nique is called intravascular optical coherence
tomography (IVOCT). In the following sections, we
describe in detail CAD, the prior state of the art in
imaging for CAD, and the IVOCT technique.
Like many other imaging techniques, a major issue
with IVOCT is that it can produce more than 500
images in a single scan, resulting in an explosion of
image data. This can be difficult and labor intensive
to analyze manually, taking up to one hour of examination for each image by a trained analyst, of which
there are not many, given the recency of the technique (Lu et al. 2012). This often precludes measurements from every frame, and plaque classification is
not done because it is infeasible in terms of time. Further, this manual process is also prone to error. In prior work (Lu et al. 2012), our group has found evidence of up to 5 percent intra- and 6 percent
inter-rater variability among analysts looking at these
images.
The goal of our work is to enable an effective detection and diagnosis of CAD, which is a necessary precursor for effective treatment. We aim to build a tool
to do this in three ways: ( 1) reduce the effort
involved, ( 2) improve the accuracy of disease mechanism identification, and ( 3) make the diagnosis
available as early in the process as possible. The
prevalence of CAD means achieving these goals can
have a major impact on health worldwide.
We anticipate fulfilling our goals in two steps. In
the first step, reported in this article, we develop an
automated method to process a single image generated by IVOCT scans. We demonstrate that it is accurate and efficient on real IVOCT data and that analysts can use the output to greatly reduce their
annotation effort. In the second step, our goal is to
integrate this approach into a real-time visualization
that accompanies an IVOCT scan. These images will
be annotated with different detected plaque types
and will be used for rapid identification of high-risk
regions for intervention, management and guidance.
Cardiovascular Artery Disease (CAD)
In this section, we discuss CAD and the state of the
art in its diagnosis and treatment.
The underlying disease process in the blood vessels
that results in coronary heart disease (heart attack) is
known as atherosclerosis. For many years, it was
thought that the main cause of a heart attack was the
buildup of fatty plaque within an artery leading to
the heart. With time, the plaque buildup would narrow the artery so much that the artery would either
close off or become clogged by a blood clot (
stenosis). The lack of oxygen-rich blood to the heart would
then lead to a heart attack. However, these types of
blockages cause only about 3 out of 10 heart attacks
(Virmani et al. 2000).
Researchers are now finding that many people
who have heart attacks do not have arteries severely
narrowed by plaque (Falk 1983). In fact, vulnerable
plaque may be buried inside the artery wall and may
not always bulge out and block the blood flow
through the artery. This is why researchers began to
look for, potentially, a different cause. What they
found was that a thin protective fibrous cap (FC)
overlying an atherosclerotic plaque (lipid pool) may
rupture, triggering the formation of a blood clot,
which may eventually lead to an acute event such as
heart attack.
Current state-of-the-art treatment of the disease
focuses on blood vessel narrowing by means of percutaneous interventions (PCIs). PCI is a procedure
that uses a catheter (a thin flexible tube) to place a
“stent” to open up blood vessels in the heart that
have been narrowed by plaque buildup (stenosis). A
stent is a flexible tube that reinforces the blood vessel wall. This needs significant imaging support to
determine how, where, or even if it should be done.
For example, the presence of significant calcification
in the vessel may prevent the stent from being
placed or from functioning as intended, triggering
additional procedures to remove the calcium or
aborting the procedure. On the other hand, if there
is a lipid pool that may rupture, a physician can
extend a stent to seal off the affected area or at least
avoid placing the stent edge in a lipid region, an
occurrence that raises the risk of a tear or damage to
the inner wall or lining of an artery. These examples
highlight the need for a reliable imaging technique
with suitable resolution to identify plaque at high
resolution (for example, thickness of vulnerable
fibrous cap << 65 μm).
The current standard intravascular imaging modality is intravascular ultrasound (IVUS). IVUS is a medical imaging methodology that uses a catheter with a
miniaturized ultrasound probe attached to the distal
end of the catheter. The proximal end of the catheter
is attached to an ultrasound device. The IVUS
machine produces a detailed cross-sectional image of
the vessel wall and plaque as a gray level intensity
image. An example of the IVUS two-dimensional
(2D) cross-sectional image is shown in figure 1,
which shows the plaque and vessel wall from which
the ultrasound wave bounces off.
When analyzing the 2D image generated by the
IVUS machine, it is possible to quantify, limited to
the IVUS resolution, the lipid plaque and the fibrous
plaque. However, quantification of the total amount
of vessel calcification by IVUS is problematic in that
its resolution is low and it cannot measure the distance between the superficial calcium and the vessel
boundary, nor can it assess the thickness of calcium
(Mintz et al. 1995).
Intravascular Optical Coherence
Tomography (IVOCT)
In this section, we introduce IVOCT and describe its