Exploring Transfer Learning for Ventricular Tachycardia Electrophysiology Studies

Arrhythmogenic sites in post-ischemic ventricular tachycardia (VT) are usually identified by looking for abnormal ventricular potentials (AVPs) in intracardiac electrograms (EGMs). Unfortunately, the accurate recognition of AVPs is a challenging problem for different reasons, including the intrinsic variability in the A VP waveform. Given the high performance of deep neural networks in several scenarios, in this work, we explored the use of transfer learning (TL) for AVPs detection in intracardiac electrophysiology. A balanced set of 1504 bipolar intracardiac EGMs was collected from nine post-ischemic VT patients. The time-frequency representation was generated for each EGM by computing the synchrosqueezed wavelet transform to be used in the re-training of the convolutional neural network. The proposed approach allows obtaining high recognition results, above 90% for all the investigated performance indexes, demonstrating the effectiveness of deep learning in the recognition of AVPs in post-ischemic VT EGMs and paving the way for its use in supporting clinicians in targeting arrhythmogenic sites. In addition, this study further confirms the efficacy of the TL approach even in case of limited dataset sizes.