1MEDIACC GmbH Berlin, Deutschland; 2Cardisio GmbH Frankfurt am Main, Deutschland; 3Internal Medicine Practice Berlin, Deutschland
Introduction: Artificial Intelligence-based 5-lead 3D-vectorcardiography (5L3DVCG-AI) offers additional information over 12-lead electrocardiography (ECG) identifying coronary vascular disease (CVD) in need for coronary intervention. So far, detection of cardiac pathologies (CVD, left ventricular hypertrophy, arrhythmias, conductance disturbances, HFpEF, HFrEF, valvular disease) has not been tested with this technology.
Hypothesis: We, thus, tested the hypothesis of the CSG-Index from 5L3DVCG-AI being able to detect mild to overt signs and / or history of any cardiac pathology.
Methods: In this monocentric retrospective study, consecutive data of 299 patients with 5L3DVCG-AI for detection of cardiac pathologies were included. The CSG-Index (cut-off -0.27) including 731 parameters and in-house features calculated in time and frequency domains (e.g. beat moments), classified patients as low or high CVD risk. Diagnosis of any cardiac pathology was based upon cardiac diagnosis according to current guidelines by 2 independent blinded cardiologists and categorised as exclusion of cardiac pathology (control), subclinical or overt cardiac pathology. Cardiovascular risk factors (CVRF) were quantified with the modified PROCAM-Score.
Results: Of 299 patients (m:w 181:118, 58.7 ± 16.7 years) of mixed ethnicity and moderate CVRF (2.3 ± 1.3), 72% were controls, 22% had subclinical and 6% overt cardiac pathology. Follow-up period was 16.2 ± 7.5 months. CVRF were significantly higher in cardiac pathologies compared to controls (2.8 ± 1.2 vs. 2.0 ± 1.2, p<0.05) whereas CSG-Index not only differentiated between controls (Co) and cardiac pathology, but also between subclinical (SCP) and overt cardiac pathology (OCP, ANOVA p<0.001, Co vs. SCP, p<0.01, and SCP vs. OCP p<0.01) with similar results for the female subpopulation (ANOVA p<0.01). CSG-Index from 5L3DVCG-AI was a better predictor for cardiac pathologies than CVRF score (ß=0.26, T=4.7, p<0.001) than CVRF-Score. ECG at rest was not able to differentiate between CVD and controls.
Conclusions: 5L3DVCG-AI, an inexpensive and easy-to-use technique may have a role in screening and identifying cardiac pathologies in need for further cardiac diagnostics, especially women, and in need for risk modification. Ongoing prospective large-scale clinical trials will have to confirm these data to verify diagnostic accuracy.