Medical simulation applies a numerical algorithm that synthesizes a clinical effect on a digital representation of patient anatomy.
Medical simulation is a technology that applies a numerical algorithm that synthesizes a clinical effect on a digital representation of patient anatomy. In the case of surgery simulation, this algorithm typically replicates a bio-mechanical interaction between a surgical tool and a tissue model.
The tissue model, used in surgery simulation, is generally obtained from patient data through a segmentation stage, which is either manual or minimally supervised, followed by a meshing stage that represents the anatomy in terms of simple shapes, such as tetrahedra or hexahedra. This meshing stage is a requirement for efficient computation, otherwise, without its inclusion, each anatomical model would be comprised of millions of voxels, infeasible for efficient computation.
In addition, there is a trade-off between the constitutive faithfulness of existing biomechanical models and their computational efficiency, with the result that there are traditionally two types of simulation:
i) Predictive simulation, features off-line finite element computations that produce results used by expert surgeons to plan their intervention;
ii)Interactive simulation, used to train surgical residents, by way of a haptic device used to input surgical gestures and output force feedback and which, until recently, emphasized relatively inexpressive but computationally efficient constitutive models, such as mass-spring systems.
New Constitutive Methods for Biomechanically Faithful Interactive Simulators
While traditionally, predictive and interactive simulation are seen as opposite poles in the spectrum, characterized by the faithfulness vs speed trade-off, recent innovations will increasingly allow interactive simulation to exploit high-fidelity nonlinear finite elements (FE).
The first of these is multigrid finite elements, which consists of a mesh representation of the anatomy at several levels of resolution. Multigrid finite elementsuse a coarse-level system, which can be solved efficiently, to accelerate the convergence of medium- and fine-level representations. The clinical relevance of this approach is that it provides a mechanism for reconciling the depiction of small critical structures, such as blood vessels, with computations over relatively large organs, such as the liver or the brain.
Another innovation is the application of pre-computations, where the topology of the tissue does not change, such as with a spatula or a clamp, as exploited by the total Lagrangian explicit dynamics (TLED) formulation. This method expresses FE computations in terms of undeformed coordinates (i.e. the tissue state prior to the simulation). As a result, TLED precomputes many variables prior to the simulation taking place. Moreover, the computations that do occur in real-time can exploit the Graphical Processor Unit (GPU), with an order of magnitude improvement over CPU implementations. This approach has since been further improved by another order of magnitude by restricting the FE system to a reduced basisof feasible deformations.
In addition to FE modeling for tissue manipulation, descriptive methods are also being developed for cutting and resection. EXtended FE Modeling (XFEM) is a FE adaptation that admits discontinuities in the "shape function," and has led to highly realistic, interactive cutting.
A second approach to cutting and resection is the application of so-called meshless methods, which is a particle-based approximation technique that obviates tissue re-meshing.
Many of these methods are, or will soon be, available in open-source. Currently, the most active open-source platform for medical simulation is Simulation Open Framework Architecture, or INRIA-SOFA, which comes with a LGPL license. SOFA's next release will include a TLED implementation. Furthermore, reduced-basis TLED FEs are available via Sourceforge. Meanwhile, University of Aachen researchers are currently integrating XFEM to the SOFA platform. There is also support for XFEM in OOFEM, but this comes with a GPL license. Neither multi-grid nor meshless methods are yet available in open-source, but planned work at Kitware, including collaborative work with Robarts Institute (London, Ont., Canada), will address this in conjunction with the SOFA platform.
Patient-specific, Descriptive Anatomical Models
An additional area where surgery simulation will - and must - prove to be clinically relevant is the application of increasingly more descriptive anatomical models, especially in clinical areas currently under-represented by the state-of-the-art, such as neurosurgery.
Currently, the emphasis of existing techniques is on generic, rather thanpatient-specificsurgery simulation. The traditional approach to simulation generally involves a significant amount of manual segmentation, which essentially precludes patient-specific modeling.
With the availability of robust segmentation methods in open-source, such as EM Segmenter, which runs on Slicer 3D (3D Slicer), and Kitware's TubeTK, which will support blood vessel segmentation, collaborative software efforts may hold the key to endowing simulation with patient-specificity.
Rigorous Clinical Specification
Finally, a key factor to producing simulators that fulfill real clinical requirements, rather than merely reflecting engineers' technical interests, is the formal specification of these clinical requirements through surgical ontologies. This area of research is dedicated to the description of surgical processes in terms of high-level steps and generally proceeds as a collaborative effort between clinicians and software engineers, on the basis of either textbook information or observing a large number of patient-specific procedures. Open-source software tools also exist in this area, namely Protege.