Premium
Incorporating Human Ventricular Fiber Architecture in Patient‐Specific Computational Models
Author(s) -
Villongco Christopher T,
Frank Lawrence R,
Krummen David E,
Nigam Vishal,
Kerckhoffs Roy CP,
Omens Jeffrey H,
McCulloch Andrew D
Publication year - 2012
Publication title -
the faseb journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.709
H-Index - 277
eISSN - 1530-6860
pISSN - 0892-6638
DOI - 10.1096/fasebj.26.1_supplement.864.19
Subject(s) - diffusion mri , voxel , tensor (intrinsic definition) , image warping , diffeomorphism , coordinate system , fractional anisotropy , magnetic resonance imaging , artificial intelligence , geometry , computer science , mathematics , mathematical analysis , medicine , radiology
Patient‐specific models of the cardiac ventricles require a realistic representation of the myofiber architecture to accurately simulate mechanical and electrical properties of healthy and failing hearts. Diffusion tensor magnetic resonance imaging (DT‐MRI) techniques have made it possible to reconstruct the fiber anatomy of the ventricles. However, in vivo DT‐MRI scans are still impractical to obtain for individual patients. The aim of this study develops a method to register DT‐MRI measurements from an explanted human heart into an anatomical model and warping the result to patient‐specific anatomies using large deformation diffeomorphic mapping. An explanted human organ donor heart was subjected to a DT‐MRI scan. The raw tensor data was registered into an anatomical model of the ventricles by fitting the measured diffusion tensor components in log‐Euclidean space. A material coordinate transformation model defined a local orthogonal coordinate system representing the fiber, sheet, and sheet‐normal directions of the myocardium from the eigenvectors of a locally interpolated diffusion tensor. Diffeomorphic maps of the deformation F between the donor and patient models were computed and used to warp the donor fiber architecture into the patient models. The registration error between the eigenvectors of the measured tensor and its interpolated tensor was quantified by a scalar metric of axes overlap. It found 74% of the registered DT‐MRI data (112,419 of 152,651 voxels) had mean angular errors of 12.5°±7° (SD) in the fiber, 20°±11.5° (SD) in the sheet, and 18°±11.6° (SD) in the sheet normal axes.