Biomedical Engineering Reference
In-Depth Information
quality meshing are the most difficult to automate and carry out efficiently.
Manual segmentation of high-resolution volumetric image is a tedious and irrepro-
ducible task, which is impractical for processing large amount of data. Fully automatic
and unsupervised methods, while having already received significant attention in
the literature are still challenging [
4
]. In particular, the segmentation of the pathology
(tumor) often requires intensivemanual interaction to achieve good or even acceptable
results.
After segmentation, finite element meshes are constructed based on surfaces
extracted from the segmentation results. The accuracy of the finite element calculation
relies heavily on element mesh quality. Although a wide range of automatic mesh
generation techniques are currently available, these are usually developed to generate
meshes fromcomputer-aided design (CAD), and therefore have difficulties generating
good-quality meshes from highly irregular surfaces such as segmented imaging data.
Even when using IA-Mesh [
5
]—one of the latest developments in meshing for
biomechanics—an experienced analyst is required to manually correct a mesh,
which is prohibitively time consuming for large applications involving many cases.
To overcome the difficulties of mesh generation in FEM, meshless method provides
an alternative way for discretizing the computational domain using unstructured could
of nodes [
6
]. Meshless nodes are placed almost arbitrarily, which provides high
flexibility of nodes location within the computational domain, even without internal
boundary geometries. In the context of patient-specific biomechanical modeling of the
brain, where intracranial structures of the brain are hard to segment accurately, the
flexibility of meshless method makes it possible to generate the computational grid of
intracranial area without well-defined tissue boundaries. To address the material
properties differences (Young's modulus and Poisson's ratio in the present study); we
employ soft tissue classification to get the fuzzy membership functions of each tissue
class for each integration points. The interpolated material properties based on fuzzy
membership functions are assigned directly to the corresponding integration points.
Noted in Fig.
1b
, the node placement is independent about the intracranial tissue
classification. Therefore in the proposed statistical meshless framework, the node
placement procedure and fuzzy tissue classification can be done simultaneously. This
could substantially reduce the modeling time, as shown within the brace in the pipeline.
This paper proposes a novel statistical meshless approach that addresses these
issues by a combination of
• Abandoning the finite element meshes and embracingmeshless computational grids
• Abandoning “hard” segmentation and utilizing statistical “soft”/“fuzzy” tissue
classifications
which could be processed simultaneously and has never been proposed in the
literature before.
We describe in detail our statistical meshless approach in Sect.
2
. In Sect.
3
,we
generate both finite element model and a statistical meshless model from a brain
MRI slice. We use the same constitutive properties, boundary conditions, and
loading, and compare simulation results obtained with both models. Section
4
contains conclusions and discussion.
