logo Back to projects page

Objectives

Computational techniques for the analysis of mechanical problems have recently moved from traditional engineering disciplines to biomedical simulations. Thus, the number of complex models describing the mechanical behavior of medical environments have increased these last years. While the development of advanced computational tools has led to interesting modeling algorithms, the relevances of these models are often criticized due to incomplete model verification and validation. The objective of this validation environment is to propose a framework and a methodology for assessing deformable models. It aims at providing tools for testing the behavior of new modeling algorithms proposed in the context of medical simulation. The framework of this validation environment is based on SOFA, where several algorithms are already implemented, allowing to share different reference models and different solutions from existing modeling methods.

Articles related to this project

- M. Marchal, J. Allard, C. Duriez, S. Cotin. "Towards a Framework for Assessing Deformable Models in Medical Simulation". In Proceedings of International Symposium on Computational Models for Biomedical Simulation. pp. 176-184. 2008. (PDF)


Measurements and metrics

To validate a deformable model used for the simulation of a medical environment, different performances criteria have to be defined. For the moment, we focus on the accuracy criterion. Validation environment users are welcome to contribute to the definition of new criteria.

Accuracy measurement

To measure accuracy performances of the different modeling methods, two different types of metrics are proposed, depending on the type of available reference data. For both types, the error can be an absolute value, taking into account the displacement value or a relative value independent of the displacement of the simulated object. We can divide the accuracy metrics into two parts, depending on the reference data markers:

  • if the reference data contains markers (the mesh of an analytical solution or solid markers inside a phantom for example):
    An error metric based on a distance between nodes/points can be defined. In the data available in this validation environment, the relative energy norm error is used.
  • if the reference data do not contain any marker (the reference data give information about their global shape (curve, surface, etc)):
    An error metric based on a measurement of the distance between the reference and the simulated shapes has to be defined. In this validation environment, the error metric used in the examples is a distance between two surfaces.


Analytical and experimental solutions

The overall objective of a validation process is to guarantee that :

  • the numerical approximation of the mathematical equations chosen for governing the model is acceptable,
  • the model provides an accurate representation of the physical behavior of the problem of interest within a given computation time.
Both assumptions need to be verified within an assessment of error in the model predictions and their achievement relies on a combination of analytical solutions and experimental data.

Analytical solutions
Experimental solutions