Deep learning methods for screening patients' S-ICD implantation eligibility.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiology

Background:

• Subcutaneous Implantable Cardioverter-Defibrillators (S-ICDs) prevent sudden cardiac death from ventricular arrhythmias.
• T Wave Over Sensing (TWOS) is a risk with S-ICDs, leading to inappropriate shocks, often predicted by a high T:R ratio.
• Current 10-second ECG screening for T:R ratio is insufficient due to temporal variations.

Purpose of the Study:

• To develop a novel Convolutional Neural Network (CNN) model for predicting T:R ratios.
• To overcome limitations of short ECG segments in assessing T:R ratio variability.
• To enhance patient selection for S-ICD implantation by providing more reliable T:R ratio analysis.

Main Methods:

• Utilized phase space reconstruction matrices with a CNN model.
• Predicted T:R ratios from 10-second ECG segments without explicit R or T wave detection.
• Developed a tool for automated, long-term patient screening and T:R ratio behavior analysis.

Main Results:

• The CNN model successfully predicts T:R ratios, mitigating TWOS risk.
• The approach allows for extended ECG analysis beyond the standard 10-second window.
• The tool offers a more comprehensive understanding of T:R ratio dynamics.

Conclusions:

• The developed CNN-based tool offers a more reliable method for assessing S-ICD patient eligibility.
• By avoiding explicit wave detection, the model circumvents TWOS issues.
• This AI-driven approach promises improved accuracy and depth in S-ICD screening processes.