Identification of the Optimal Artificial Intelligence Algorithm for predicting Myocardial Infarction: A Review

Background and aims: Over the past decade, there has been a dramatic increase in cardiovasculardisease incidents. A great deal of previous research into cardiovascular disease has focused onMyocardial Infarction (MI). It can lead to malignant arrhythmia, Heart Failure (HF), and suddendeath. Several studies reveal that AI in medicine can enhance modeling, prediction, treatment, anddiagnosis. Achieving higher accuracy in HF population prediction models is a major problem thathas not been reached with traditional predictive models. Consequently, predictive models basedon machine learning techniques that were externally validated could have tremendous help andbe an alternative. This study aims to present the most accurate algorithm in MI prediction models.Method: To identify relevant studies for this analysis, a comprehensive search strategy was developedusing the PubMed database. Exclusion criteria were applied to remove studies that focusedon mortality, systematic reviews, reviews, and protocols. The inclusion criteria were studies thatfocused on MI disease, were written in English, and included inpatients from the last decade upto ۱۷th November ۲۰۲۲. To ensure consistency in the data collected, vital signs and discrete physiologicalmeasurements of MI patients with no previous history of heart failure were monitored.Prediction decompensation was then used to highlight important features from a combination ofmodel inputs from dissimilar data. This approach aimed to identify the most accurate algorithmfor MI prediction models and contribute to the development of effective machine-learning techniquesfor cardiovascular disease management.Results: A considerable amount of literature have been published about AI algorithms in medicineprediction models. Generally, ۴۲۴ papers were included in the study, and we have investigated ۳۱full texts. Ultimately, inclusion criteria resulted in ۹ articles. The most striking result to emergefrom the included studies is that the most used algorithms in order are Random Forest (RF) andLogistic Regression (LR). Additionally, the most accurate algorithm was Deep Neural Networkwhich has above ۹۰% accuracy. Moreover, among those used several algorithms simultaneously,Deep Neural Networks had a better Power.Conclusion: Our findings reveal that the most commonly used algorithms in medical predictionmodels are Random Forest and Logistic Regression. However, the most accurate algorithm isthe Deep Neural Network, which has demonstrated an accuracy rate above ۹۰%. Furthermore,when multiple algorithms are used simultaneously, the Deep Neural Network shows better power.These results suggest that AI algorithms have great potential in improving medical predictionmodels, and further research is needed to explore their full capabilities.

کلیدواژه ها:

Artificial Intelligence ، Machine Learning ، Deep Learning ، Neural Network ، Myocardial Infarction ، Heart Failure