Heart attacks, which are among the major causes of death globally, have claimed the lives of numerous people. In order to preserve human lives, early identification of cardiac disease is beneficial. The major test for diagnosing cardiac disorders is the electrocardiogram (ECG), yet many doctors have been unable to diagnose patients’ conditions early using the ECG readings. Numerous machine-learning, deep-learning techniques have been created to forecast cardiac problems sooner for follow-up therapies, but the current techniques do not provide the best outcome. In this study, a hybrid model CNN with LSTM was proposed utilizing 1120 paper-based ECG records from the clinics. The features from the ECG records could be extracted using GLCM, and the noise from the ECG signaling data could be reduced using Median-Filter. In-order to use the PSO, ant colony optimizer (ACO) was used to increase the performance Accuracy and compared the results with particle existing deep learning models and it achieves the accuracy 98.04%,98.08%F1-Score,98.04% precision and 98.01%recall. This method may be quickly and accurately used to classify important issues and is easily adaptable to other biological signals and various DL models.