Work

Cross-correlation as an alternative for Local Activation Times for the analysis of reentries in Directed Graph Mapping

Year: 2025

Type: article

Abstract: The detection of reentries is crucial for the treatment of many arrhythmias . We have previously introduced a tool for their detection – Directed Graph Mapping (DGM), a method that uses local activation time (LAT) maps to build directed networks that follow the direction of electrical propagation. DGM has been shown to be capable of finding arrhythmic drivers in clinical and simulated atrial tachycardia (AT). However, LATs are less reliable for noisy or fractionated data, such as in atrial fibrillation (AF). Here, we show that by replacing the LAT-based method – DGM-LAT – with a pipeline based on signal cross-correlation – DGM-signal – this issue can be greatly mitigated. We compare the methods across 379 AT simulations and 21 clinical AT cases, also testing varying signal-to-noise ratios (SNRs) in the simulated data (white Gaussian noise, SNR = 10 dB, 5 dB, 0 dB, −5 dB) to assess robustness. Performance, measured as accuracy, sensitivity, specificity, and precision was evaluated by comparing network cycles with ground-truth reentry locations. Both DGM-LAT and DGM-signal reached 100% accuracy on clinical and noiseless simulated data. As noise increased in simulated data, DGM-signal consistently had comparable specificity and superior accuracy and sensitivity for every noise level . In particular, at SNR = −5 dB, DGM-LAT no longer found any reentries , while DGM-signal still found 70% of them. Our findings show that cross-correlation greatly improves the robustness of DGM, and suggest that it may be a helpful tool to identify reentry circuits in noisy clinical settings, aiding in catheter ablation procedures .

Source: Biomedical Signal Processing and Control

DOI: 10.1016/j.bspc.2025.107716