient-descent optimization strategy. For these results, we have introduced an automatic mechanism for culling points that exhibit bad positioning, which is determined from the surface normals. In future we will pursue more systematic approaches to such practical problems, such as including the surface normal information into the shape-space description. This work will also progress on the clinical front.

Cited References

  1. S Ardekani, R.G. Weiss, A.C. Lardo, R.T. George, J.A.C. Lima, K.C. Wu, M.I. Miller, R.L. Winslow, and L. Younes. "Computational method for identifying and quantifying shape features of human left ventricular remodeling". Annals of Biomedical Engineering, 2009.
  2. S. Ardekani, R.G. Weiss, A.C. Lardo, M.I. Miller, R.L. Winslow, and L. Younes. "Cardiac motion analysis in ischemic and non-ischemic cardiomyopathy using parallel transport". In Proc. IEEE Int. Sym. on Biomedical Imaging (ISBI '09), June 2009.