Correlation analysis of deep learning methods in S-ICD screening.

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Subcutaneous implantable cardiac defibrillator (S-ICD) eligibility can vary due to dynamic ECG signals.
• Current S-ICD screening methods face practical limitations in acquiring sufficient ECG data duration.
• This study investigates deep learning for improved S-ICD screening.

Purpose of the Study:

• To explore the efficacy of deep learning methods in S-ICD screening.
• To assess the potential of deep learning to overcome limitations in ECG data acquisition for S-ICD eligibility.
• To introduce and evaluate a novel concept, favorable ratio time (FVR), in S-ICD vector analysis.

Main Methods:

• A retrospective study utilizing a deep learning tool for descriptive analysis of T:R ratios over 24-hour ECG recordings.
• Analysis of 28 vectors from 14 patients.
• Spearman's rank correlation test to compare deep learning outcomes with a gold standard S-ICD simulator.

Main Results:

• The study analyzed data from 14 patients (mean age 63.7 years, 71.4% male).
• Key metrics included mean T:R (0.21 ± 0.11), standard deviation of T:R (0.08 ± 0.04), and favorable ratio time (FVR) (79% ± 30%).
• Statistically significant strong correlations (p < .001) were found between the deep learning tool's outcomes and the S-ICD simulator.

Conclusions:

• Deep learning presents a practical software solution for analyzing extended ECG data, surpassing current S-ICD screening limitations.
• This approach can enhance patient selection for S-ICD therapy and guide vector selection in eligible patients.
• Further research is required for clinical translation of this deep learning methodology.