PAMM · Proc. Appl. Math. Mech. 17, 205 – 206 (2017) / DOI 10.1002/pamm.201710072
Development of patient-speciﬁc computational models of the human
aortic valve under healthy and diseased conditions
Sergio Morales Ortuno
and Oliver Röehrle
Stuttgart Research Centre for Simulation Technology (SimTech), University of Stuttgart, 70569 Stuttgart, Germany
Institute of Applied Mechanics (CE), University of Stuttgart, 70569 Stuttgart, Germany
A 3-D ﬁnite element modelling (FEM) approach of the human aortic valve (AV) is developed to study its extra-cellular
matrix (ECM) destructive remodelling caused presumably by ventricular diseases, assuming that the function, deformation
and performance of the valve is strongly dependent on its geometry.
2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
The main goal of the present work is to investigate the apparent link between hearts presenting particular cardiomyopathies
and an ECM remodelling  resulting along the aortic leaﬂets from the diseased hearts. To achieve this goal, a FEM of the
aortic leaﬂets is developed, in which, a patient-speciﬁc geometry obtained by non-invasive geometry extraction methods is
used for such model. A series of valvular geometries is extracted from medical imaging data (MRI and 3D-TEE) to replicate
the patient-speciﬁc human AV’s geometry. The geometry encompasses the asymmetry and regional thicknesses of the leaﬂets,
both inﬂuencing the bio-mechanical behaviour of the valve. Additionally, case-speciﬁc boundary conditions are applied to the
models to reproduce the bending and tensile stretches suffered by the cusps along the heart cycle. The overall aim is to use the
model as a tool to predict the kinematics of the leaﬂets of healthy and diseased AVs and compare it with the dynamics of the
valves observed in medical imaging data, with the purpose of generating useful and sufﬁcient data to study the development
of calciﬁc AV disease from initial alterations in the tissue to end-stage calciﬁcation.
2 Geometry acquisition
Traditional image processing ﬁlters cannot satisfactorily identify the edge regions on MRI and Ultrasound (US) imaging.
These edges are a result of abrupt discontinuities of the brightness on the image, which for this case, correspond to the areas
where blood and tissue meet. Edge localization of the leaﬂets is the ﬁrst step employed in this work for the patient-speciﬁc
model generation. Intensity and local phase-based methods for AV boundary detection are investigated in this work and
detailed in the following.
2.1 Intensity based and intensity invariant approaches
This approach ﬁrst uses of a pre-processing step which highlights the possible valvular tissue while minimizing the noise and
background effect. This step is a so called Frangi ﬁltering . The idea is to extract the principal directions in which the local
second order (Hessian) structure of the image (I) can be decomposed. The enhancement of the possible leaﬂet structures on
the 2-D input image is then obtained by combining the results from two geometric ratios from the second order structure of the
image. The ﬁrst one accounts for the bob-like
structures, while the second one quantiﬁes the magnitude of the derivatives of
the pixels with the purpose of suppressing the background areas where its value it is small due to the lack of contrast. Once the
structures on the region of interested are highlighted, it is then when an edge detection method is applied since better results
are expected. The following step or line-like leaﬂet geometry recognition is based on the work by Schneider et al. , adapted
to aortic leaﬂet delineation of our speciﬁc data. This is done by applying a Thin Tissue Detector (TTD), that makes use of
the characteristics of the previously highlighted input images’ (Ω) gradient ﬁeld (∇(·)) and a convolved version of it (Ω
The TTD value is at the end computed by combining three values, which represent the average (Θ) of the angles between
gradient vectors in a neighbourhood, the ﬂux (Φ) of these vectors across the neighbourhood boundary faces and the number
and proximity (Π) of strong edges to a pixel.
An alternative approach to the intensity-based approach is employed to delineate the leaﬂet geometry from the US imaging
data. This method offers better performance than the previously introduced approach due to the characteristics of the low
contrast between boundaries present in the 2D-US images in hand. A method which makes use of the local phase features of
the image is implemented. Edges of the valvular tissue are then detected by using the feature assymetry  measure, which
makes use of the difference between an even and odd ﬁltered responses of the input image .
Corresponding author: e-mail email@example.com, phone + 49 711 685 65837 fax + 49 711 685 66347
In computer vision a Blob refers to a region of the image in which the properties of its points are similar and/or constant.
2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim