CML-Cardio: a cascade machine learning model to predict cardiovascular disease risk as a primary prevention strategy

Area of Science:

• Computational biology
• Medical informatics
• Machine learning applications in healthcare

Background:

• Cardiovascular diseases (CVDs) are a leading cause of global mortality, with projected deaths reaching 23 million annually by 2030.
• Current population awareness strategies for CVD prevention remain largely ineffective.
• Automating risk assessment is crucial for early intervention and improved patient outcomes.

Purpose of the Study:

• To introduce CML-Cardio, a novel cascade machine learning application designed for automated cardiovascular disease risk classification.
• To address the challenge of highly imbalanced data in CVD risk prediction.
• To develop a prototype for primary CVD prevention.

Main Methods:

• Collected data on diabetes, blood pressure, and other risk factors from a private company.
• Developed a two-stage cascade machine learning model.
• Stage 1: Binary classification using logistic regression, SVM, random forest, XGBoost, CatBoost, and multilayer perceptron to identify low vs. attention-needed risk.
• Stage 2: Anomaly detection to analyze intermediate risk patterns in the attention-needed group.

Main Results:

• The initial binary classification stage achieved an average accuracy of 0.86 ± 0.03 and an F-score of 0.85 ± 0.04.
• The complete cascade model demonstrated an average accuracy of 0.80 ± 0.07 and an F-score of 0.70 ± 0.07.
• Feature impact analysis was performed to interpret model predictions.

Conclusions:

• The CML-Cardio tool effectively automates cardiovascular disease risk classification using a cascade machine learning approach.
• The developed prototype serves as a valuable tool for primary CVD prevention strategies.
• This machine learning application shows promise in enhancing early detection and management of cardiovascular risks.