Machine learning-based phenogrouping in heart failure to identify responders to cardiac resynchronization therapy

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Heart failure (HF) is a complex condition with heterogeneous patient populations.
• Cardiac resynchronization therapy (CRT) effectiveness varies among HF patients.
• Predicting CRT response remains a clinical challenge.

Purpose of the Study:

• To develop and validate a machine learning (ML) algorithm for phenogrouping HF patients.
• To identify HF patient subgroups with differential response to CRT.
• To integrate complex echocardiographic and clinical data for improved patient stratification.

Main Methods:

• Utilized unsupervised ML (Multiple Kernel Learning and K-means clustering) on 1106 HF patients from the MADIT-CRT trial.
• Integrated baseline clinical parameters, echocardiographic data (volume, deformation) for phenogrouping.
• Compared CRT-D treatment effects on primary outcomes and volume response across identified phenogroups.

Main Results:

• Identified four distinct HF phenogroups based on integrated data.
• Phenogroups differed significantly in baseline characteristics, biomarkers, and ventricular function.
• Two phenogroups demonstrated a substantially better treatment effect of CRT-D (HR 0.35-0.36, P<0.001) compared to others (interaction P=0.02).

Conclusions:

• Unsupervised ML can create clinically meaningful classifications of heterogeneous HF cohorts.
• Integrating imaging and clinical data aids in identifying patients likely to benefit from CRT.
• This approach may optimize patient selection for CRT and improve therapeutic outcomes.