Page 24 - Artificial Intelligence for Computational Modeling of the Heart

Preface xxv

patterns could be extracted from the data, when the output of
physiological modeling is considered as additional information.
This could finally promote the use of digital twins of human
organs and systems where they are most needed: at the best
side.

Computational modeling of the heart: from

physiology understanding to patient-specific
simulations
Finding mathematical laws that explain physiological systems
has been of scientific interest for centuries. Leonardo Da Vinci’s
Vitruvian man provides an outstanding example of the quest for
mathematical laws of human proportions and beauty. However,
it is commonly acknowledged that mathematical models of car-
diac cells and tissue were first proposed in the 1950’s, as the un-
derstanding of biology increased in the scientific community. In
particular, Hodgin and Huxley’s work has been seminal in deriving
what is known to be the first model of the cellular action poten-
tial [12]. Since then, tremendous progress has been made in mod-
eling heart biophysiology, from the cardiac action potential [13],
sarcomeres function [14], excitation-contraction coupling [15],
metabolism [16], blood perfusion and then going up to the tissue
level, organ level and finally circulatory and human body levels [7].
Multiple models have been proposed and made available in public
repositories (e.g. CellML [17]), with varying degrees of complexity
and fidelity with respect to wet lab experiments. However, it still
remains an art to identify which model to apply, what simplifi-
cations to make and which parameters to adjust to simulate the
biophysics under investigation.
Supported by the growing computational power along with in-
creasing amount of clinical data, researchers have started to in-
vestigate methods to perform patient-specific simulations [5,6,
18]. In this endeavor, a new set of challenges arises. How detailed
should the model be to capture the disease of the patient? What
data is available in a clinical setup to estimate a useful patient-
specific model? If some parameters are not clinically measurable,
what should their value be? What is a good validation strategy?
How should these models be integrated into the clinical work-
flow? These are only a small sample of the questions that need
to be answered for a clinically useful virtual heart. Recently pub-
lished works are suggesting promising results, like for instance
non-invasive characterization of cardiac function in heart fail-
ure [6], planning and guidance of ablation therapy for cardiac ar-