IEEE Spectrum November, 2014 - 37

Making the Mock-up
Here's how to create a personalized model of the heart of a patient who has suffered a heart attack. It's
necessary to map the pattern of scar tissue and the orientation of muscle fibers, because these factors
determine how electrical signals move through the cardiac tissue to produce a regular heartbeat.

take mri scans
of the patient's heart.

Use image-processing tools
to identify the heart's scar tissue [tan]
and the semifunctional adjacent tissue [purple].

Use image-processing tools
to locate the walls of the heart's chambers.

overlay that UniqUe pattern
of scar tissue on the 3-D model.

To validate our method and get this
test to the clinic, our team is now creating individualized virtual hearts for postheart-attack patients who have an ejection
fraction greater than 35 percent (who
therefore don't qualify for an implanted
device). Clinical recommendations for
these patients are sparse, but we can run
our simulations and make our predictions. If we discern patients in that group
who are at high risk for a lethal arrhythmia, we can recommend that they receive
implants despite their relatively high ejection fractions. We're working toward a
day when cardiologists routinely order
these virtual tests as a noninvasive way
of screening their patients and gauging
their risk of sudden cardiac death.

With that data, construct a 3-D model
depicting that heart's unique anatomy.

Using another image-analysis
program, determine the orientation of
the heart's muscle fibers.

s o how d o w e m A k e A v i r t uA l
heart? To be clinically useful, our model
must represent the individual's unique
anatomy and the pattern of scar tissue
from the heart attack. We start with the
patient's magnetic resonance imaging
(MRI) or computed tomography (CT)
scans, which produce images representing slices of the heart. We use imageprocessing techniques to identify the
muscle tissue in the walls of the heart's
chambers and to map the damaged
heart's scar tissue. Then we use that
information to build a geometric model.
For the final step we use the images to
estimate the orientation of the muscle
fibers, which determines how electrical signals propagate through the tissue.

Once we have this patient-specific geometric structure, we overlay a computational model of the inner workings of a
generic heart. We need to represent activity at the cellular and molecular levels,
where ionic exchanges across cardiac-cell
membranes trigger contractions and
where currents flow from cell to cell. The
result is a personalized computer model
that can be likened to Google Earth-think
of it as "Google Heart"-which allows us
to zoom in on a disease component and
then zoom out to explore the phenomena at the organ level.
In another application of virtualheart technology, we're looking at a
treatment for arrhythmia that goes a
step beyond the implanted defibrilla-

SPECTRUM.IEEE.ORG

nORTh aMERICan

nOv 2014

http://SPECTRUM.IEEE.ORG

Table of Contents for the Digital Edition of IEEE Spectrum November, 2014