Validation of a novel automated signal analysis tool for ablation of Wolff-Parkinson-White Syndrome

Abstract

Background: In previous pilot work we demonstrated that a novel automated signal analysis tool could accurately identify successful ablation sites during Wolff-Parkinson-White (WPW) ablation at a single center.

Objective: We sought to validate and refine this signal analysis tool in a larger multi-center cohort of children with WPW.

Methods: A retrospective review was performed of signal data from children with WPW who underwent ablation at two pediatric arrhythmia centers from 2008-2015. All patients with WPW ≤ 21 years who underwent invasive electrophysiology study and ablation with ablation signals available for review were included. Signals were excluded if temperature or power delivery was inadequate or lesion time was < 5 seconds. Ablation lesions were reviewed for each patient. Signals were classified as successful if there was loss of antegrade and retrograde accessory pathway (AP) conduction or unsuccessful if ablation did not eliminate AP conduction. Custom signal analysis software analyzed intracardiac electrograms for amplitudes, high and low frequency components, integrated area, and signal timing components to create a signal score. We validated the previously published signal score threshold 3.1 in this larger, more diverse cohort and explored additional scoring options. Logistic regression with lasso regularization using Youden's index criterion and a cost-benefit criterion to identify thresholds was considered as a refinement to this score.

Results: 347 signals (141 successful, 206 unsuccessful) in 144 pts were analyzed [mean age 13.2 ± 3.9 years, 96 (67%) male, 66 (45%) left sided APs]. The software correctly identified the signals as successful or unsuccessful in 276/347 (80%) at a threshold of 3.1. The performance of other thresholds did not significantly improve the predictive ability. A signal score threshold of 3.1 provided the following diagnostic accuracy for distinguishing a successful from unsuccessful signal: sensitivity 83%, specificity 77%, PPV 71%, NPV 87%.

Conclusions: An automated signal analysis software tool reliably distinguished successful versus unsuccessful ablation electrograms in children with WPW when validated in a large, diverse cohort. Refining the tools using an alternative threshold and statistical method did not improve the original signal score at a threshold of 3.1. This software was effective across two centers and multiple operators and may be an effective tool for ablation of WPW.

Fig 1. Novel signal analysis composite score (signal score).

This figure demonstrates the generation of the composite Signal Score. A.) The high frequency (HF) and low frequency (LF) components of the signal on the distal ablation catheter are filtered and the onset of the LF and peak of the R wave on surface lead I are identified. B.) The area under the curve of the high frequency signal within a window of 50 msec, from 25 msec before to 25 msec after the peak of the LF signal, is calculated as the composite signal score (C.).

Fig 2. Example of successful and unsuccessful signals.

(A) An example of a successful signal with a score 4.57. (B) An example of an unsuccessful signal in the same patient with a score of 2.23.

Fig 3. Box plot of successful vs. unsuccessful ablation signal scores.

The figure demonstrates box plots of the composite signal score for successful and unsuccessful signals, with the threshold value of 3.1 highlighted with the dashed line. The bold line in the middle of the box represents the median, and the box defines the 25^th-75^th percentiles (the interquartile range [IQR]). The lower whisker denotes the 25^th percentile– 1.5*IQR and the upper whisker denotes the 75^th percentile + 1.5*IQR. Observations outside of this range are plotted as dots.

Fig 4. ROC curve of signal analysis software.

A receiver-operator-curve of the signal analysis tool demonstrates an area-under-the-curve of 0.80.

Fig 5. Predictive value of signal analysis tool.

The figure demonstrates the sensitivity, specificity, positive and negative predictive values of the signal analysis tool. (PPV = positive predictive value, NPV = negative predictive ability).

Fig 6. Predictive ability by accessory pathway location.

The figure demonstrates the predictive ability of the signal analysis tool by accessory pathway location on the tricuspid and mitral annuli. The numbers next to the accessory pathway location indicate the percentage of signals in that location that were classified correctly and the numbers in parenthesis indicated the number correctly classified signals over the total number of signals in that location. (LAL = left antero-lateral, LL = left lateral, LPL = left postero-lateral, LPS = left postero-septal, RA = right anterior, RAL = right antero-lateral, RAS = right antero-septal, RMS = right mid-septal, RL = right lateral, RP = right posterior, RPL = right poster-lateral, other locations include middle cardiac vein, CS diverticulum, and right atrial appendage).