Machine learning as a new frontier in mitral valve surgical strategy

Area of Science:

• Cardiovascular Surgery
• Medical Informatics
• Machine Learning

Background:

• Mitral valve repair for ischemic mitral regurgitation (MR) is limited by significant rates of recurrent regurgitation and poor patient outcomes.
• Predicting recurrent MR typically relies on complex, subjective echocardiographic and clinical measurements not routinely collected.

Purpose of the Study:

• To develop predictive models for recurrent MR or 1-year mortality in patients undergoing mitral valve repair.
• To leverage machine learning (ML) and the Society of Thoracic Surgeons (STS) database for this prediction.

Main Methods:

• Utilized the STS database, containing routinely collected demographic and clinical parameters.
• Applied machine learning (ML) methodologies to accommodate data collinearity and identify significant predictors.
• Developed three distinct ML models to predict recurrent MR or death.

Main Results:

• The developed ML models demonstrated good predictive performance with an area under the curve (AUC) ranging from 0.72 to 0.75.
• Data-driven analysis identified key predictors of recurrent MR, including three revascularization targets, peripheral vascular disease, and beta-blocker usage.

Conclusions:

• Machine learning offers a robust approach to analyze complex clinical data for predicting outcomes in ischemic MR.
• This data-driven methodology can identify significant predictors, potentially improving patient management and personalized medicine.
• The study highlights the potential of integrating diverse data types (imaging, waveform, genomic) with ML for future advancements in cardiovascular care.