Changes in the ST-T segment of the electrocardiogram (ECG) are related to ischemic cardiac events, representing a risk factor for malignant arrhythmias and even associated with an acute myocardial infarction. Early diagnosis is vital for effective interventions and improved patient outcomes. The automatic analysis of ECG from long-term recordings or cardiac stress tests is crucial for detecting such abnormality. Recent advances in machine learning have enhanced traditional diagnostic methods, showing potential for increased detection performance. This study proposes a deep-learning approach to identify ST-T wave changes in the ECG. The convolutional neural networks produced 92.1% accuracy using the European ST-T Database, overcoming the support vector machine and gradient boosting machine performances.