Page 154 - Artificial Intelligence for Computational Modeling of the Heart

126 Chapter 4 Data-driven reduction of cardiac models

Figure 4.5. Features describing a stenosis.

4.1.2.4 Local features
These features characterize the anatomy only at the specific
location at which the FFR prediction is performed. Herein, the
actual radius of the vessel, and the reference radius of the hypo-
thetically healthy vessel are used (if the current location is not
stenosed, the two values are identical). Additionally, a branch spe-
cific ischemic weight was defined, representing the potential con-
tribution of the branch to the overall ischemic state of the individ-
ual. Its value is initially set based on the reference radiuses of all
branches in the anatomical model, and then adapted as described
in the following.

4.1.2.5 Features defined based on the proximal and distal
vasculature
Given the significant inter-dependence of coronary hemody-
namics at different locations features describing the proximal and
distal vasculature are of significant importance. A first step in the
definition of the proximal and distal features is the identification
of all proximal and distal stenoses. Stenoses are identified auto-
matically: all narrowings with a radius reduction larger than 10%
are marked as stenoses. Next, all identified stenoses are ranked
based on the degree of radius reduction, and the most severe four
proximal and distal stenoses are retained. For each stenosis the
following anatomical characteristics and their non-linear product
combinations are computed (Fig. 4.5):
• Minimum radius
• Proximal radius
• Distal radius
• Length of the stenotic segment with minimum radius
• Total stenotic length