<p>Early and accurate detection of myocardial infarction (MI) is pivotal in enhancing cardiac function, as it can result in a heart attack. Detecting MI arrhythmias from electrocardiogram (ECG) signals presents challenges in accurately extracting patterns, such as Q, R, S and T wave segments. This paper proposes an efficient automated MI detection based on optimized Stockwell transform (OptST), Renyi entropy, dispersion entropy, and extreme learning machine (ELM) algorithm. The OptST is employed to decompose the ECG signal to four frequency band. The renyi and dispersion entropy features are then extracted in each frequency band for training and validation of ELM model to classify the five MI diseases. To identify clinically significant features, a Wilcoxon rank sum test is conducted, and to test the robustness of the proposed method 5-fold cross-validation has been applied. The simulation results have demonstrated that the proposed model has achieved a highest accuracy of 96.86%, specificity of 98.04%, recall of 92.16%, and F-score of 92.18% for sine activation function at hidden number of 300.</p>