PO-05-229 DEEP LEARNING CLASSIFICATION OF PAROXYSMAL SUPRAVENTRICULAR TACHYCARDIA TYPES USING SINUS RHYTHM ELECTROCARDIOGRAMS

Only sinus rhythm electrocardiograms (ECGs) are often available in patients with paroxysmal supraventricular tachycardia (PSVT). However, ECGs have a limited value classifying PSVT types. We aimed to investigate the deep learning approach of Atrioventricular Nodal Reentry Tachycardia (AVNRT) from concealed (AVRT) using ECGs. collected AVNRT or conceald AVRT, and their 12-lead rhythm. The diagnosis each patient was validated an electrophysiology study. ResNet-34 used for model, it trained classify underlying AVRT. Due training dataset, transfer performed by pre-training open-source ECG database (PhysioNet/CinC Challenge 2021). Ten-fold cross-validation validate performance. Integrated gradient analysis visualize which segment is important model two arrhythmias. A total 696 305 AVRT were analyzed. Compared group, group significantly younger had higher male proportion; age 48 vs. 51 years (p<0.001); 63.3 34.3% (p<0.001). After pre-training, performance improved; AUROC 0.73 (95% CI 0.69-0.76) 0.67 (0.62-0.72) without pre-training; p<0.001 (Figure A). Deep showed modest arrhythmias; AUPRC, F1-score, sensitivity, specificity 0.84 (0.81-0.87), 0.75 (0.70-0.80), 0.68 (0.61-0.75), 0.71 (0.65-0.77), respectively. focused on PR intervals T waves identifying AVNRT, P-waves B). classification feasible