Use of an artificial neural network to localize accessory pathways of Wolff–Parkinson–White syndrome with12-lead electrocardiogram

Abstract

Today, radio-frequency ablation has been shown to be a safe and effective method to treat paroxysmal tachycardia with Wolff–Parkinson–White syndrome. The many criteria reported for localizing the sites of accessory pathways from a 12-lead electrocardiogram have not proven adequate to differentiate the correct sites of accessory pathways for all situations. This study trained an artificial neural network to differentiate the varied features needed to localize 10 sites of accessory pathways. One hundred fifty patients underwent successful catheter ablation, with manifest single and antegradely conducting accessory pathways. Using the two electrocardiogram features of polarity of delta wave and R wave's share of QRS complex, an artificial neural network learned the characteristics of electrocardiogram waves for each site of the 10 accessory pathways through 90 learning cases, and an applicable network model was developed for testing. In 58 of 60 test cases (96.7%), sites of accessory pathways were localized correctly by the network. Based on the method employed in the present study, it thus becomes possible to predict the sites of accessory pathways with Wolff–Parkinson–White syndrome in more detail by using an artificial neural network with a 12-lead electrocardiogram. In the future, when this method is incorporated into a conventional automatic electrocardiogram system which could analyze delta waves and ORS complex, it will become useful to automatically diagnose the locations of the accessory pathways with Wolff–Parkinson–White syndrome in clinical practice.

Keywords:

• Artificial neural network
• accessory pathways
• Wolff–Parkinson–White syndrome
• electrocardiogram

Acknowledgment

This study was supported by The Hori Information Science Promotion Foundation in Japan.