Parallel deformation of heterogeneous ChainMail models: Application to interactive deformation of large medical volumes.

In this work we present a new solution for correctly handling heterogeneous materials in ChainMail models, which are widely used in medical applications. Our core method relies on two main components: (1) a novel timestamp-based propagation scheme that tracks the propagation speed of a deformation through the model and allows to correct ambiguous configurations, and (2) a novel relaxation stage that performs an energy minimization process taking into account the heterogeneity of the model. In addition, our approach extends the SP-ChainMail algorithm by supporting interactive topology changes and handling multiple concurrent deformations, increasing its range of applicability. Finally, we present an improved blocking scheme that efficiently handles the sparse computation, greatly increasing the performance of our algorithm. Our proposed solution has been applied to interactive deformation of large medical datasets. The simulation model is directly generated from the input dataset and a user defined material transfer function, while the visualization of the deformations is performed by rendering the resampled deformed model using direct volume rendering techniques. In our results, we show that our parallel pipeline is capable of interactively deforming models with several million elements. A comparison is finally discussed, analyzing the properties of our approach with respect to previous work. The results show that our algorithm correctly handles very large heterogeneous ChainMail models in an interactive manner, increasing the applicability of the ChainMail approach for more demanding scenarios both in response time and material modeling.