Statistical Models of Appearance for Functional Analysis of Cardiac MRI

We present a framework for the analysis of cardiac MRI using Statistical Models of Appearance. This thesis makes three major contributions. The first contribution involves the introduction of a new algorithm for fitting 3-D Active Appearance Models on cardiac MRI, using the inverse compositional image alignment algorithm. We observe a 60-fold increase in fitting speed and an accuracy that is on par with Gauss-Newton optimization. The second contribution involves an investigation of the use of wavelets in hierarchical Active Shape Models, as a potential way of making them more expressively powerful. The third contribution involves an investigation of the use of adaptive filtering for high quality resampling of 4-D cardiac MR images. We show the high quality results that are derived by the use of adaptive filtering, and describe the ways in which it could improve the automated analysis of medical images.