Machine learning techniques in cardiac risk assessment.

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Cardiac surgery carries significant mortality risk.
• Accurate prediction of this risk is crucial for patient management.
• Machine learning offers potential for improved risk prediction models.

Purpose of the Study:

• To predict patient mortality risk during or shortly after cardiac surgery.
• To leverage machine learning for enhanced predictive accuracy.
• To identify key risk factors influencing post-cardiac surgery outcomes.

Main Methods:

• Utilized a dataset from Acıbadem Maslak Hospital.
• Employed European System for Cardiac Operative Risk Evaluation (EuroSCORE) factors.
• Developed predictive models using five machine learning algorithms on numeric and categorical datasets.
• Evaluated model performance via 10-fold cross-validation.

Main Results:

• The C4.5 decision tree algorithm achieved the highest accuracy (0.989) on the numeric dataset.
• Pulmonary hypertension, recent myocardial infarction, and thoracic aorta surgery were identified as primary mortality risk factors.
• A dynamic web application was developed based on the best-performing model.

Conclusions:

• The C4.5 model demonstrated superior performance in predicting cardiac surgery mortality.
• Using numerical risk factor data enhances machine learning model performance.
• Hospital-specific assessment systems utilizing local data are beneficial for clinical decision-making.