Multi Classification of ECG Signals based on Convolution and Residual Neural Networks for Effective Forecasting of Cardiovascular Disease

Cardiovascular disease (CVD) is a major contributor to death rates around the globe. Early diagnosis of cardiac illnesses is critical for efficient therapy, and an electrocardiogram (ECG) is vital for identification. Deep learning techniques have made significant advancements in the classification of ECG signals, achieving a level of accuracy comparable to that of cardiologists. In a medical situation, a cardiologist uses conventional 12-lead ECG data to determine a diagnosis. This work describes a multi-class classifier that can identify five distinct forms of cardiovascular illnesses: NORM, hypertrophy, myocardial infarction, ST-T abnormalities, and conduction disturbances. The model utilizes the abundant input from the usual 12-lead ECG data and acquires knowledge of patterns at the beat, and rhythm, and uses the dataset PTB-XL. Convolution with residual connection with Bi-LSTM performance at different filter sizes, when the filter size increases, we get a little more improvement in the model performance. The comparative study of the performance of 3 different classifiers LSTM, RNN, and Bi-LSTM, evaluates the performance of Accuracy, f1-score, recall, and precision. The performance at CRDM at filter size 11 accuracy, precision, F1-score &recall is 95.34%,91.54%,93.67% and 92.99%.