Development of a Preprocessor for the Simulation of Human Bones.

bones5

Abstract

One approach for the simulation of human bones is the Finite Cell Method (FCM), which takes advantage of voxel-based data obtained from computed tomography scans (CT-scans). But before entering the simulation phase, some operations on the initial data are needed, such as reading & visualizing the data and extracting the volume of interest (VOI). Goal of this project was to implement a preprocessor which operates interactively on patient‘s CT-scans and provides the appropriate input to the Finite Cell Method.

Starting point

CT-scanner (Fig.1) produce series of slices (Fig.2). The initial form of a slice is a matrix of numbers (Hounsfield Units) which is stored in DICOM file-format (contains patient, scan & image info). Combining all slices allows to reconstruct the scanned body, but includes also the soft tissues (Fig.3,4). The latter needs to be removed if only the bones are of interest.

pipeline_bones

Medical Processor in Action

By using the Graphical User Interface, the user can open any patient‘s DICOM-files & apply various rendering techniques to it(Fig.5). A box widget enables the user to dynamically choose a Volume-Of-Interest (VOI), which can be inspected seperately.

bones6

After choosing the VOI, the voxel data of the selected VOI can be exported. As a result, the extracted volume is a subset of the initial data that consists of voxels again.
Because each voxel has a value on the Hounsfield scale, which describes the relative attenuation of X-rays from substances, iso-surfaces can be used to approximate the geometry of different body-parts (here the bones).

bones2

In order to ensure that the exported data contains the correct values of the VOI, ParaView was used to visualize them (Fig.6).

bones3

Future Work-Outlook

Additional functionalities should be added to the medical preprocessor in order to make it complete. For instance, after selecting the portion that shall be simulated, the VOI needs to be cleaned from the soft tissue by a threshold based segmentation. Also, the preprocessor should be able to automatically identify Neumann and Dirichlet boundary conditions.

bones4