Using ensemble of ensemble machine learning methods to predict outcomes of cardiac resynchronization

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiology

Background:

• Cardiac resynchronization therapy (CRT) is a treatment for heart failure, but approximately 30% of patients do not respond.
• Identifying CRT responders is crucial for effective treatment and resource allocation.

Purpose of the Study:

• To develop and evaluate a novel machine learning (ML) system for predicting patient response to CRT.
• To compare the predictive performance of the developed system against traditional ML methods and a convolutional neural network (CNN).

Main Methods:

• Retrospective analysis of 1664 patient electronic health records from 2002-2017.
• Development of an ensemble of ensemble (EoE) ML system with supervised and unsupervised layers.
• Comparison of EoE model performance against traditional ML and CNN models trained on ECG waveforms.

Main Results:

• The EoE model demonstrated improved performance with increased feature utilization.
• The EoE model achieved an area under the receiver operating characteristic curve of 0.76 and an F1-score of 0.73.
• A CNN model directly applied to raw ECG waveforms showed less promising results.

Conclusions:

• The proposed CRT risk calculator significantly outperforms clinical guidelines and traditional ML methods in identifying likely CRT responders.
• This tool can enhance heart failure patient care management by identifying high-risk individuals for CRT.