Advancements in Predictive Analytics for Early Detection of Myocardial Infarction in High-Risk Populations

Early Detection of Myocardial Infarction Using Predictive Analytics

Abstract

Background: The ability to diagnose myocardial infarction at a relatively early stage is considered to be highly important with the aim of increasing patient survival rate, especially in conditions of increased risk. The existing risk assessment models are somewhat misleading in terms of predictive validity.

Objective: This study evaluates the ability of the various machine learning models in the prediction of MI with especial emphasis on the neural networks’ performance as well as the comparison with the other traditional and other types of sophisticated models.

Methods: A cross-sectional study was conducted to collect the data from Electronic Health Records (EHRs) of the high-risk patients. The machine learning algorithms chosen were logistic regression, random forest, gradient boosting machines, and neural networks. The outcomes were measured in terms of accuracy, sensitivity, specificity and Area under the curve of the receiver operating characteristic (AUC-ROC).

Results: Thus, as it can be seen, the neural networks proved to be the most effective as they have the highest accuracy (91.3%) and AUC-ROC of (0.95). These models have shown better predictive accuracy compared with other techniques and most of the other forms of the machine learning models.

Conclusion: Neural networks improve the early identification of MI in high-risk groups adding potential betterment in clinical prognosis. These new models are still not implemented in everyday clinical practice, and their integration could revolutionise patient management, providing earlier and better-targeted treatments.