Semi-automated generation of individual computational models of the human head and torso from MR images

Benjamin Kalloch, Jens Bode, Mikhail V. Kozlov, André Pampel, Mario Hlawitschka, Bernhard Sehm, Arno Villringer, Harald E Möller, Pierre-Louis Bazin

Research output: Contribution to journal/periodicalArticleScientificpeer-review

43 Downloads (Pure)

Abstract

PURPOSE: Simulating the interaction of the human body with electromagnetic fields is an active field of research. Individualized models are increasingly being used, as anatomical differences affect the simulation results. We introduce a processing pipeline for creating individual surface-based models of the human head and torso for application in simulation software based on unstructured grids. The pipeline is designed for easy applicability and is publicly released on figshare.

METHODS: The pipeline covers image acquisition, segmentation, generation of segmentation masks, and surface mesh generation of the single, external boundary of each structure of interest. Two gradient-echo sequences are used for image acquisition. Structures of the head and body are segmented using several atlas-based approaches. They consist of bone/skull, subarachnoid cerebrospinal fluid, gray matter, white matter, spinal cord, lungs, the sinuses of the skull, and a combined class of all other structures including skin. After minor manual preparation, segmentation images are processed to segmentation masks, which are binarized images per segmented structure free of misclassified voxels and without an internal boundary. The proposed workflow is applied to 2 healthy subjects.

RESULTS: Individual differences of the subjects are well represented. The models are proven to be suitable for simulation of the RF electromagnetic field distribution.

CONCLUSION: Image segmentation, creation of segmentation masks, and surface mesh generation are highly automated. Manual interventions remain for preparing the segmentation images prior to segmentation mask generation. The generated surfaces exhibit a single boundary per structure and are suitable inputs for simulation software.

Original languageEnglish
Pages (from-to)2090-2105
JournalMagnetic Resonance in Medicine
Volume81
DOIs
Publication statusPublished - 2019

Keywords

  • Journal Article

Fingerprint Dive into the research topics of 'Semi-automated generation of individual computational models of the human head and torso from MR images'. Together they form a unique fingerprint.

  • Cite this

    Kalloch, B., Bode, J., Kozlov, M. V., Pampel, A., Hlawitschka, M., Sehm, B., Villringer, A., Möller, H. E., & Bazin, P-L. (2019). Semi-automated generation of individual computational models of the human head and torso from MR images. Magnetic Resonance in Medicine, 81, 2090-2105. https://doi.org/10.1002/mrm.27508