Cardiovascular disease (CVD) is one of the leading causes of human deaths. More than 86% of the population is suffering from CVD worldwide. This life-threatening disease can be detected at an early stage by different medical diagnoses including electrocardiogram (ECG), echo test, and stress analysis. ECG is basically the electrical activity of our heart. By proper analysis of ECG signals, early detection of CVD is possible and subsequently with proper medication, we can save many human lives. Recently, various deep learning models are widely used in different machine learning problems due to their higher accuracy than the conventional machine learning approaches. In this chapter, we review the existing different deep learning approaches to analyze ECG signals for the detection of CVD. We first describe different deep learning models like 1D convolutional neural network (CNN), 2D CNN, recurrent neural network, and long short-term Memory in detail. Then, we summarize their applications for the detection of CVD and discuss all the advantages and limitations. Finally, we discuss the developments of other deep learning methods that can be applied for future diagnoses.