This study proposes an integrated deep learning approach for detecting and localizing Myocardial Infarction (MI) in electrocardiogram (ECG) signals. The accurate detection and localization of MI remains a critical challenge in cardiology, particularly due to the limited availability of properly labeled ECG data for certain MI types. The main objective of this research is to develop a robust automated system combining Generative Adversarial Networks (GAN) and Long Short-Term Memory (LSTM) models for improving detection and localization Myocardial Infarction (MI). This research utilizes the PTB-XL dataset, focusing on five classes: Normal (NORM), Anterior Myocardial Infar ction (AMI), Inferior Myocardial Infarction (IMI), Lateral Myocardial Infarction (LMI), and Posterior Myocardial Infarction (PMI). To address significant class imbalance, particularly in LMI and PMI cases, we implement a GAN based data augmentation strategy that successfully increased underrepresented classes while maintaining signal characteristics. Our LSTM model achieved robust performance with an overall accuracy of 83.65%, precision of 83.60%, recall of 83.65%, and F1-score of 83.20%. Notably, the model demonstrated exceptional performance in detecting PMI (F1-score: 99.83%) and LMI (F1 score: 95.88%), while showing moderate performance in distinguishing between AMI and IMI patterns. The results demonstrate the effectiveness of combining GANs for data augmentation with LSTM networks for MI detection and localization, though challenges remain in differentiating certain MI subtypes. This approach shows promise for improving automated cardiac diagnosis, with potential applications in clinical settings pending further validation on diverse datasets.