ABSTRACT
The microcirculation plays a critical role in organ homeostasis and in disease pathogenesis. Much effort has
been dedicated to developing methods to image the microcirculation, however developing quantitative methods
to assess organ-specific microcirculation remains an ongoing challenge. Identifying microvascular phenotypes
from existing imaging modalities would help overcome these limitations. Most vascular imaging studies focus on
larger vessels (> 1mm) due to limited instrument resolution. However, these studies often collect time-course
data containing dynamic information that reflects blood flow. Since the microcirculation is primarily responsible
for regulating flow, blood flow data reflects microvascular function when there is no proximal stenosis. Thus, we
can use time-course dynamic data from imaging studies to identify microvascular phenotypes without directly
imaging the micro-vessels. Our central hypothesis is that the time course of contrast material in blood vessels
and the dynamics of contrast material in tissue regions contain intravascular and tissue parameters, respectively,
which reflect the status of the microcirculation. We propose to develop robust image analysis techniques to
discover image-based microvascular phenotypes. We will initially focus on the coronary microcirculation, given
the broad public health implications of Ischemic Heart Disease. In Aim 1, we will develop, test, and validate (a)
a recently-developed Hybrid Intelligence (HI) approach to segment major vessel segments and myocardial tissue
regions in clinical coronary angiograms, and (b) methods to estimate parameters of blood flow in segmented
vessels and perfusion in segmented tissue regions. In Aim 2, we will determine the optimal imaging biomarkers
for coronary microvascular function using two leading methods currently used to quantify coronary
microvasculature. First, we will compare vessel-specific parameters and tissue-based parameters to global and
regional myocardial blood flow as measured by Rubidium-82 perfusion cardiac PET. Then, we will compare our
parameters against TIMI frame count measurements, an established yet laborious method to quantify coronary
flow on coronary angiograms. These studies will develop a novel imaging technology to establish coronary-
angiogram based microvascular phenotypes and biomarkers. These methods are also applicable to additional
angiography datasets (2D projection x time) including cerebral, renal, pulmonary, and peripheral vascular
angiograms, and could be extended to 4D datasets (3D imaging x time) as seen in perfusion computed
tomography and magnetic resonance imaging studies. They would therefore allow for assessment of organ-
specific microcirculation from existing imaging studies and allow for microvascular phenotyping to greatly
improve clinical care and accelerate research in this urgently needed area.