This project compares five deep learning architectures—1D CNN, LSTM, GRU, CNN-LSTM, and Transformer Encoder—for multi-label classification of 12-lead ECG signals using the PTB-XL dataset. ECG diagnostic labels were mapped into five superclasses: NORM, MI, STTC, CD, and HYP. The 1D CNN achieved the strongest overall performance, suggesting that convolutional models can effectively learn local ECG waveform patterns for automated diagnostic classification.