A machine learning method integrating ECG and gated SPECT for cardiac resynchronization therapy decision support.

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Cardiac resynchronization therapy (CRT) is a key treatment for heart failure.
• Mechanical dyssynchrony can predict patient response to CRT.
• Predicting CRT responders is crucial for optimizing treatment outcomes.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting CRT response.
• To integrate electrocardiography (ECG), gated SPECT myocardial perfusion imaging (GMPS), and clinical variables.
• To compare ML model performance against traditional guideline-based criteria.

Main Methods:

• A prospective cohort of 153 CRT-eligible patients was analyzed.
• Patients were classified as responders (LVEF increase ≥5%) and super-responders (LVEF increase ≥15%).
• Machine learning models (Prediction Analysis of Microarrays, Naïve Bayes) were developed and compared to guideline variables.

Main Results:

• The ML model (PAM) achieved an AUC of 0.80, outperforming guideline variables (AUC 0.72).
• ML models demonstrated improved sensitivity and specificity for predicting CRT response compared to guideline criteria.
• Neural network models showed a trend towards better prediction, though not statistically significant.

Conclusions:

• Machine learning methods show potential for improved prediction of CRT and super-response.
• Gated SPECT MPI (GMPS) was a critical data source for these predictive parameters.
• Further research is necessary to validate these ML models in larger cohorts.