An Image Registration-Based Morphing Technique for Generating Subject-Specific Brain Finite Element Models

• Published in: Annals of biomedical engineering Vol. 48; no. 10; pp. 2412 - 2424
• Main Authors: Giudice, J. Sebastian, Alshareef, Ahmed, Wu, Taotao, Gancayco, Christina A., Reynier, Kristen A., Tustison, Nicholas J., Druzgal, T. Jason, Panzer, Matthew B.
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.10.2020
• Subjects: Adult; Aged; Aged, 80 and over; Algorithms; Biochemistry; Biological and Medical Physics; Biomechanical Phenomena; Biomedical and Life Sciences; Biomedical Engineering and Bioengineering; Biomedicine; Biophysics; Brain - diagnostic imaging; Brain Concussion - diagnostic imaging; Classical Mechanics; Female; Finite Element Analysis; Football - injuries; Humans; Magnetic Resonance Imaging; Male; Middle Aged; Models, Anatomic; Original Article; Patient-Specific Modeling; Young Adult; Magnetic resonance imaging (MRI); Computational mechanics; Traumatic brain injury (TBI); Personalized medicine
• ISSN: 0090-6964; 1573-9686; 1573-9686
• DOI: 10.1007/s10439-020-02584-z

Finite element (FE) models of the brain are crucial for investigating the mechanisms of traumatic brain injury (TBI). However, FE brain models are often limited to a single neuroanatomy because the manual development of subject-specific models is time consuming. The objective of this study was to develop a pipeline to automatically generate subject-specific FE brain models using previously developed nonlinear image registration techniques, preserving both external and internal neuroanatomical characteristics. To verify the morphing-induced mesh distortions did not influence the brain deformation response, strain distributions predicted using the morphed model were compared to those from manually created voxel models of the same subject. Morphed and voxel models were generated for 44 subjects ranging in age, and simulated using head kinematics from a football concussion case. For each subject, brain strain distributions predicted by each model type were consistent, and differences in strain prediction was less than 4% between model type. This automated technique, taking approximately 2 h to generate a subject-specific model, will facilitate interdisciplinary research between the biomechanics and neuroimaging fields and could enable future use of biomechanical models in the clinical setting as a tool for improving diagnosis.