Machine learning models using symptoms and clinical variables to predict coronary artery disease on coronary

Yangjie Yu1, Weikai Li2, Jiajia Wu3

  • 1Department of Cardiology, Huashan Hospital, Fudan University, Shanghai, China.

Insights

Machine learning models can predict coronary artery disease (CAD) using clinical data, potentially reducing the need for invasive coronary angiography (CAG). These algorithms show promise in excluding or confirming severe CAD, improving patient care.

Area of Science:

  • Cardiovascular Medicine
  • Medical Informatics
  • Artificial Intelligence in Healthcare

Background:

  • Coronary angiography (CAG) is invasive and costly, with many patients undergoing the procedure showing no significant coronary artery disease (CAD).
  • There is a need for non-invasive methods to accurately predict the likelihood of CAD before resorting to CAG.

Purpose of the Study:

  • To develop and evaluate machine learning algorithms for predicting severe CAD using routinely collected clinical variables.
  • To assess the performance of these algorithms in identifying patients who may not require CAG.

Main Methods:

  • A cross-sectional study involving 2602 patients suspected of CAD was conducted, with a training set of 2082 patients and a test set of 520 patients.
  • LASSO regression was used for feature selection, and logistic regression, Support Vector Machine, and XGBoost algorithms were employed to predict severe CAD.
  • Performance metrics including Area Under the Receiver Operating Characteristic Curve (AUC), Positive Predictive Value (PPV), and Negative Predictive Value (NPV) were analyzed.

Main Results:

  • The logistic regression model achieved an AUC of 0.77 in the training set and 0.75 in the test set.
  • Setting the probability threshold below 0.1 screened out 11 patients in the test set with a 90.9% Negative Predictive Value (NPV).
  • A threshold of 0.9 yielded a Positive Predictive Value (PPV) of 97.4%, indicating high confidence in positive predictions.

Conclusions:

  • Machine learning algorithms utilizing data from hospital information systems can effectively assist in the exclusion and confirmation of severe CAD.
  • These predictive models have the potential to significantly reduce the number of unnecessary invasive coronary angiographies, leading to improved patient outcomes and reduced healthcare costs.
Abstract