Artificial Intelligence-assisted Diagnostics in Angina with No Obstructive Coronary Artery Disease

Ischemic heart disease (IHD) affects 126 million people and remains the number
one cause of death and disability worldwide. Angina pectoris is an umbrella
term describing symptoms of chest pain, chest tightness, or dyspnea related to
IHD. Angina pectoris is diagnosed in 180.000 people in the Netherlands yearly.
Diagnostics in angina pectoris remain focused on detecting obstructive
epicardial coronary artery disease (CAD), which may lead to evidence-based
treatment including medical therapy and coronary revascularization. However,
among patients undergoing coronary angiography for angina pectoris, 40-70% have
no CAD.(1) This is more common in women than in men and is associated with poor
quality of life, increased risk of cardiovascular events, and high medical
expenses due to ongoing symptoms, repeat (invasive) investigations and hospital
admissions.(2-5) Although the reasons for angina pectoris without epicardial
CAD are multifactorial, many of these patients have a disorder of coronary
artery vasomotor function.(6) This term relates to the occurrence of vasospasm
in the epicardial coronary artery or the microvasculature, to microvascular
vasodilator dysfunction, or to a combination of these disorders. These
conditions can be diagnosed during invasive coronary angiography by applying
additional testing, referred to as coronary function testing (CFT). CFT
involves the administration of acetylcholine to study endothelium-dependent
vasodilatation and the susceptibility to coronary vasospasm, and the
administration of adenosine to study endothelium-independent vasodilatation.
The response to acetylcholine administration in terms of reproduction of
anginal symptoms, electrocardiographic changes related to myocardial ischemia,
and angiographic narrowing of the coronary arteries determines the likelihood
of coronary vasospasm. Coronary blood flow responses to adenosine
administration define the vasodilator capacity of the coronary circulation as
the most important diagnostic marker of coronary microvascular dysfunction.
Together, these results inform on the specific coronary vasomotor disorder in
the individual patient. When performed routinely in patients with angina and no
obstructive coronary artery disease (ANOCA), a vasomotor disorder is identified
in up to 85% of patients.(6) Of these, the majority consists of forms of
coronary vasospasm (80%), while only a few percent of patients have coronary
microvascular dysfunction as the final diagnosis.(7) This is crucial, since
coronary microvascular dysfunction can be diagnosed using non-invasive
techniques, but invasive CFT is required to diagnose coronary vasospasm. Hence,
invasive diagnostics are currently a must to adequately diagnose patients with
ANOCA.

However, particularly in the diagnosis of coronary vasospasm, several caveats
exist. The diagnosis of coronary vasospasm is based on international diagnostic
criteria.(8) For epicardial coronary spasm, provocation should induce
epicardial coronary artery constriction of >90% in lumen diameter compared to
the maximally vasodilated state, in combination with recognizable angina, and
ischemic changes in the 12-lead electrocardiogram (ECG). For microvascular
spasm, provocation should induce recognizable angina and ischemic changes in
the 12-lead ECG. When patients do not fulfil these diagnostic criteria, the
test is considered equivocal (1-2 positive criteria) or negative (0 positive
criteria). An equivocal test result occurs in 25% of patients undergoing
coronary function testing. The interpretation of the ECG is the most important
caveat in the diagnosis of vasospasm because the sensitivity for myocardial
ischemia of the standard 12-lead surface ECG is suboptimal in this patient
population and the interpretation of subtle beat-to-beat ECG changes during
acetylcholine administration is inherently difficult. These caveats of the ECG
lead to a risk of misdiagnosis in this complex patient population. Besides the
resulting uncertainty, physicians may not initiate medical therapy or are
reluctant to intensify medical therapy in absence of a definitive diagnosis,
leading to ongoing symptoms, repeat investigations and hospital admissions.
Improving diagnostic certainty in coronary function testing is therefore an
important goal to improve care for ANOCA patients.

A readily available technique could improve detection of myocardial ischemia
during coronary function testing. During CFT, a coronary guide wire is
routinely advanced in the coronary artery which also allows obtaining an
intracoronary ECG by attaching a sterile alligator clamp to a standard
electrocardiogram lead.(9) This allows continuous recording of an intracoronary
ECG throughout CFT on the same monitor as the 12-lead surface ECG. Pilot data
indeed support that this technique can identify electrocardiographic changes
related to myocardial ischemia during coronary function testing that are not
apparent on the standard 12-lead surface ECG and can thereby increase
sensitivity for myocardial ischemia during CFT.

Beyond improving invasive diagnosis of coronary vasospasm, the potential for
non-invasive diagnostics deserves particular attention to reduce the risks
associated with invasive coronary function testing and facilitate early
diagnosis in this complex and rapidly expanding patient population. The
diagnostic criteria for coronary vasospasm mainly drive on the association of
reproducible anginal complaints and the simultaneous occurrence of ischaemic
changes in the electrocardiogram, which together make a diagnosis of coronary
vasospasm. Simultaneous invasive documentation of the presence or absence of
epicardial vasospasm further allows to distinguish epicardial and microvascular
vasospasm. The latter, however, currently bears little to no consequences both
in terms of subsequent medical management, and in terms of prognosis. Hence,
the non-invasive documentation of ischemic electrocardiographic changes in
relation to spontaneous reproducible anginal complaints would allow to diagnose
the presence of the vasospastic form of coronary vasomotor dysfunction. Since
spontaneous episodes of anginal complaints are frequent in ANOCA patients, and
tools for ambulatory electrocardiographic monitoring as well as its analysis
have improved tremendously over time, we hypothesize that a large proportion of
patients could be adequately diagnosed using non-invasive ambulatory
electrocardiographic monitoring.

Both for invasive and non-invasive diagnostics, the accuracy and feasibility of
electrocardiographic data play a crucial role. Electrocardiographic data
represent an area where machine learning techniques have progressively shown
large potential for improving feasibility and accuracy of spot recordings,
beat-to-beat data, and large longitudinal data registrations. If properly
built, machine learning algorithms have a great potential in facilitating
diagnosis both during CFT by allowing beat-to-beat quantification of ischemic
ECG changes on both 12-lead ECG and intracoronary ECG registrations, as well as
for ambulatory electrocardiographic registrations, where their use facilitates
identification of potentially subtle ECG changes during prolonged recordings.

The goal of the (ai)NOCA study is to address these issues by aiming to optimize
invasive diagnosis using intracoronary ECG registrations to enhance diagnostic
efficacy for identification inducible myocardial ischemia during acetylcholine
provocation testing, as well as by developing (machine learning) algorithms to
enhance diagnostic efficacy of outpatient Holter ECG recordings to allow
diagnosis of coronary vasospasm in the outpatient setting. Together, these
goals lead to optimization of the diagnostic pathway for patients with ANOCA.

1.1 Primary Objectives
1. To document the prevalence of ischemic intracoronary electrocardiogram
changes in ANOCA patients with equivocal acetylcholine provocation test results.
2. To document the diagnostic efficacy of the 12-lead Holter ECG to identify
ischemic ECG changes during acetylcholine-provoked chest pain.
3. To document the diagnostic efficacy of outpatient 12-lead Holter ECG
monitoring to identify patients with a positive acetylcholine provocation test.

1.2 Secondary Objectives
1. To develop an algorithm to optimize identification of ischemic ECG changes
on outpatient 12-lead Holter ECG monitoring.
2. To develop an algorithm that allows beat-to-beat detection of ischemic ECG
changes on the intracoronary ECG during coronary function testing.
3. To develop an algorithm that allows beat-to-beat detection of ischemic ECG
changes on the 12-lead surface electrocardiogram.
4. To develop an algorithm for quantification of acetylcholine-induced ischemic
ECG changes on the intracoronary and surface electrocardiogram.

Observational study with non-invasive and invasive measurements.

There is no additional burden for the patients of obtaining an intracoronary
electrocardiogram. Risks related to the electrophysiology part of this study
are negligible. There is no direct benefit of participation in the study, but
the study can only be performed in this patient population to advance care of
future patients with the same clinical condition.