1Universitätsklinikum Würzburg Med. Klinik und Poliklinik I, Klinische Elektrophysiologie Würzburg, Deutschland; 2University Clinic Wuerzburg Infection Control and Antimicrobial Stewardship Unit Würzburg, Deutschland; 3Universitätsklinikum Würzburg Deutsches Zentrum für Herzinsuffizienz Würzburg, Deutschland; 4Universitätsklinikum Würzburg Medizinische Klinik und Poliklinik I Würzburg, Deutschland; 5Universitätsklinikum Würzburg Medizinische Klinik I, Kardiologie Würzburg, Deutschland
Introduction: Atrial cardiomyopathy (AC) establishes links between atrial fibrillation (AF), left atrial (LA) mechanical dysfunction, structural remodeling and thromboembolic events. Early diagnosis of AC may impact AF treatment and stroke risk prevention. To date, however, no defined echocardiography-derived marker exists for the diagnosis of AC. Modern endocardial contact-mapping provides high-resolution electroanatomical (EA) maps of the LA thus allowing to display the myocardial substrate based on impaired signal amplitude and to characterize AC. Correlation of invasively assessed AC using a novel, multipolar mapping catheter (Octaray®, Biosense Webster, limited market release) and LA echo parameters could form the basis of an echo-parameter set for noninvasive prediction of AC.
Aims: Definition of an echocardiography-derived parameter set for non-invasive prediction of i) invasively defined AC and ii) AF recurrence after AF ablation.
Methods: We identified n=50 patients (pts.) fulfilling the selection criteria who underwent primary pulmonary vein isolation (PVI) for paroxysmal or persistent AF between 08/22 and 05/23. Inclusion criteria were: i) EA mapping using a novel multipolar mapping catheter (Octaray®); ii) acquisition of voltage maps in sinus rhythm (SR) with ≥5000 points/map; iii) transthoracic echocardiography acquired in SR ≤48 hours before PVI. Exclusion criterion was previous LA ablation. We generated EA maps with voltage thresholds of 0.2–1.0 mV and assessed total LA low voltage area (LVA) in every patient. LVA thresholds for the classification of AC are not yet established. Based on the assumption of a healthier and a sicker subgroup in the study cohort, cluster analysis was performed using the Gaussian Mixture Model (GMM). A support vector machine (SVM) was trained to determine the predictive capacity of selected parameters. Echo feature selection was based on the Boruta algorithm. Ongoing analysis comprises i) semi-automated investigation of voltage distribution as a novel measure of AC and ii) non-invasive prediction of AF-recurrence following PVI.
Results: Mean age of the studied sample was 63±11 years, 62% were men, 64% showed persistent AF, median CHA2DS2-VASc score was 2 (1,3) and NTproBNP was 292 (±309) pg/ml. A median of 5771 (5217,6988) points/map were acquired. GMM yielded clusters of mild and severe AC. Mean LVA was 4.6 cm2 in group mild AC and 29.6 cm2 in group severe AC. Echo parameters differed between groups mild AC vs severe AC: LA reservoir strain (LArS) R 2D: 24.5% (22,29) vs 15% (12,19), p<0.001; LArS R 3D: 17.5% (15,21) vs 12.5% (8,15), p<0.01; LArS P 2D: 22% (19,25) vs 15% (11,18), p<0.001; LA conduit strain R 2D: 13% (8,15) vs 7.5% (3,13), p<0.01; LA volume index (LAVI)/a’ 3D: 297 (231,365) vs 510 (326-781), p<0.01. Consistent distribution of NTproBNP (mild AC: 125 pg/ml, severe AC: 408 pg/ml, p<0.0001) and CHA2DS2-VASc (mild AC: 1, severe AC: 3, p<0.0001) served as proof of concept. Applying selected echo parameters, the SVM achieved correct group alignment with a mean AUC of 0.9.
Conclusion: LA echo parameters differed significantly between groups with small vs large areas of LVA. Machine learning supported algorithms (SVM) allowed non-invasive prediction of subgroups. Thus, the dataset could form the basis for non-invasive diagnosis and characterization of AC, thereby potentially improving i) individualized outcome prediction of AF-ablation and ii) individualized assessment of stroke risks.