Similarity Prediction of Intracardiac Electrograms Images Using Regression Model Based on Siamese Network Architecture

Pace mapping allows accurate localizing of the source of cardiac arrhythmia when the arrhythmia is non-inducible during electrophysiological study. It compares the electrocardiographic (ECG) morphology of clinical ventricular tachycardia (VT) with the paced one. Often intracardiac electrograms (EGMs) from implantable cardiac devices are the only source of ECG morphology. However, when presented as images by programmer devices, EGM can only be assessed for similarity by subjective eyeballing. This study aimed to develop regression models for objectively measuring similarities between images of two distinct signals. Six regression models based on Siamese neural network architecture with two types of custom similarity layers were trained on 40,000 pairs of augmented images from 17,421 digital intracardiac EGMs. The best performance showed the model with Efficientnet_b0 as the backbone for Manhattan similarity
$(MSE=0,002,\ R2=0.98)$
and Resnet18 as the backbone for Pearson correlation
$(MSE=0,009,\ R2=0.973)$
. Regression models based on the Siamese network architecture with custom similarity layers provide a promising tool for objectively measuring the similarity of EGM images acquired from implanted devices.