<p>Electrocardiogram signals play a crucial role in the early prevention and diagnosis of cardiovascular diseases. ECG signals can be used for the diagnosis of various diseases, but their pathological features differ only slightly, which poses a challenge for automatic classification models. Moreover, signal processing methods often have difficulty extracting hidden features. Therefore, researchers are turning to convolutional neural networks to solve the feature extraction problem. However, CNNs cannot fully capture the pathological differences between different diseases and often lack temporal information. In response to these challenges, we introduce TECC (Transformer-enabled ECG Classification), a novel network that incorporates a transformer encoder. TECC aims to improve the extraction of pathological features and temporal data. The proposed model utilises the multichannel information from standard 12-lead ECG recordings and learns patterns at beat, rhythm, and channel levels. In addition, we propose a weighted fusion method to further improve the performance of the model. The experimental results show that our model achieves the highest performance on five classification tasks on the PTB-XL dataset, with a macro-averaged ROC-AUC value of 92.61%, an average accuracy of 89.26%, and a maximum F1 score of 81.28%. The outstanding performance of the TECC model in classifying various diseases emphasises its potential to assist heart disease experts in detecting various diseases and provide a way to explore further potential information contained in ECG signals.</p>