CVAR-Seg: An Automated Signal Segmentation Pipeline for Conduction Velocity and Amplitude Restitution.

Area of Science:

• Cardiovascular Electrophysiology
• Biomedical Signal Processing

Background:

• S1S2 stimulation protocols are crucial for atrial fibrillation risk assessment.
• Clinical studies generate large datasets requiring automated analysis.
• Existing methods struggle with noise and proximity of stimulation artifacts to atrial activity.

Purpose of the Study:

• To develop an automated pipeline (CVAR-Seg) for analyzing S1S2 stimulation data.
• To address challenges of arbitrary protocols, noise robustness, and artifact proximity.
• To enable precise evaluation of local activation times, electrogram amplitude, and conduction velocity.

Main Methods:

• Time interval detection for S1 basic cycle length estimation.
• Amplitude thresholding and peak detection for stimulus segmentation.
• Matched filtering to eliminate stimulation artifacts.
• Non-linear signal energy operator for atrial activity segmentation.
• Geodesic and Euclidean distances for conduction velocity approximation.

Main Results:

• CVAR-Seg achieved median local activation time error <1 ms at signal-to-noise ratios as low as 0 dB.
• The pipeline demonstrated robustness against clinically relevant noise levels.
• Proof-of-concept testing on a paroxysmal atrial fibrillation patient yielded plausible restitution curves.

Conclusions:

• The CVAR-Seg pipeline offers fast, automated, robust, and accurate atrial signal evaluation.
• It overcomes challenges of artifact proximity, enabling standardized analysis of large datasets.
• Open availability of CVAR-Seg facilitates research and clinical application without human bias.