This review summarizes the current practices and performance outcomes in the use of deep learning for EEG classification. It addresses critical questions about which EEG classification tasks have been explored with deep learning, what input formulations have been used for training deep networks, and whether specific deep learning network structures are suitable for specific tasks. A systematic literature review of 90 studies was conducted, analyzing EEG classification using deep learning from Web of Science and PubMed databases. The studies were analyzed based on task type, EEG preprocessing methods, input type, and deep learning architecture. For EEG classification tasks, convolutional neural networks (CNNs), recurrent neural networks (RNNs), and deep belief networks (DBNs) outperformed stacked auto-encoders (SAEs) and multi-layer perceptron neural networks (MLPNNs) in classification accuracy. The tasks that used deep learning fell into five general groups: emotion recognition, motor imagery, mental workload, seizure detection, and sleep scoring. For each task, specific input formulations, major characteristics, and end classifier recommendations were described. The review highlights that CNNs, RNNs, and DBNs are effective for various EEG classification tasks. CNNs are suitable for tasks involving images or spectrograms, RNNs are effective for tasks involving time-series data, and DBNs are suitable for tasks involving calculated features. The review also discusses the importance of input formulation, with signal values, calculated features, and images being common input types. The results show that CNNs, RNNs, and DBNs are effective for different tasks, with CNNs showing the highest accuracy in some cases. The review also discusses the importance of activation functions, with ReLU being the most commonly used. The review concludes that deep learning has the potential to improve EEG classification accuracy and that future research should focus on optimizing input formulations and architecture design for specific tasks. The review provides practical suggestions for selecting hyperparameters and recommends specific architectures for different tasks.This review summarizes the current practices and performance outcomes in the use of deep learning for EEG classification. It addresses critical questions about which EEG classification tasks have been explored with deep learning, what input formulations have been used for training deep networks, and whether specific deep learning network structures are suitable for specific tasks. A systematic literature review of 90 studies was conducted, analyzing EEG classification using deep learning from Web of Science and PubMed databases. The studies were analyzed based on task type, EEG preprocessing methods, input type, and deep learning architecture. For EEG classification tasks, convolutional neural networks (CNNs), recurrent neural networks (RNNs), and deep belief networks (DBNs) outperformed stacked auto-encoders (SAEs) and multi-layer perceptron neural networks (MLPNNs) in classification accuracy. The tasks that used deep learning fell into five general groups: emotion recognition, motor imagery, mental workload, seizure detection, and sleep scoring. For each task, specific input formulations, major characteristics, and end classifier recommendations were described. The review highlights that CNNs, RNNs, and DBNs are effective for various EEG classification tasks. CNNs are suitable for tasks involving images or spectrograms, RNNs are effective for tasks involving time-series data, and DBNs are suitable for tasks involving calculated features. The review also discusses the importance of input formulation, with signal values, calculated features, and images being common input types. The results show that CNNs, RNNs, and DBNs are effective for different tasks, with CNNs showing the highest accuracy in some cases. The review also discusses the importance of activation functions, with ReLU being the most commonly used. The review concludes that deep learning has the potential to improve EEG classification accuracy and that future research should focus on optimizing input formulations and architecture design for specific tasks. The review provides practical suggestions for selecting hyperparameters and recommends specific architectures for different tasks.