Detection of Coronary Artery Disease by non-invasive Cardisiography – Results of the prospective monocentric CARDIAC – pilot study

Background:
Methods for detecting coronary artery disease (CAD) and differentiating extracardiac chest pain are often time-consuming, costly and have limited availability. Therefore, the development and improvement of non-invasive and cost-effective methods is of particular interest. This study aimed to assess the diagnostic efficacy of non-invasive cardisiography (CSG) in detecting significant CAD against the reference of cardiac magnetic resonance (CMR) stress testing.

Methods:
In a single-center, prospective diagnostic pilot study, 141 patients with planned assessment of myocardial ischemia by stress CMR were enrolled between June 2021 and February 2022. Patients with acute coronary syndrome, pacemaker or ICD-implantation, or cardiac surgery were excluded. All patients with suspected relevant CAD underwent CSG. Myocardial ischaemia was then assessed by stress CMR. A positive stress test was defined by a perfusion deficit of ≥ 12% myocardium or wall motion abnormalities in ≥ 2/17 segments. The primary endpoint was diagnostic accuracy, defined as the combined endpoint of sensitivity and specificity of CSG as a screening test in detecting relevant myocardial ischaemia compared with CMR.

Results:
141 patients (mean age 62.3 ± 12.2 years) were included in the study. 69.5% of the patients were male, 71.4% had arterial hypertension, 20.3% had type 2 diabetes mellitus, 43.5% had a positive history of nicotine use, and a large proportion displayed other risk factors such as overweight (mean BMI 27.4 kg/m2 ± 4.6) or dyslipidaemia (62.6%). 61.0% of the patients had known CAD. 71.2% had normal LV function (LV-EF > 54%). Prevalent late gadolinium enhancement on CMR was found in 71.9%, with 34,0% of all patients exhibiting a previous myocardial infarction. Stress CMR was positive for myocardial ischemia in 16 (11.0%) patients. CSG sensitivity was 0.44 (CI 0.19 - 0.70) and specificity 0.54 (CI 0.44 - 0.65). Positive and negative predictive values of CSG were 0.10 (CI 0.04 - 0.20) and 0.88 (CI 0.78 - 0.94), respectively. 82.7% of the 58 false CSG-positive patients had structural myocardial changes (e.g. dilated left ventricle in 19.9%, left ventricular hypertrophy in 48.9%), 32.7% had a non-significant perfusion deficit in CMR. Evaluation of an updated ML algorithm (n = 113, 28 patient measurements could not be analysed) showed similar performance with sensitivity and specificity of 0.50 (CI 0.23 - 0.76) and 0.53 (CI 0.43 - 0.63) respectively.

Discussion:
In the high-risk CAD population studied, the findings indicate that current CSG algorithms may not be efficient for differentiating between a significant myocardial ischemia caused by CAD, subclinical ischemia (< 12% of myocardium) and structural changes (e.g. LV-dilatation/-hypertrophy) that are possibly causing cellular hypoperfusion below current CMR cut-off values for clinically relevant ischemia. A possible reason for the limited differentiation among different types of heart diseases may be the fact that the ML algorithm has not been trained with non-invasive imaging correlated data at this point in time. Furthermore, coronary angiography without functional stenosis assessment was used for initial algorithm training whereas this study is the first one to assess patients with CMR. Further studies are needed to investigate whether adaptation and training of the algorithm with stress test data, including functional stenosis assessment, and structural information could improve diagnostic accuracy