Accurate Interactive Visualization of Large Deformations and Variability in Biomedical Image Ensembles

摘要: Large image deformations pose a challenging problem for the visualization and statistical analysis of 3D ensembles which have multitude applications in biology medicine. Simple linear interpolation tangent space ensemble introduces artifactual anatomical structures that hamper application targeted visual shape techniques. In this work we make use theory stationary velocity fields to facilitate interactive non-linear plausible extrapolation high quality rendering large devise an efficient warping method on GPU. This does not only improve existing techniques, but opens up field novel methods analysis. Taking advantage warping, showcase four visualizations: 1) browsing on-the-fly computed group mean shapes learn about differences between specific classes, 2) reformation investigate complex morphologies single view, 3) likelihood volumes gain concise overview variability 4) streamline show variation detail, specifically uncovering its component tangential reference surface. Evaluation real world dataset shows presented outperforms state-of-the-art terms while retaining frame rates. A case study with domain expert was performed are applied standard model structures, namely skull mandible different rodents, compare influence phylogeny, diet geography shape. The visualizations enable instance distinguish (population-)normal pathological morphology, assist correlation extrinsic factors potentially support assessment quality.