A novel method to correct repolarization time estimation from unipolar electrograms distorted by standard filtering.

Area of Science:

• Cardiac Electrophysiology
• Computational Biology
• Medical Signal Processing

Background:

• Accurate patient-specific ventricular repolarization times (RTs) are crucial for identifying arrhythmogenic cardiac tissue and sudden cardiac death risk.
• Unipolar electrograms (UEs) are used to estimate RTs, with the Wyatt method being a common approach.
• High-pass filtering of UEs, necessary for signal processing, can distort T-wave phase and compromise RT estimation accuracy.

Purpose of the Study:

• To investigate the impact of high-pass filtering on RT estimation from UEs.
• To develop an improved method for estimating RTs from filtered UEs.
• To enhance the reliability of repolarization maps for identifying patient-specific abnormalities.

Main Methods:

• Generated synthetic UEs with known RTs and applied high-pass filtering mimicking CARTO filter settings.
• Trained a deep neural network (DNN) to estimate RT and confidence from filtered synthetic UEs.
• Validated the DNN on filtered ex-vivo human UEs and patient UEs from CARTO, comparing performance against the Wyatt method.

Main Results:

• Even a 2 Hz high-pass filter significantly errors RT estimation using the Wyatt method.
• The DNN outperformed the Wyatt method in 62.75% of cases, demonstrating significantly lower absolute error (p=8.99E-13) with a median of 16.91 ms on ex-vivo UEs.
• The DNN successfully computed an RT map from patient UEs acquired with CARTO.

Conclusions:

• DNNs trained on synthetic data can effectively improve RT estimation from filtered UEs.
• This DNN-based approach offers more reliable repolarization mapping compared to traditional methods.
• Improved repolarization maps aid in the precise identification of patient-specific repolarization abnormalities and associated risks.