SUBJECT: Ph.D. Dissertation Defense
BY: Minliang Liu
TIME: Monday, November 2, 2020, 10:00 a.m.
PLACE:, Online
TITLE: Identification of in vivo Material Properties of Ascending Thoracic Aortic Aneurysm: towards Non-invasive Risk Assessment
COMMITTEE: Dr. Wei Sun, Co-Chair (BME)
Dr. Jerry Qi, Co-Chair (ME)
Dr. Bradley Leshnower (Emory School of Medicine)
Dr. Rudolph Gleason (ME)
Dr. John Oshinski (BME)


Advances in imaging techniques and numerical methods have made it possible to investigate biomechanics of the cardiovascular system on a patient-specific level. For the four key components in a in vivo patient-specific biomechanical analysis (geometries, loading and boundary conditions, and material hyperelastic properties and material failure properties), patient-specific geometries and physiological loading conditions can be obtained at a high level of spatial and temporal resolutions from clinical diagnostic imaging tools, such as CT scans, and blood pressure measurements, respectively. However, accurate identification of the unknown in vivo patient-specific hyperelastic properties, which are nonlinear and anisotropic, has been a challenging problem in the field of cardiovascular biomechanics for several decades. Furthermore, since patient-specific failure properties cannot be obtained noninvasively from clinical images, an accurate failure metric that incorporates uncertainties of failure properties, needs to be developed for patient-specific biomechanical assessment. The objective of this thesis was to develop a novel computational framework to identify in vivo patient-specific hyperleastic properties for biomechanical risk assessment of ascending thoracic aortic aneurysm (ATAA). A novel inverse method was developed for in vivo hyperleastic property identification from clinical 3D CT image data. The developed inverse approach was validated by using numerical examples as well as clinical CT images and matching tissue samples. To describe the shape probability distribution, statistical shape model (SSM) was built from ATAA geometries. A machine learning (ML) approach was investigated for fast in vivo material property identification (i.e., within seconds), virtual geometries sampled from the SSM were used to train and test the ML-model. To assess ATAA risk, a novel probabilistic and anisotropic failure metric was derived by using uniaxial failure testing data. To evaluate the performance of risk assessment methods (e.g., with and without patient-specific hyperelastic properties), ATAA risks were numerically-reconstructed by using additional patient data. The results highlighted the potentially important roles of patient-specific hyperelastic properties and probabilistic failure metric.