Semiautomatic Generation of a Three-Dimensional Human Anatomical Model of Bone for Biomedical Applications: First Approach

Third-dimensional human anatomical models have been used to better understand the human body. However, one of their most important uses is the personalization of any type of treatment. 3D human models of bones from the lower extremities are crucial for visualization, treatment planning, and orthotic and prosthetic design. Nonetheless, this 3D models must be compatibles with any modeling software, like those working with the finite element method. Therefore, a methodology to generate 3D anatomical models compatible with CAD systems was proposed. A couple of sets of Computed Tomography (CT) images in DICOM (Digital Imaging and Communications in Medicine) format was used. To reduce the computational cost, different optimization algorithms were implemented. A Gaussian low-pass filter was implemented for bone segmentation, and the thresholding method was applied. Morphological operations to eliminate sparse pixels, gaps and to generate smooth edges were performed. The Marching Cubes algorithm was used to generate a three-dimensional mesh. The 3D model scale was adjusted by the metadata in the DICOM files. The 3D model surfaces were smoothed by a Gaussian filter applied to the segmented images and an averaging filter used on the model’s vertices. To verify the dimensions of the reconstructed model, they were compared with those obtained from a model reconstructed in InVesalius 3.1. Approximately, it took 1.95 min to generate the exported 3D model in a CAD-compatible STL file.