The use of artificial intelligence and deep learning methods in subcutaneous implantable cardioverter defibrillator screening to optimise selection in special patient populations

Elrefai, M., Abouelasaad, M, Conibear, I, Wiles, B, Dunn, A, Coniglio, S, Zemkoho, A and Roberts, P (2022) The use of artificial intelligence and deep learning methods in subcutaneous implantable cardioverter defibrillator screening to optimise selection in special patient populations. EP Europace, 24 (Supplement_1). (doi:10.1093/europace/euac053.448).

Abstract

Funding Acknowledgements:
Type of funding sources: Private company. Main funding source(s): Dr.Mohamed ElRefai is receiving an unrestricted grant from Boston Scientific.
Introduction: Adult congenital heart disease (ACHD) and hypertrophic cardiomyopathy (HCM) patients who require defibrillator therapy are often relatively young and may require several generator replacements in their lifetime. The increased risk of complications associated with transvenous ICDs make the subcutaneous (S-ICD) a valuable alternative. However, higher S-ICD ineligibility rates (20-40% in ACHD and 7-38% in HCM) and higher inappropriate shock rates (10.5% in ACHD and 12.5% in HCM) are observed in these populations. Unfavourable T:R ratios and dynamic changes in the R and T wave amplitudes are the primarily factors behind ineligibility and inappropriate shocks, which are most commonly caused by T wave over-sensing.
Purpose: We report a novel application of deep learning methods used to autonomously screen patients for S-ICD eligibility over a longer period than conventional screening. We hypothesise that this screening approach might achieve better patient selection and optimise S-ICD vector selection in challenging patient cohorts.
Methods: Adult patients with ACHD or HCM and a control group of normal subjects were fitted with a 24-hour ambulatory ECG with the leads placed to record their S-ICD vectors. T: R ratio throughout the recordings was analysed utilising phase space reconstruction matrices to convert the ECG signal into compressed pixel images. Whilst a convolutional neural network model was trained to provide an in-depth description of the T: R variation plot for each vector T: R variation was compared statistically using a one-way ANOVA test.
Results: 20 patients (age 44.1 ±11.68, 60% male, 7 HCM, 6 ACHD and 7 control subjects) were enrolled. A significant difference was observed in the mean and median T: R values between the three groups (p<0.001). There was no difference observed in the standard deviation of T: R between the control subjects and HCM group. However, there was a statistically significant difference in the standard deviation of T: R between the control subjects and the ACHD group (p= 0.01). [see Figure].
Conclusions: T:R ratio, a main determinant for S-ICD eligibility, is significantly higher in ACHD and HCM when compared to normal hearts and it also has more tendency to fluctuate in ACHD patients when compared to HCM and normal hearts populations. We hypothesise that our novel model could be used to select S-ICD eligible patients by better characterisation of T:R ratio reducing the risk of T wave oversensing and inappropriate shocks particularly in the ACHD patients’ cohort

This record has no associated files available for download.

Contributors

Author: S Coniglio

Author: A Zemkoho

Download statistics