Leveraging free-text clinical records for heart disease classification through structured feature mapping

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Cardiovascular Disease Research

Background:

• Hypertension and diabetes are primary risk factors for heart disease, a leading cause of death globally.
• Effective management of heart disease relies on early identification and monitoring of these risk factors.
• Machine learning offers potential for automated risk factor detection and personalized preventive strategies.

Purpose of the Study:

• To develop and evaluate a machine learning model for automatic extraction of heart disease risk factors.
• To predict the presence or absence of heart disease using multimodal data sources.
• To enhance early intervention and patient management for cardiovascular diseases.

Main Methods:

• Utilized a combination of unstructured clinical narratives (PrevComp corpus) and structured datasets (UCI heart disease dataset).
• Employed the Light Gradient Boosting Machine (LightGBM) algorithm for classification.
• Developed a model to predict the presence or absence of heart disease based on extracted risk factors.

Main Results:

• The LightGBM classification model achieved a predictive accuracy of 83% for heart disease.
• Demonstrated robust performance in identifying patients at high risk for heart disease.
• The model's accuracy supports its potential for improving clinical decision-making.

Conclusions:

• Machine learning, specifically LightGBM, can effectively identify heart disease risk factors from diverse data.
• The developed model shows promise for accurate patient risk stratification.
• This approach can significantly contribute to personalized medicine and improved cardiovascular outcomes.