Predicting the response to cardiac resynchronization therapy (CRT) using the deep learning approach

Abstract

Cardiac Resynchronization Therapy Defibrillator (CRT-D) is a method to improve heart rate variability and arrhythmia-related symptoms in heart failure patients. According to clinical reports, CRT is not entirely safe and risk-free like other surgery. It can reduce heart failure risks, shorten hospital stays, and enhance the patients' quality of life. The present study aims to perform the proper selection of patients before surgery to avoid potential costs. This article focuses on the data collection of heart failure patients' activities, the process of features effective extraction, and identifying an optimal pattern using a Deep Learning (DL) algorithm. Also, the main tasks of the proposed methods include the use of qualitative indicators for initial feature extraction, oversampling from minority class, Density-Based Spatial Clustering of Applications with Noise (DBSCAN) hierarchical clustering, selecting features from Low error clusters, selecting samples from high error clusters, and classification using customized DL configuration. The research data collection consisted of 209 patients with 60 demographic, clinical, laboratory, ECG, and echo features. In addition, features were analyzed based on their significance in predicting CRT response status. The DL algorithm, which used dense layers and convolution for its architecture, was employed to heart failure patients optimally identify the treatment status. The proposed method predicted the response to cardiac resynchronization therapy with an error rate of 91.85% and an Area Under Curve (AUC) of 0.957 and a sensitivity of 94.22%.