Publication details
Towards an efficient data assimilation in physically-based medical simulations
| Authors | |
|---|---|
| Year of publication | 2015 |
| Type | Article in Proceedings |
| Conference | Proceedings of IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 2015 |
| MU Faculty or unit | |
| Citation | |
| Web | http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=7359884&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D7359884 |
| Doi | http://dx.doi.org/10.1109/BIBM.2015.7359884 |
| Field | Informatics |
| Keywords | Data Assimilation; Kalman Filtering; Non-linear elasticity; Finite Element Method; Patient-specific Modeling |
| Description | Computer simulation of soft tissues is rapidly be- coming an important aspect of medical training, pre-operative planning and intra-operative navigation. Whereas in medical training, generic models are usually employed, both planing and navigation require patient-specific modeling. However, creating a patient-specific model is a challenging task, as many of the mechanical parameters of the organ tissues are unknown. One way of addressing the issue is to extend the deterministic simulation by methods based on stochastic modeling. In this paper we focus on parameter estimation in models with large number of degrees of freedom based on a variant of Kalman filtering. The main contribution of the paper is a detailed description of an integration of two advanced concepts of numerical modeling: we employ a state-of-the-art method of data assimilation based on reduced-order Kalman filtering in order to perform parameter estimation of a finite-element model of non-linear elasticity used in medical simulations. In order to assess the method, we present a preliminary evaluation of the accuracy of the parameter estimation as well as the performance using synthetic data with added noise. We also evaluate the parallelized version of the prediction phase and finally we describe further perspectives which, as we believe, will bring the data assimilation of models with many parameters closer to the real-time processing. |