Background: Acute Myocardial Infarction (AMI) remains a major cause of morbidity and mortality worldwide. Early and accurate diagnosis is critical to implementing timely treatment strategies and improving patient outcomes. Traditional ECG interpretation can be time-consuming and prone to human error, highlighting the need for automated diagnostic solutions. This study aims to develop and evaluate a machine learning-based system for the automated detection of AMI using 12-lead Electrocardiogram (ECG) recordings obtained from real-world hospital data. Methods: A observational study was conducted using ECG data collected from hospital. The study focuses on leveraging machine learning algorithms to differentiate between normal and AMI-affected ECGs. The proposed system includes a robust signal preprocessing pipeline capable of extracting clean, one-dimensional ECG signals from images or PDF files. Feature extraction techniques were applied to the preprocessed signals, followed by classification using an ensemble machine learning model. The ensemble integrates Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), Gaussian Naive Bayes (GNB), and Logistic Regression. Performance metrics, including accuracy, sensitivity, and specificity, were calculated to evaluate the model. Results: The ensemble model achieved an average classification accuracy of 80%. Notably, the model demonstrated high sensitivity in detecting myocardial infarction, achieving a sensitivity rate of 96%, while also effectively identifying normal ECG cases. Conclusion: The findings suggest that machine learning can significantly enhance the rapid and accurate diagnosis of AMI, supporting clinical decision-making and potentially improving patient outcomes. The proposed system demonstrates promise for integration into real-world healthcare settings, offering automated and reliable ECG analysis.